Merge pull request #340 from jschanck/saber

Update Saber and add AVX2 implementation

crypto_kem/firesaber/META.yml

@@ -14,4 +14,13 @@ principal-submitters:
   - Frederik Vercauteren
 implementations:
     - name: clean
-      version: https://github.com/KULeuven-COSIC/SABER/commit/14ede83f1ff3bcc41f0464543542366c68b55871
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/1ae84c3c/saber
+    - name: avx2
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/1ae84c3c/saber
+      supported_platforms:
+          - architecture: x86_64
+            operating_systems:
+                - Linux
+                - Darwin
+            required_flags:
+                - avx2

crypto_kem/firesaber/avx2/LICENSE

@@ -0,0 +1 @@
+Public Domain

crypto_kem/firesaber/avx2/Makefile

@@ -0,0 +1,22 @@
+# This Makefile can be used with GNU Make or BSD Make
+
+LIB=libfiresaber_avx2.a
+HEADERS=api.h cbd.h pack_unpack.h poly.h SABER_indcpa.h SABER_params.h verify.h 
+OBJECTS=cbd.o kem.o pack_unpack.o poly.o poly_mul.o SABER_indcpa.o verify.o 
+
+CFLAGS=-O3 -mavx2 -Wall -Wextra -Wpedantic -Wvla -Werror -Wredundant-decls -Wmissing-prototypes -std=c99 -I../../../common $(EXTRAFLAGS)
+
+all: $(LIB)
+
+%.o: %.s $(HEADERS)
+	$(AS) -o $@ $<
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_kem/firesaber/avx2/SABER_indcpa.c

@@ -0,0 +1,125 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "fips202.h"
+#include "pack_unpack.h"
+#include "poly.h"
+#include "randombytes.h"
+#include <stdint.h>
+#include <string.h>
+
+#define h1 (1 << (SABER_EQ - SABER_EP - 1))
+#define h2 ((1 << (SABER_EP - 2)) - (1 << (SABER_EP - SABER_ET - 1)) + (1 << (SABER_EQ - SABER_EP - 1)))
+
+void PQCLEAN_FIRESABER_AVX2_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]) {
+    size_t i, j;
+
+    poly A[SABER_L][SABER_L];
+    poly *skpv1 = A[0]; // use first row of A to hold sk temporarily
+    toom4_points skpv1_eval[SABER_L];
+    poly res[SABER_L];
+
+    uint8_t rand[SABER_NOISESEEDBYTES];
+    uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+
+    randombytes(seed_A, SABER_SEEDBYTES);
+    shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
+
+    randombytes(rand, SABER_NOISESEEDBYTES);
+    PQCLEAN_FIRESABER_AVX2_GenSecret(skpv1, rand);
+    PQCLEAN_FIRESABER_AVX2_POLVECq2BS(sk, skpv1); // pack secret key
+
+    for (j = 0; j < SABER_L; j++) {
+        PQCLEAN_FIRESABER_AVX2_toom4_eval(&skpv1_eval[j], &skpv1[j]);
+    }
+
+    PQCLEAN_FIRESABER_AVX2_GenMatrix(A, seed_A); // sample matrix A
+    PQCLEAN_FIRESABER_AVX2_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const toom4_points *)skpv1_eval, 1); // Matrix in transposed order
+
+    // rounding
+    for (i = 0; i < SABER_L; i++) {
+        for (j = 0; j < SABER_N; j++) {
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
+        }
+    }
+
+    PQCLEAN_FIRESABER_AVX2_POLVECp2BS(pk, res); // pack public key
+}
+
+
+void PQCLEAN_FIRESABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]) {
+    size_t i, j;
+
+    poly A[SABER_L][SABER_L];
+    poly res[SABER_L];
+    toom4_points skpv1_eval[SABER_L];
+
+    poly *temp = A[0]; // re-use stack space
+    poly *vprime = &A[0][0];
+    poly *message = &A[0][1];
+
+    const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+    uint8_t *msk_c = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+
+    PQCLEAN_FIRESABER_AVX2_GenSecret(temp, noiseseed);
+    for (j = 0; j < SABER_L; j++) {
+        PQCLEAN_FIRESABER_AVX2_toom4_eval(&skpv1_eval[j], &temp[j]);
+    }
+
+    PQCLEAN_FIRESABER_AVX2_GenMatrix(A, seed_A);
+    PQCLEAN_FIRESABER_AVX2_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const toom4_points *)skpv1_eval, 0); // 0 => not transposed
+
+    // rounding
+    for (i = 0; i < SABER_L; i++) { //shift right EQ-EP bits
+        for (j = 0; j < SABER_N; j++) {
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
+        }
+    }
+    PQCLEAN_FIRESABER_AVX2_POLVECp2BS(ciphertext, res);
+
+    // vector-vector scalar multiplication with mod p
+    PQCLEAN_FIRESABER_AVX2_BS2POLVECp(temp, pk);
+    PQCLEAN_FIRESABER_AVX2_InnerProd(vprime, temp, skpv1_eval);
+    PQCLEAN_FIRESABER_AVX2_BS2POLmsg(message, m);
+
+    for (i = 0; i < SABER_N; i++) {
+        vprime->coeffs[i] += h1 - (message->coeffs[i] << (SABER_EP - 1));
+        vprime->coeffs[i] &= SABER_P - 1;
+        vprime->coeffs[i] >>= SABER_EP - SABER_ET;
+    }
+
+    PQCLEAN_FIRESABER_AVX2_POLT2BS(msk_c, vprime);
+}
+
+
+void PQCLEAN_FIRESABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]) {
+    size_t i;
+
+    poly temp[SABER_L];
+    toom4_points sksv_eval[SABER_L];
+
+    const uint8_t *packed_cm = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+    poly *v = &temp[0];
+    poly *cm = &temp[1];
+
+    PQCLEAN_FIRESABER_AVX2_BS2POLVECq(temp, sk);
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_FIRESABER_AVX2_toom4_eval(&sksv_eval[i], &temp[i]);
+    }
+
+    PQCLEAN_FIRESABER_AVX2_BS2POLVECp(temp, ciphertext);
+    PQCLEAN_FIRESABER_AVX2_InnerProd(v, temp, sksv_eval);
+
+    PQCLEAN_FIRESABER_AVX2_BS2POLT(cm, packed_cm);
+
+    for (i = 0; i < SABER_N; i++) {
+        v->coeffs[i] += h2 - (cm->coeffs[i] << (SABER_EP - SABER_ET));
+        v->coeffs[i] &= SABER_P - 1;
+        v->coeffs[i] >>= SABER_EP - 1;
+    }
+
+    PQCLEAN_FIRESABER_AVX2_POLmsg2BS(m, v);
+}

crypto_kem/firesaber/avx2/SABER_indcpa.h

@@ -0,0 +1,13 @@
+#ifndef INDCPA_H
+#define INDCPA_H
+#include "SABER_params.h"
+#include <stdint.h>
+
+void PQCLEAN_FIRESABER_AVX2_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]);
+
+void PQCLEAN_FIRESABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]);
+
+void PQCLEAN_FIRESABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]);
+
+
+#endif

crypto_kem/firesaber/avx2/SABER_params.h

@@ -0,0 +1,41 @@
+#ifndef PARAMS_H
+#define PARAMS_H
+
+
+/* Don't change anything below this line */
+#define SABER_L 4
+#define SABER_MU 6
+#define SABER_ET 6
+
+#define SABER_N 256
+
+#define SABER_EP 10
+#define SABER_P (1 << SABER_EP)
+
+#define SABER_EQ 13
+#define SABER_Q (1 << SABER_EQ)
+
+#define SABER_SEEDBYTES 32
+#define SABER_NOISESEEDBYTES 32
+#define SABER_KEYBYTES 32
+#define SABER_HASHBYTES 32
+
+#define SABER_POLYCOINBYTES (SABER_MU * SABER_N / 8)
+
+#define SABER_POLYBYTES (SABER_EQ * SABER_N / 8)
+#define SABER_POLYVECBYTES (SABER_L * SABER_POLYBYTES)
+
+#define SABER_POLYCOMPRESSEDBYTES (SABER_EP * SABER_N / 8)
+#define SABER_POLYVECCOMPRESSEDBYTES (SABER_L * SABER_POLYCOMPRESSEDBYTES)
+
+#define SABER_SCALEBYTES_KEM (SABER_ET * SABER_N / 8)
+
+#define SABER_INDCPA_PUBLICKEYBYTES (SABER_POLYVECCOMPRESSEDBYTES + SABER_SEEDBYTES)
+#define SABER_INDCPA_SECRETKEYBYTES (SABER_POLYVECBYTES)
+
+#define SABER_PUBLICKEYBYTES (SABER_INDCPA_PUBLICKEYBYTES)
+#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES + SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
+
+#define SABER_BYTES_CCA_DEC (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM)
+
+#endif

crypto_kem/firesaber/avx2/api.h

@@ -0,0 +1,18 @@
+#ifndef PQCLEAN_FIRESABER_AVX2_API_H
+#define PQCLEAN_FIRESABER_AVX2_API_H
+
+
+#define PQCLEAN_FIRESABER_AVX2_CRYPTO_ALGNAME "FireSaber"
+#define PQCLEAN_FIRESABER_AVX2_CRYPTO_BYTES 32
+#define PQCLEAN_FIRESABER_AVX2_CRYPTO_CIPHERTEXTBYTES 1472
+#define PQCLEAN_FIRESABER_AVX2_CRYPTO_PUBLICKEYBYTES 1312
+#define PQCLEAN_FIRESABER_AVX2_CRYPTO_SECRETKEYBYTES 3040
+
+int PQCLEAN_FIRESABER_AVX2_crypto_kem_keypair(unsigned char *pk, unsigned char *sk);
+
+int PQCLEAN_FIRESABER_AVX2_crypto_kem_enc(unsigned char *ct, unsigned char *k, const unsigned char *pk);
+
+int PQCLEAN_FIRESABER_AVX2_crypto_kem_dec(unsigned char *k, const unsigned char *ct, const unsigned char *sk);
+
+
+#endif /* PQCLEAN_FIRESABER_AVX2_API_H */

crypto_kem/firesaber/avx2/cbd.c

@@ -0,0 +1,48 @@
+#include "SABER_params.h"
+#include "api.h"
+#include "cbd.h"
+#include <stdint.h>
+/*---------------------------------------------------------------------
+This file has been adapted from the implementation
+(available at, Public Domain https://github.com/pq-crystals/kyber)
+of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------------------------*/
+
+
+static uint64_t load_littleendian(const uint8_t *x, int bytes) {
+    int i;
+    uint64_t r = x[0];
+    for (i = 1; i < bytes; i++) {
+        r |= (uint64_t)x[i] << (8 * i);
+    }
+    return r;
+}
+
+void PQCLEAN_FIRESABER_AVX2_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]) {
+    uint32_t t, d, a[4], b[4];
+    int i, j;
+
+    for (i = 0; i < SABER_N / 4; i++) {
+        t = (uint32_t) load_littleendian(buf + 3 * i, 3);
+        d = 0;
+        for (j = 0; j < 3; j++) {
+            d += (t >> j) & 0x249249;
+        }
+
+        a[0] = d & 0x7;
+        b[0] = (d >> 3) & 0x7;
+        a[1] = (d >> 6) & 0x7;
+        b[1] = (d >> 9) & 0x7;
+        a[2] = (d >> 12) & 0x7;
+        b[2] = (d >> 15) & 0x7;
+        a[3] = (d >> 18) & 0x7;
+        b[3] = (d >> 21);
+
+        s[4 * i + 0] = (uint16_t)(a[0] - b[0]);
+        s[4 * i + 1] = (uint16_t)(a[1] - b[1]);
+        s[4 * i + 2] = (uint16_t)(a[2] - b[2]);
+        s[4 * i + 3] = (uint16_t)(a[3] - b[3]);
+    }
+}

crypto_kem/firesaber/avx2/cbd.h

@@ -0,0 +1,16 @@
+#ifndef CBD_H
+#define CBD_H
+/*---------------------------------------------------------------------
+This file has been adapted from the implementation
+(available at, Public Domain https://github.com/pq-crystals/kyber)
+of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------------------------*/
+#include "SABER_params.h"
+#include <stdint.h>
+
+void PQCLEAN_FIRESABER_AVX2_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]);
+
+
+#endif

crypto_kem/firesaber/avx2/kem.c

@@ -0,0 +1,77 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "api.h"
+#include "fips202.h"
+#include "randombytes.h"
+#include "verify.h"
+#include <stddef.h>
+#include <stdint.h>
+
+
+int PQCLEAN_FIRESABER_AVX2_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
+    size_t i;
+
+    PQCLEAN_FIRESABER_AVX2_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
+    for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
+        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
+    }
+
+    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES); // Then hash(pk) is appended.
+
+    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES); // Remaining part of sk contains a pseudo-random number.
+    // This is output when check in PQCLEAN_FIRESABER_AVX2_crypto_kem_dec() fails.
+    return (0);
+}
+
+int PQCLEAN_FIRESABER_AVX2_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t *pk) {
+
+    uint8_t kr[64]; // Will contain key, coins
+    uint8_t buf[64];
+
+    randombytes(buf, 32);
+
+    sha3_256(buf, buf, 32); // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
+
+    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES); // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
+
+    sha3_512(kr, buf, 64);               // kr[0:63] <-- Hash(buf[0:63]);
+    // K^ <-- kr[0:31]
+    // noiseseed (r) <-- kr[32:63];
+    PQCLEAN_FIRESABER_AVX2_indcpa_kem_enc(c, buf, kr + 32, pk); // buf[0:31] contains message; kr[32:63] contains randomness r;
+
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC);
+
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}
+
+int PQCLEAN_FIRESABER_AVX2_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
+    size_t i;
+    uint8_t fail;
+    uint8_t cmp[SABER_BYTES_CCA_DEC];
+    uint8_t buf[64];
+    uint8_t kr[64]; // Will contain key, coins
+    const uint8_t *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
+
+    PQCLEAN_FIRESABER_AVX2_indcpa_kem_dec(buf, sk, c); // buf[0:31] <-- message
+
+    // Multitarget countermeasure for coins + contributory KEM
+    for (i = 0; i < 32; i++) { // Save hash by storing h(pk) in sk
+        buf[32 + i] = sk[SABER_SECRETKEYBYTES - 64 + i];
+    }
+
+    sha3_512(kr, buf, 64);
+
+    PQCLEAN_FIRESABER_AVX2_indcpa_kem_enc(cmp, buf, kr + 32, pk);
+
+    fail = PQCLEAN_FIRESABER_AVX2_verify(c, cmp, SABER_BYTES_CCA_DEC);
+
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC); // overwrite coins in kr with h(c)
+
+    PQCLEAN_FIRESABER_AVX2_cmov(kr, sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES, fail);
+
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}

crypto_kem/firesaber/avx2/pack_unpack.c

@@ -0,0 +1,151 @@
+#include "SABER_params.h"
+#include "pack_unpack.h"
+#include "poly.h"
+#include <string.h>
+
+void PQCLEAN_FIRESABER_AVX2_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = (uint8_t) ((in[0] & 0x3f) | (in[1] << 6));
+        out[1] = (uint8_t) (((in[1] >> 2) & 0x0f) | (in[2] << 4));
+        out[2] = (uint8_t) (((in[2] >> 4) & 0x03) | (in[3] << 2));
+        in += 4;
+        out += 3;
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]) {
+    /* This function does not reduce its output mod T */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = in[0];
+        out[1] = (in[0] >> 6) | (in[1] << 2);
+        out[2] = (in[1] >> 4) | (in[2] << 4);
+        out[3] = (in[2] >> 2);
+        in += 3;
+        out += 4;
+    }
+}
+
+static void POLq2BS(uint8_t bytes[SABER_POLYBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x1f) | (in[1] << 5));
+        out[2] = (uint8_t) (in[1] >> 3);
+        out[3] = (uint8_t) (((in[1] >> 11) & 0x03) | (in[2] << 2));
+        out[4] = (uint8_t) (((in[2] >> 6) & 0x7f) | (in[3] << 7));
+        out[5] = (uint8_t) (in[3] >> 1);
+        out[6] = (uint8_t) (((in[3] >> 9) & 0x0f) | (in[4] << 4));
+        out[7] = (uint8_t) (in[4] >> 4);
+        out[8] = (uint8_t) (((in[4] >> 12) & 0x01) | (in[5] << 1));
+        out[9] = (uint8_t) (((in[5] >> 7) & 0x3f) | (in[6] << 6));
+        out[10] = (uint8_t) (in[6] >> 2);
+        out[11] = (uint8_t) (((in[6] >> 10) & 0x07) | (in[7] << 3));
+        out[12] = (uint8_t) (in[7] >> 5);
+        in += 8;
+        out += 13;
+    }
+}
+
+static void BS2POLq(poly *data, const uint8_t bytes[SABER_POLYBYTES]) {
+    /* This function does not reduce its output mod Q */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (in[0]) | (in[1] << 8);
+        out[1] = (in[1] >> 5) | (in[2] << 3) | (in[3] << 11);
+        out[2] = (in[3] >> 2) | (in[4] << 6);
+        out[3] = (in[4] >> 7) | (in[5] << 1) | (in[6] << 9);
+        out[4] = (in[6] >> 4) | (in[7] << 4) | (in[8] << 12);
+        out[5] = (in[8] >> 1) | (in[9] << 7);
+        out[6] = (in[9] >> 6) | (in[10] << 2) | (in[11] << 10);
+        out[7] = (in[11] >> 3) | (in[12] << 5);
+        in += 13;
+        out += 8;
+    }
+}
+
+static void POLp2BS(uint8_t bytes[SABER_POLYCOMPRESSEDBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x03) | (in[1] << 2));
+        out[2] = (uint8_t) (((in[1] >> 6) & 0x0f) | (in[2] << 4));
+        out[3] = (uint8_t) (((in[2] >> 4) & 0x3f) | (in[3] << 6));
+        out[4] = (uint8_t) (in[3] >> 2);
+        in += 4;
+        out += 5;
+    }
+}
+
+static void BS2POLp(poly *data, const uint8_t bytes[SABER_POLYCOMPRESSEDBYTES]) {
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = in[0] | (in[1] << 8);
+        out[1] = (in[1] >> 2) | (in[2] << 6);
+        out[2] = (in[2] >> 4) | (in[3] << 4);
+        out[3] = (in[3] >> 6) | (in[4] << 2);
+        in += 5;
+        out += 4;
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLq2BS(bytes + i * SABER_POLYBYTES, &data[i]);
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLq(&data[i], bytes + i * SABER_POLYBYTES);
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLp2BS(bytes + i * SABER_POLYCOMPRESSEDBYTES, &data[i]);
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLp(&data[i], bytes + i * SABER_POLYCOMPRESSEDBYTES);
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]) {
+    size_t i, j;
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            data->coeffs[j * 8 + i] = ((bytes[j] >> i) & 0x01);
+        }
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data) {
+    size_t i, j;
+    memset(bytes, 0, SABER_KEYBYTES);
+
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            bytes[j] = bytes[j] | ((data->coeffs[j * 8 + i] & 0x01) << i);
+        }
+    }
+}

crypto_kem/firesaber/avx2/pack_unpack.h

@@ -0,0 +1,28 @@
+#ifndef PACK_UNPACK_H
+#define PACK_UNPACK_H
+#include "SABER_params.h"
+#include "poly.h"
+#include <stdint.h>
+#include <stdio.h>
+
+void PQCLEAN_FIRESABER_AVX2_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data);
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]);
+
+
+void PQCLEAN_FIRESABER_AVX2_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]);
+
+void PQCLEAN_FIRESABER_AVX2_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]);
+
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]);
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]);
+
+
+void PQCLEAN_FIRESABER_AVX2_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]);
+
+void PQCLEAN_FIRESABER_AVX2_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data);
+
+
+#endif

crypto_kem/firesaber/avx2/poly.c

@@ -0,0 +1,62 @@
+#include "cbd.h"
+#include "fips202.h"
+#include "pack_unpack.h"
+#include "poly.h"
+
+
+void PQCLEAN_FIRESABER_AVX2_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const toom4_points s_eval[SABER_L], int transpose) {
+    size_t i, j;
+    toom4_points_product c_eval;
+
+    if (transpose) {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_FIRESABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[0][i], &s_eval[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_FIRESABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[j][i], &s_eval[j], 1);
+            }
+            PQCLEAN_FIRESABER_AVX2_toom4_interp(&c[i], &c_eval);
+        }
+    } else {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_FIRESABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[i][0], &s_eval[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_FIRESABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[i][j], &s_eval[j], 1);
+            }
+            PQCLEAN_FIRESABER_AVX2_toom4_interp(&c[i], &c_eval);
+        }
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_InnerProd(poly *c, const poly b[SABER_L], const toom4_points s_eval[SABER_L]) {
+    size_t i;
+    toom4_points_product c_eval; //Holds results for 9 Karatsuba at a time
+
+    PQCLEAN_FIRESABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &b[0], &s_eval[0], 0);
+    for (i = 1; i < SABER_L; i++) {
+        PQCLEAN_FIRESABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &b[i], &s_eval[i], 1);
+    }
+
+    PQCLEAN_FIRESABER_AVX2_toom4_interp(c, &c_eval);
+}
+
+void PQCLEAN_FIRESABER_AVX2_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYVECBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_FIRESABER_AVX2_BS2POLVECq(A[i], buf + i * SABER_POLYVECBYTES);
+    }
+}
+
+void PQCLEAN_FIRESABER_AVX2_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYCOINBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_NOISESEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_FIRESABER_AVX2_cbd(s[i].coeffs, buf + i * SABER_POLYCOINBYTES);
+    }
+}

crypto_kem/firesaber/avx2/poly.h

@@ -0,0 +1,38 @@
+#ifndef POLY_H
+#define POLY_H
+#include "SABER_params.h"
+#include <immintrin.h>
+#include <stdint.h>
+
+typedef union {
+    uint16_t coeffs[SABER_N];
+    __m256i dummy;
+} poly;
+
+typedef union {
+    uint16_t coeffs[4 * SABER_N];
+    __m256i dummy;
+} toom4_points;
+
+typedef union {
+    uint16_t coeffs[8 * SABER_N];
+    __m256i dummy;
+} toom4_points_product;
+
+void PQCLEAN_FIRESABER_AVX2_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const toom4_points s_eval[SABER_L], int transpose);
+
+void PQCLEAN_FIRESABER_AVX2_InnerProd(poly *c, const poly b[SABER_L], const toom4_points s_eval[SABER_L]);
+
+void PQCLEAN_FIRESABER_AVX2_GenMatrix(poly a[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]);
+
+void PQCLEAN_FIRESABER_AVX2_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]);
+
+
+void PQCLEAN_FIRESABER_AVX2_toom4_interp(poly *res_avx, const toom4_points_product *c_eval);
+
+void PQCLEAN_FIRESABER_AVX2_toom4_eval(toom4_points *b_eval, const poly *b);
+
+void PQCLEAN_FIRESABER_AVX2_toom4_mul_A_by_B_eval(toom4_points_product *c_eval, const poly *a_avx, const toom4_points *b_eval, int accumulate);
+
+
+#endif

crypto_kem/firesaber/avx2/verify.c

@@ -0,0 +1,35 @@
+#include "verify.h"
+
+/*-------------------------------------------------
+This file has been adapted from the implementation
+(available at https://github.com/pq-crystals/kyber) of
+"CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------*/
+
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+uint8_t PQCLEAN_FIRESABER_AVX2_verify(const uint8_t *a, const uint8_t *b, size_t len) {
+    uint64_t r;
+    size_t i;
+    r = 0;
+
+    for (i = 0; i < len; i++) {
+        r |= a[i] ^ b[i];
+    }
+
+    r = (~r + 1); // Two's complement
+    r >>= 63;
+    return (uint8_t) r;
+}
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_FIRESABER_AVX2_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b) {
+    size_t i;
+
+    b = -b;
+    for (i = 0; i < len; i++) {
+        r[i] ^= b & (x[i] ^ r[i]);
+    }
+}

crypto_kem/firesaber/avx2/verify.h

@@ -0,0 +1,22 @@
+#ifndef VERIFY_H
+#define VERIFY_H
+/*-------------------------------------------------
+This file has been adapted from the implementation
+(available at https://github.com/pq-crystals/kyber) of
+"CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------*/
+
+#include <stddef.h>
+#include <stdint.h>
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+uint8_t PQCLEAN_FIRESABER_AVX2_verify(const uint8_t *a, const uint8_t *b, size_t len);
+
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_FIRESABER_AVX2_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b);
+
+
+#endif

crypto_kem/firesaber/clean/LICENSE

@@ -1,8 +1 @@
-SABER_v1.1
-
-Public domain
-
-Authors: Jan-Pieter D'Anvers, Angshuman Karmakar, Sujoy Sinha Roy,
-Frederik Vercauteren
+Public Domain

crypto_kem/firesaber/clean/Makefile

@@ -1,10 +1,10 @@
 # This Makefile can be used with GNU Make or BSD Make
 
 LIB=libfiresaber_clean.a
-HEADERS=api.h cbd.h poly.h poly_mul.h SABER_indcpa.h SABER_params.h verify.h pack_unpack.h 
+HEADERS=api.h cbd.h pack_unpack.h poly.h poly_mul.h SABER_indcpa.h SABER_params.h verify.h 
 OBJECTS=cbd.o kem.o pack_unpack.o poly.o poly_mul.o SABER_indcpa.o verify.o 
 
-CFLAGS=-O3 -Wall -Wextra -Wpedantic -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common $(EXTRAFLAGS)
+CFLAGS=-O3 -Wall -Wextra -Wpedantic -Wvla -Werror -Wredundant-decls -Wmissing-prototypes -std=c99 -I../../../common $(EXTRAFLAGS)
 
 all: $(LIB)
 

crypto_kem/firesaber/clean/SABER_indcpa.c

@@ -3,296 +3,111 @@
 #include "fips202.h"
 #include "pack_unpack.h"
 #include "poly.h"
-#include "poly_mul.h"
 #include "randombytes.h"
 #include <stdint.h>
 #include <string.h>
 
+#define h1 (1 << (SABER_EQ - SABER_EP - 1))
+#define h2 ((1 << (SABER_EP - 2)) - (1 << (SABER_EP - SABER_ET - 1)) + (1 << (SABER_EQ - SABER_EP - 1)))
 
+void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]) {
+    size_t i, j;
 
-/*-----------------------------------------------------------------------------------
-    This routine generates a=[Matrix K x K] of 256-coefficient polynomials
+    poly A[SABER_L][SABER_L];
+    poly s[SABER_L];
+    poly res[SABER_L];
 
-#define h1 4 //2^(EQ-EP-1)
+    uint8_t rand[SABER_NOISESEEDBYTES];
+    uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
 
-#define h2 ( (1<<(SABER_EP-2)) - (1<<(SABER_EP-SABER_ET-1)) + (1<<(SABER_EQ-SABER_EP-1)) )
+    randombytes(seed_A, SABER_SEEDBYTES);
+    shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
 
-static void InnerProd(uint16_t pkcl[SABER_K][SABER_N], uint16_t skpv[SABER_K][SABER_N], uint16_t mod, uint16_t res[SABER_N]);
-static void MatrixVectorMul(polyvec *a, uint16_t skpv[SABER_K][SABER_N], uint16_t res[SABER_K][SABER_N], uint16_t mod, int16_t transpose);
+    randombytes(rand, SABER_NOISESEEDBYTES);
+    PQCLEAN_FIRESABER_CLEAN_GenSecret(s, rand);
+    PQCLEAN_FIRESABER_CLEAN_POLVECq2BS(sk, s);
 
-static void POL2MSG(const uint16_t *message_dec_unpacked, unsigned char *message_dec);
+    PQCLEAN_FIRESABER_CLEAN_GenMatrix(A, seed_A); // sample matrix A
+    PQCLEAN_FIRESABER_CLEAN_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const poly *)s, 1); // Matrix in transposed order
 
-static void GenMatrix(polyvec *a, const unsigned char *seed) {
-    unsigned char buf[SABER_K * SABER_K * (13 * SABER_N / 8)];
 
-    uint16_t temp_ar[SABER_N];
-
-    int i, j, k;
-    uint16_t mod = (SABER_Q - 1);
-
-    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
-
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_K; j++) {
-            PQCLEAN_FIRESABER_CLEAN_BS2POL(buf + (i * SABER_K + j) * (13 * SABER_N / 8), temp_ar);
-            for (k = 0; k < SABER_N; k++) {
-                a[i].vec[j].coeffs[k] = (temp_ar[k])& mod ;
-            }
-        }
-    }
-}
-
-
-void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(unsigned char *pk, unsigned char *sk) {
-    polyvec a[SABER_K];
-
-    uint16_t skpv[SABER_K][SABER_N];
-
-    unsigned char seed[SABER_SEEDBYTES];
-    unsigned char noiseseed[SABER_COINBYTES];
-    int32_t i, j;
-    uint16_t mod_q = SABER_Q - 1;
-
-
-    uint16_t res[SABER_K][SABER_N];
-
-    randombytes(seed, SABER_SEEDBYTES);
-
-    // for not revealing system RNG state
-    shake128(seed, SABER_SEEDBYTES, seed, SABER_SEEDBYTES);
-    randombytes(noiseseed, SABER_COINBYTES);
-
-    GenMatrix(a, seed);   //sample matrix A
-
-    // generate secret from constant-time binomial distribution
-    PQCLEAN_FIRESABER_CLEAN_GenSecret(skpv, noiseseed);
-
-    // do the matrix vector multiplication and rounding
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = 0;
-        }
-    }
-    MatrixVectorMul(a, skpv, res, SABER_Q - 1, 1);
-
-    // now rounding
-    for (i = 0; i < SABER_K; i++) {
+    // rounding
+    for (i = 0; i < SABER_L; i++) {
         for (j = 0; j < SABER_N; j++) {
-            // shift right 3 bits
-            res[i][j] = (res[i][j] + h1) & (mod_q);
-            res[i][j] = (res[i][j] >> (SABER_EQ - SABER_EP));
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
         }
     }
 
-    // unload and pack sk=3 x (256 coefficients of 14 bits)
-    PQCLEAN_FIRESABER_CLEAN_POLVEC2BS(sk, skpv, SABER_Q);
-
-    // unload and pack pk=256 bits seed and 3 x (256 coefficients of 11 bits)
-    // load the public-key coefficients
-    PQCLEAN_FIRESABER_CLEAN_POLVEC2BS(pk, res, SABER_P);
-
-
-    // now load the seedbytes in PK. Easy since seed bytes are kept in byte format.
-    for (i = 0; i < SABER_SEEDBYTES; i++) {
-        pk[SABER_POLYVECCOMPRESSEDBYTES + i] = seed[i];
-    }
-
+    PQCLEAN_FIRESABER_CLEAN_POLVECp2BS(pk, res); // pack public key
 }
 
 
-void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(const unsigned char *message_received, unsigned char *noiseseed, const unsigned char *pk, unsigned char *ciphertext) {
-    uint32_t i, j, k;
-    polyvec a[SABER_K];
-    unsigned char seed[SABER_SEEDBYTES];
-    // public key of received by the client
-    uint16_t pkcl[SABER_K][SABER_N];
-    uint16_t skpv1[SABER_K][SABER_N];
-    uint16_t message[SABER_KEYBYTES * 8];
-    uint16_t res[SABER_K][SABER_N];
-    uint16_t mod_p = SABER_P - 1;
-    uint16_t mod_q = SABER_Q - 1;
-    uint16_t vprime[SABER_N];
-    unsigned char msk_c[SABER_SCALEBYTES_KEM];
+void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]) {
+    size_t i, j;
 
-    // extract the seedbytes from Public Key.
-    for (i = 0; i < SABER_SEEDBYTES; i++) {
-        seed[i] = pk[ SABER_POLYVECCOMPRESSEDBYTES + i];
-    }
+    poly A[SABER_L][SABER_L];
+    poly res[SABER_L];
+    poly s[SABER_L];
+    poly *temp = A[0]; // re-use stack space
+    poly *vprime = &A[0][0];
+    poly *message = &A[0][1];
 
-    GenMatrix(a, seed);
+    const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+    uint8_t *msk_c = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
 
-    // generate secret from constant-time binomial distribution
-    PQCLEAN_FIRESABER_CLEAN_GenSecret(skpv1, noiseseed);
+    PQCLEAN_FIRESABER_CLEAN_GenSecret(s, noiseseed);
+    PQCLEAN_FIRESABER_CLEAN_GenMatrix(A, seed_A);
+    PQCLEAN_FIRESABER_CLEAN_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const poly *)s, 0); // 0 => not transposed
 
-    // matrix-vector multiplication and rounding
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = 0;
-        }
-    }
-    MatrixVectorMul(a, skpv1, res, SABER_Q - 1, 0);
-
-    // now rounding
-    //shift right 3 bits
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = ( res[i][j] + h1 ) & mod_q;
-            res[i][j] = (res[i][j] >> (SABER_EQ - SABER_EP) );
-        }
-    }
 
-    PQCLEAN_FIRESABER_CLEAN_POLVEC2BS(ciphertext, res, SABER_P);
-
-    // ************client matrix-vector multiplication ends************
-
-    // now calculate the v'
-    // unpack the public_key
-    // pkcl is the b in the protocol
-    PQCLEAN_FIRESABER_CLEAN_BS2POLVEC(pk, pkcl, SABER_P);
-    for (i = 0; i < SABER_N; i++) {
-        vprime[i] = 0;
-    }
-    for (i = 0; i < SABER_K; i++) {
+    // rounding
+    for (i = 0; i < SABER_L; i++) { //shift right EQ-EP bits
         for (j = 0; j < SABER_N; j++) {
-            skpv1[i][j] = skpv1[i][j] & (mod_p);
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
         }
     }
+    PQCLEAN_FIRESABER_CLEAN_POLVECp2BS(ciphertext, res);
 
     // vector-vector scalar multiplication with mod p
-    InnerProd(pkcl, skpv1, mod_p, vprime);
+    PQCLEAN_FIRESABER_CLEAN_BS2POLVECp(temp, pk);
+    PQCLEAN_FIRESABER_CLEAN_InnerProd(vprime, temp, s);
+    PQCLEAN_FIRESABER_CLEAN_BS2POLmsg(message, m);
 
-    // addition of h1 to vprime
     for (i = 0; i < SABER_N; i++) {
-        vprime[i] = vprime[i] + h1;
+        vprime->coeffs[i] += h1 - (message->coeffs[i] << (SABER_EP - 1));
+        vprime->coeffs[i] &= SABER_P - 1;
+        vprime->coeffs[i] >>= SABER_EP - SABER_ET;
     }
 
-    // unpack message_received;
-    for (j = 0; j < SABER_KEYBYTES; j++) {
-        for (i = 0; i < 8; i++) {
-            message[8 * j + i] = ((message_received[j] >> i) & 0x01);
-        }
-    }
-
-    // message encoding
-    for (i = 0; i < SABER_N; i++) {
-        message[i] = (message[i] << (SABER_EP - 1));
-    }
-
-    for (k = 0; k < SABER_N; k++) {
-        vprime[k] = ( (vprime[k] - message[k]) & (mod_p) ) >> (SABER_EP - SABER_ET);
-    }
-
-
-    PQCLEAN_FIRESABER_CLEAN_pack_6bit(msk_c, vprime);
-
-    for (j = 0; j < SABER_SCALEBYTES_KEM; j++) {
-        ciphertext[SABER_POLYVECCOMPRESSEDBYTES + j] = msk_c[j];
-    }
+    PQCLEAN_FIRESABER_CLEAN_POLT2BS(msk_c, vprime);
 }
 
 
-void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_dec(const unsigned char *sk, const unsigned char *ciphertext, unsigned char message_dec[]) {
-    uint32_t i, j;
-    // secret key of the server
-    uint16_t sksv[SABER_K][SABER_N];
-    uint16_t pksv[SABER_K][SABER_N];
-    uint8_t scale_ar[SABER_SCALEBYTES_KEM];
-    uint16_t mod_p = SABER_P - 1;
-    uint16_t v[SABER_N];
-    uint16_t op[SABER_N];
+void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]) {
+    size_t i;
 
-    // sksv is the secret-key
-    PQCLEAN_FIRESABER_CLEAN_BS2POLVEC(sk, sksv, SABER_Q);
-    // pksv is the ciphertext
-    PQCLEAN_FIRESABER_CLEAN_BS2POLVEC(ciphertext, pksv, SABER_P);
+    poly temp[SABER_L];
+    poly s[SABER_L];
 
-    // vector-vector scalar multiplication with mod p
-    for (i = 0; i < SABER_N; i++) {
-        v[i] = 0;
-    }
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            sksv[i][j] = sksv[i][j] & (mod_p);
-        }
-    }
-    InnerProd(pksv, sksv, mod_p, v);
+    const uint8_t *packed_cm = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+    poly *v = &temp[0];
+    poly *cm = &temp[1];
 
-    //Extraction
-    for (i = 0; i < SABER_SCALEBYTES_KEM; i++) {
-        scale_ar[i] = ciphertext[SABER_POLYVECCOMPRESSEDBYTES + i];
-    }
+    PQCLEAN_FIRESABER_CLEAN_BS2POLVECq(s, sk);
+    PQCLEAN_FIRESABER_CLEAN_BS2POLVECp(temp, ciphertext);
+    PQCLEAN_FIRESABER_CLEAN_InnerProd(&temp[0], temp, s);
 
-    PQCLEAN_FIRESABER_CLEAN_un_pack6bit(scale_ar, op);
+    PQCLEAN_FIRESABER_CLEAN_BS2POLT(cm, packed_cm);
 
-    //addition of h1
     for (i = 0; i < SABER_N; i++) {
-        v[i] = ( ( v[i] + h2 - (op[i] << (SABER_EP - SABER_ET)) ) & (mod_p) ) >> (SABER_EP - 1);
+        v->coeffs[i] += h2 - (cm->coeffs[i] << (SABER_EP - SABER_ET));
+        v->coeffs[i] &= SABER_P - 1;
+        v->coeffs[i] >>= SABER_EP - 1;
     }
 
-    // pack decrypted message
-    POL2MSG(v, message_dec);
-}
-static void MatrixVectorMul(polyvec *a, uint16_t skpv[SABER_K][SABER_N], uint16_t res[SABER_K][SABER_N], uint16_t mod, int16_t transpose) {
-    uint16_t acc[SABER_N];
-    int32_t i, j, k;
-
-    if (transpose == 1) {
-        for (i = 0; i < SABER_K; i++) {
-            for (j = 0; j < SABER_K; j++) {
-                PQCLEAN_FIRESABER_CLEAN_pol_mul((uint16_t *)&a[j].vec[i], skpv[j], acc, SABER_Q, SABER_N);
-
-                for (k = 0; k < SABER_N; k++) {
-                    res[i][k] = res[i][k] + acc[k];
-                    //reduction mod p
-                    res[i][k] = (res[i][k] & mod);
-                    //clear the accumulator
-                    acc[k] = 0;
-                }
-            }
-        }
-    } else {
-        for (i = 0; i < SABER_K; i++) {
-            for (j = 0; j < SABER_K; j++) {
-                PQCLEAN_FIRESABER_CLEAN_pol_mul((uint16_t *)&a[i].vec[j], skpv[j], acc, SABER_Q, SABER_N);
-                for (k = 0; k < SABER_N; k++) {
-                    res[i][k] = res[i][k] + acc[k];
-                    // reduction
-                    res[i][k] = res[i][k] & mod;
-                    // clear the accumulator
-                    acc[k] = 0;
-                }
-            }
-        }
-    }
-}
-
-static void POL2MSG(const uint16_t *message_dec_unpacked, unsigned char *message_dec) {
-    int32_t i, j;
-
-    for (j = 0; j < SABER_KEYBYTES; j++) {
-        message_dec[j] = 0;
-        for (i = 0; i < 8; i++) {
-            message_dec[j] = message_dec[j] | (uint8_t) (message_dec_unpacked[j * 8 + i] << i);
-        }
-    }
-}
-
-
-static void InnerProd(uint16_t pkcl[SABER_K][SABER_N], uint16_t skpv[SABER_K][SABER_N], uint16_t mod, uint16_t res[SABER_N]) {
-    uint32_t j, k;
-    uint16_t acc[SABER_N];
-
-    // vector-vector scalar multiplication with mod p
-    for (j = 0; j < SABER_K; j++) {
-        PQCLEAN_FIRESABER_CLEAN_pol_mul(pkcl[j], skpv[j], acc, SABER_P, SABER_N);
-
-        for (k = 0; k < SABER_N; k++) {
-            res[k] = res[k] + acc[k];
-            // reduction
-            res[k] = res[k] & mod;
-            // clear the accumulator
-            acc[k] = 0;
-        }
-    }
+    PQCLEAN_FIRESABER_CLEAN_POLmsg2BS(m, v);
 }

crypto_kem/firesaber/clean/SABER_indcpa.h

@@ -1,9 +1,13 @@
 #ifndef INDCPA_H
 #define INDCPA_H
+#include "SABER_params.h"
+#include <stdint.h>
 
-void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(unsigned char *pk, unsigned char *sk);
-void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(const unsigned char *message, unsigned char *noiseseed, const unsigned char *pk, unsigned char *ciphertext);
-void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_dec(const unsigned char *sk, const unsigned char *ciphertext, unsigned char *message_dec);
+void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]);
 
-#endif
+void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]);
+
+void PQCLEAN_FIRESABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]);
 
+
+#endif

crypto_kem/firesaber/clean/SABER_params.h

@@ -1,49 +1,41 @@
 #ifndef PARAMS_H
 #define PARAMS_H
 
-#include "api.h"
 
-#define SABER_K 4
+/* Don't change anything below this line */
+#define SABER_L 4
 #define SABER_MU 6
 #define SABER_ET 6
 
-#define SABER_EQ 13
-#define SABER_EP 10
-
 #define SABER_N 256
-#define SABER_Q 8192
-#define SABER_P 1024
 
-#define SABER_SEEDBYTES       32
-#define SABER_NOISESEEDBYTES  32
-#define SABER_COINBYTES       32
-#define SABER_KEYBYTES        32
-
-#define SABER_HASHBYTES       32
+#define SABER_EP 10
+#define SABER_P (1 << SABER_EP)
 
-#define SABER_POLYBYTES       416 //13*256/8 
+#define SABER_EQ 13
+#define SABER_Q (1 << SABER_EQ)
 
-#define SABER_POLYVECBYTES    (SABER_K * SABER_POLYBYTES)
+#define SABER_SEEDBYTES 32
+#define SABER_NOISESEEDBYTES 32
+#define SABER_KEYBYTES 32
+#define SABER_HASHBYTES 32
 
-#define SABER_POLYVECCOMPRESSEDBYTES (SABER_K * 320) //10*256/8 NOTE : changed till here due to parameter adaptation
+#define SABER_POLYCOINBYTES (SABER_MU * SABER_N / 8)
 
-#define SABER_CIPHERTEXTBYTES (SABER_POLYVECCOMPRESSEDBYTES)
+#define SABER_POLYBYTES (SABER_EQ * SABER_N / 8)
+#define SABER_POLYVECBYTES (SABER_L * SABER_POLYBYTES)
 
-#define SABER_SCALEBYTES (SABER_DELTA*SABER_N/8)
+#define SABER_POLYCOMPRESSEDBYTES (SABER_EP * SABER_N / 8)
+#define SABER_POLYVECCOMPRESSEDBYTES (SABER_L * SABER_POLYCOMPRESSEDBYTES)
 
-#define SABER_SCALEBYTES_KEM ((SABER_ET)*SABER_N/8)
+#define SABER_SCALEBYTES_KEM (SABER_ET * SABER_N / 8)
 
 #define SABER_INDCPA_PUBLICKEYBYTES (SABER_POLYVECCOMPRESSEDBYTES + SABER_SEEDBYTES)
 #define SABER_INDCPA_SECRETKEYBYTES (SABER_POLYVECBYTES)
 
 #define SABER_PUBLICKEYBYTES (SABER_INDCPA_PUBLICKEYBYTES)
+#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES + SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
 
-#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES +  SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
-
-#define SABER_BYTES_CCA_DEC   (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM) /* Second part is for Targhi-Unruh */
-
-
-
+#define SABER_BYTES_CCA_DEC (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM)
 
 #endif
-

crypto_kem/firesaber/clean/api.h

@@ -1,14 +1,18 @@
 #ifndef PQCLEAN_FIRESABER_CLEAN_API_H
 #define PQCLEAN_FIRESABER_CLEAN_API_H
 
+
 #define PQCLEAN_FIRESABER_CLEAN_CRYPTO_ALGNAME "FireSaber"
-#define PQCLEAN_FIRESABER_CLEAN_CRYPTO_SECRETKEYBYTES 3040
-#define PQCLEAN_FIRESABER_CLEAN_CRYPTO_PUBLICKEYBYTES (4*320+32)
 #define PQCLEAN_FIRESABER_CLEAN_CRYPTO_BYTES 32
 #define PQCLEAN_FIRESABER_CLEAN_CRYPTO_CIPHERTEXTBYTES 1472
+#define PQCLEAN_FIRESABER_CLEAN_CRYPTO_PUBLICKEYBYTES 1312
+#define PQCLEAN_FIRESABER_CLEAN_CRYPTO_SECRETKEYBYTES 3040
 
 int PQCLEAN_FIRESABER_CLEAN_crypto_kem_keypair(unsigned char *pk, unsigned char *sk);
-int PQCLEAN_FIRESABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *ss, const unsigned char *pk);
-int PQCLEAN_FIRESABER_CLEAN_crypto_kem_dec(unsigned char *ss, const unsigned char *ct, const unsigned char *sk);
 
-#endif /* api_h */
+int PQCLEAN_FIRESABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *k, const unsigned char *pk);
+
+int PQCLEAN_FIRESABER_CLEAN_crypto_kem_dec(unsigned char *k, const unsigned char *ct, const unsigned char *sk);
+
+
+#endif /* PQCLEAN_FIRESABER_CLEAN_API_H */

crypto_kem/firesaber/clean/cbd.c

@@ -1,3 +1,7 @@
+#include "SABER_params.h"
+#include "api.h"
+#include "cbd.h"
+#include <stdint.h>
 /*---------------------------------------------------------------------
 This file has been adapted from the implementation
 (available at, Public Domain https://github.com/pq-crystals/kyber)
@@ -6,12 +10,8 @@ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------------------------*/
 
-#include "SABER_params.h"
-#include "api.h"
-#include "cbd.h"
-#include <stdint.h>
 
-static uint64_t load_littleendian(const unsigned char *x, int bytes) {
+static uint64_t load_littleendian(const uint8_t *x, int bytes) {
     int i;
     uint64_t r = x[0];
     for (i = 1; i < bytes; i++) {
@@ -20,10 +20,7 @@ static uint64_t load_littleendian(const unsigned char *x, int bytes) {
     return r;
 }
 
-
-void PQCLEAN_FIRESABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
-    uint16_t Qmod_minus1 = SABER_Q - 1;
-
+void PQCLEAN_FIRESABER_CLEAN_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]) {
     uint32_t t, d, a[4], b[4];
     int i, j;
 
@@ -34,19 +31,18 @@ void PQCLEAN_FIRESABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
             d += (t >> j) & 0x249249;
         }
 
-        a[0] =  d & 0x7;
-        b[0] = (d >>  3) & 0x7;
-        a[1] = (d >>  6) & 0x7;
-        b[1] = (d >>  9) & 0x7;
+        a[0] = d & 0x7;
+        b[0] = (d >> 3) & 0x7;
+        a[1] = (d >> 6) & 0x7;
+        b[1] = (d >> 9) & 0x7;
         a[2] = (d >> 12) & 0x7;
         b[2] = (d >> 15) & 0x7;
         a[3] = (d >> 18) & 0x7;
         b[3] = (d >> 21);
 
-        r[4 * i + 0] = (uint16_t)(a[0]  - b[0]) & Qmod_minus1;
-        r[4 * i + 1] = (uint16_t)(a[1]  - b[1]) & Qmod_minus1;
-        r[4 * i + 2] = (uint16_t)(a[2]  - b[2]) & Qmod_minus1;
-        r[4 * i + 3] = (uint16_t)(a[3]  - b[3]) & Qmod_minus1;
-
+        s[4 * i + 0] = (uint16_t)(a[0] - b[0]);
+        s[4 * i + 1] = (uint16_t)(a[1] - b[1]);
+        s[4 * i + 2] = (uint16_t)(a[2] - b[2]);
+        s[4 * i + 3] = (uint16_t)(a[3] - b[3]);
     }
 }

crypto_kem/firesaber/clean/cbd.h

@@ -1,6 +1,5 @@
 #ifndef CBD_H
 #define CBD_H
-
 /*---------------------------------------------------------------------
 This file has been adapted from the implementation
 (available at, Public Domain https://github.com/pq-crystals/kyber)
@@ -8,10 +7,10 @@ of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
 by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------------------------*/
-
-#include "poly.h"
+#include "SABER_params.h"
 #include <stdint.h>
 
-void PQCLEAN_FIRESABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf);
+void PQCLEAN_FIRESABER_CLEAN_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]);
+
 
 #endif

crypto_kem/firesaber/clean/kem.c

@@ -1,96 +1,77 @@
 #include "SABER_indcpa.h"
 #include "SABER_params.h"
+#include "api.h"
 #include "fips202.h"
 #include "randombytes.h"
 #include "verify.h"
+#include <stddef.h>
 #include <stdint.h>
-#include <stdio.h>
-#include <string.h>
 
-int PQCLEAN_FIRESABER_CLEAN_crypto_kem_keypair(unsigned char *pk, unsigned char *sk) {
-    int i;
 
-    // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
-    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(pk, sk);
+int PQCLEAN_FIRESABER_CLEAN_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
+    size_t i;
 
-    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
+    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
     for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
-        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];
+        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
     }
 
-    // Then hash(pk) is appended.
-    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES);
+    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES); // Then hash(pk) is appended.
 
-    // Remaining part of sk contains a pseudo-random number.
-    // This is output when check in crypto_kem_dec() fails.
-    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES );
+    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES); // Remaining part of sk contains a pseudo-random number.
+    // This is output when check in PQCLEAN_FIRESABER_CLEAN_crypto_kem_dec() fails.
     return (0);
 }
 
-int PQCLEAN_FIRESABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *ss, const unsigned char *pk) {
-    // Will contain key, coins
-    unsigned char kr[64];
-    unsigned char buf[64];
+int PQCLEAN_FIRESABER_CLEAN_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t *pk) {
 
-    randombytes(buf, 32);
+    uint8_t kr[64]; // Will contain key, coins
+    uint8_t buf[64];
 
-    // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
-    sha3_256(buf, buf, 32);
+    randombytes(buf, 32);
 
-    // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
-    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES);
+    sha3_256(buf, buf, 32); // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
 
-    // kr[0:63] <-- Hash(buf[0:63]);
-    sha3_512(kr, buf, 64);
+    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES); // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
 
+    sha3_512(kr, buf, 64);               // kr[0:63] <-- Hash(buf[0:63]);
     // K^ <-- kr[0:31]
     // noiseseed (r) <-- kr[32:63];
-    // buf[0:31] contains message; kr[32:63] contains randomness r;
-    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(buf, kr + 32, pk,  ct);
+    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(c, buf, kr + 32, pk); // buf[0:31] contains message; kr[32:63] contains randomness r;
 
-    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC);
 
-    // hash concatenation of pre-k and h(c) to k
-    sha3_256(ss, kr, 64);
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
 
     return (0);
 }
 
+int PQCLEAN_FIRESABER_CLEAN_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
+    size_t i;
+    uint8_t fail;
+    uint8_t cmp[SABER_BYTES_CCA_DEC];
+    uint8_t buf[64];
+    uint8_t kr[64]; // Will contain key, coins
+    const uint8_t *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
 
-int PQCLEAN_FIRESABER_CLEAN_crypto_kem_dec(unsigned char *ss, const unsigned char *ct, const unsigned char *sk) {
-    int i;
-    unsigned char fail;
-    unsigned char cmp[SABER_BYTES_CCA_DEC];
-    unsigned char buf[64];
-
-    // Will contain key, coins
-    unsigned char kr[64];
-    const unsigned char *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
-
-    // buf[0:31] <-- message
-    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_dec(sk, ct, buf);
-
+    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_dec(buf, sk, c); // buf[0:31] <-- message
 
     // Multitarget countermeasure for coins + contributory KEM
-    // Save hash by storing h(pk) in sk
-    for (i = 0; i < 32; i++) {
+    for (i = 0; i < 32; i++) { // Save hash by storing h(pk) in sk
         buf[32 + i] = sk[SABER_SECRETKEYBYTES - 64 + i];
     }
 
     sha3_512(kr, buf, 64);
 
-    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(buf, kr + 32, pk, cmp);
-
+    PQCLEAN_FIRESABER_CLEAN_indcpa_kem_enc(cmp, buf, kr + 32, pk);
 
-    fail = PQCLEAN_FIRESABER_CLEAN_verify(ct, cmp, SABER_BYTES_CCA_DEC);
+    fail = PQCLEAN_FIRESABER_CLEAN_verify(c, cmp, SABER_BYTES_CCA_DEC);
 
-    // overwrite coins in kr with h(c)
-    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC); // overwrite coins in kr with h(c)
 
     PQCLEAN_FIRESABER_CLEAN_cmov(kr, sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES, fail);
 
-    // hash concatenation of pre-k and h(c) to k
-    sha3_256(ss, kr, 64);
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
 
     return (0);
 }

crypto_kem/firesaber/clean/pack_unpack.c

@@ -1,254 +1,151 @@
+#include "SABER_params.h"
 #include "pack_unpack.h"
+#include "poly.h"
+#include <string.h>
 
-void PQCLEAN_FIRESABER_CLEAN_pack_3bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
-    for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 3 * j;
-        offset_data = 8 * j;
-        bytes[offset_byte + 0] = (data[offset_data + 0] & 0x7) |
-                                 ((data[offset_data + 1] & 0x7) << 3) |
-                                 ((data[offset_data + 2] & 0x3) << 6);
-        bytes[offset_byte + 1] = ((data[offset_data + 2] >> 2 ) & 0x01)  |
-                                 ((data[offset_data + 3] & 0x7) << 1) |
-                                 ((data[offset_data + 4] & 0x7) << 4) |
-                                 (((data[offset_data + 5]) & 0x01) << 7);
-        bytes[offset_byte + 2] = ((data[offset_data + 5] >> 1 ) & 0x03) |
-                                 ((data[offset_data + 6] & 0x7) << 2) |
-                                 ((data[offset_data + 7] & 0x7) << 5);
+void PQCLEAN_FIRESABER_CLEAN_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = (uint8_t) ((in[0] & 0x3f) | (in[1] << 6));
+        out[1] = (uint8_t) (((in[1] >> 2) & 0x0f) | (in[2] << 4));
+        out[2] = (uint8_t) (((in[2] >> 4) & 0x03) | (in[3] << 2));
+        in += 4;
+        out += 3;
     }
 }
 
-void PQCLEAN_FIRESABER_CLEAN_un_pack3bit(const uint8_t *bytes, uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
-    for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 3 * j;
-        offset_data = 8 * j;
-        data[offset_data + 0] = (bytes[offset_byte + 0]) & 0x07;
-        data[offset_data + 1] = ((bytes[offset_byte + 0]) >> 3 ) & 0x07;
-        data[offset_data + 2] = (((bytes[offset_byte + 0]) >> 6 ) & 0x03) |
-                                (((bytes[offset_byte + 1]) & 0x01) << 2);
-        data[offset_data + 3] = ((bytes[offset_byte + 1]) >> 1 ) & 0x07;
-        data[offset_data + 4] = ((bytes[offset_byte + 1]) >> 4 ) & 0x07;
-        data[offset_data + 5] = (((bytes[offset_byte + 1]) >> 7 ) & 0x01) |
-                                (((bytes[offset_byte + 2]) & 0x03) << 1);
-        data[offset_data + 6] = ((bytes[offset_byte + 2] >> 2) & 0x07);
-        data[offset_data + 7] = ((bytes[offset_byte + 2] >> 5) & 0x07);
+void PQCLEAN_FIRESABER_CLEAN_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]) {
+    /* This function does not reduce its output mod T */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = in[0];
+        out[1] = (in[0] >> 6) | (in[1] << 2);
+        out[2] = (in[1] >> 4) | (in[2] << 4);
+        out[3] = (in[2] >> 2);
+        in += 3;
+        out += 4;
     }
 }
 
-void PQCLEAN_FIRESABER_CLEAN_pack_4bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data;
-
-    for (j = 0; j < SABER_N / 2; j++) {
-        offset_data = 2 * j;
-        bytes[j] = (data[offset_data] & 0x0f) |
-                   ((data[offset_data + 1] & 0x0f) << 4);
+static void POLq2BS(uint8_t bytes[SABER_POLYBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x1f) | (in[1] << 5));
+        out[2] = (uint8_t) (in[1] >> 3);
+        out[3] = (uint8_t) (((in[1] >> 11) & 0x03) | (in[2] << 2));
+        out[4] = (uint8_t) (((in[2] >> 6) & 0x7f) | (in[3] << 7));
+        out[5] = (uint8_t) (in[3] >> 1);
+        out[6] = (uint8_t) (((in[3] >> 9) & 0x0f) | (in[4] << 4));
+        out[7] = (uint8_t) (in[4] >> 4);
+        out[8] = (uint8_t) (((in[4] >> 12) & 0x01) | (in[5] << 1));
+        out[9] = (uint8_t) (((in[5] >> 7) & 0x3f) | (in[6] << 6));
+        out[10] = (uint8_t) (in[6] >> 2);
+        out[11] = (uint8_t) (((in[6] >> 10) & 0x07) | (in[7] << 3));
+        out[12] = (uint8_t) (in[7] >> 5);
+        in += 8;
+        out += 13;
     }
 }
 
-void PQCLEAN_FIRESABER_CLEAN_un_pack4bit(const unsigned char *bytes, uint16_t *ar) {
-    uint32_t j;
-    uint32_t offset_data;
-
-    for (j = 0; j < SABER_N / 2; j++) {
-        offset_data = 2 * j;
-        ar[offset_data] = bytes[j] & 0x0f;
-        ar[offset_data + 1] = (bytes[j] >> 4) & 0x0f;
+static void BS2POLq(poly *data, const uint8_t bytes[SABER_POLYBYTES]) {
+    /* This function does not reduce its output mod Q */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (in[0]) | (in[1] << 8);
+        out[1] = (in[1] >> 5) | (in[2] << 3) | (in[3] << 11);
+        out[2] = (in[3] >> 2) | (in[4] << 6);
+        out[3] = (in[4] >> 7) | (in[5] << 1) | (in[6] << 9);
+        out[4] = (in[6] >> 4) | (in[7] << 4) | (in[8] << 12);
+        out[5] = (in[8] >> 1) | (in[9] << 7);
+        out[6] = (in[9] >> 6) | (in[10] << 2) | (in[11] << 10);
+        out[7] = (in[11] >> 3) | (in[12] << 5);
+        in += 13;
+        out += 8;
     }
 }
 
-void PQCLEAN_FIRESABER_CLEAN_pack_6bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+static void POLp2BS(uint8_t bytes[SABER_POLYCOMPRESSEDBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
     for (j = 0; j < SABER_N / 4; j++) {
-        offset_byte = 3 * j;
-        offset_data = 4 * j;
-        bytes[offset_byte + 0] = (data[offset_data + 0] & 0x3f) |
-                                 ((data[offset_data + 1] & 0x03) << 6);
-        bytes[offset_byte + 1] = ((data[offset_data + 1] >> 2) & 0x0f) |
-                                 ((data[offset_data + 2] & 0x0f) << 4);
-        bytes[offset_byte + 2] = ((data[offset_data + 2] >> 4) & 0x03) |
-                                 ((data[offset_data + 3] & 0x3f) << 2);
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x03) | (in[1] << 2));
+        out[2] = (uint8_t) (((in[1] >> 6) & 0x0f) | (in[2] << 4));
+        out[3] = (uint8_t) (((in[2] >> 4) & 0x3f) | (in[3] << 6));
+        out[4] = (uint8_t) (in[3] >> 2);
+        in += 4;
+        out += 5;
     }
 }
 
-
-void PQCLEAN_FIRESABER_CLEAN_un_pack6bit(const unsigned char *bytes, uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+static void BS2POLp(poly *data, const uint8_t bytes[SABER_POLYCOMPRESSEDBYTES]) {
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
     for (j = 0; j < SABER_N / 4; j++) {
-        offset_byte = 3 * j;
-        offset_data = 4 * j;
-        data[offset_data + 0] = bytes[offset_byte + 0] & 0x3f;
-        data[offset_data + 1] = ((bytes[offset_byte + 0] >> 6) & 0x03) |
-                                ((bytes[offset_byte + 1] & 0x0f) << 2);
-        data[offset_data + 2] = ((bytes[offset_byte + 1] & 0xff) >> 4) |
-                                ((bytes[offset_byte + 2] & 0x03) << 4);
-        data[offset_data + 3] = ((bytes[offset_byte + 2] & 0xff) >> 2);
+        out[0] = in[0] | (in[1] << 8);
+        out[1] = (in[1] >> 2) | (in[2] << 6);
+        out[2] = (in[2] >> 4) | (in[3] << 4);
+        out[3] = (in[3] >> 6) | (in[4] << 2);
+        in += 5;
+        out += 4;
     }
 }
 
-
-static void POLVECp2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 10) / 8;
-        for (j = 0; j < SABER_N / 4; j++) {
-            offset_byte = offset_byte1 + 5 * j;
-            offset_data = 4 * j;
-            bytes[offset_byte + 0] = (data[i][offset_data + 0] & (0xff));
-            bytes[offset_byte + 1] = ((data[i][offset_data + 0] >> 8) & 0x03) |
-                                     ((data[i][offset_data + 1] & 0x3f) << 2);
-            bytes[offset_byte + 2] = ((data[i][offset_data + 1] >> 6) & 0x0f) |
-                                     ((data[i][offset_data + 2] & 0x0f) << 4);
-            bytes[offset_byte + 3] = ((data[i][offset_data + 2] >> 4) & 0x3f) |
-                                     ((data[i][offset_data + 3] & 0x03) << 6);
-            bytes[offset_byte + 4] = ((data[i][offset_data + 3] >> 2) & 0xff);
-        }
+void PQCLEAN_FIRESABER_CLEAN_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLq2BS(bytes + i * SABER_POLYBYTES, &data[i]);
     }
 }
 
-static void BS2POLVECp(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 10) / 8;
-        for (j = 0; j < SABER_N / 4; j++) {
-            offset_byte = offset_byte1 + 5 * j;
-            offset_data = 4 * j;
-            data[i][offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                       ((bytes[offset_byte + 1] & 0x03) << 8);
-            data[i][offset_data + 1] = ((bytes[offset_byte + 1] >> 2) & (0x3f)) |
-                                       ((bytes[offset_byte + 2] & 0x0f) << 6);
-            data[i][offset_data + 2] = ((bytes[offset_byte + 2] >> 4) & (0x0f)) |
-                                       ((bytes[offset_byte + 3] & 0x3f) << 4);
-            data[i][offset_data + 3] = ((bytes[offset_byte + 3] >> 6) & (0x03)) |
-                                       ((bytes[offset_byte + 4] & 0xff) << 2);
-        }
+void PQCLEAN_FIRESABER_CLEAN_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLq(&data[i], bytes + i * SABER_POLYBYTES);
     }
 }
 
-
-
-static void POLVECq2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 13) / 8;
-        for (j = 0; j < SABER_N / 8; j++) {
-            offset_byte = offset_byte1 + 13 * j;
-            offset_data = 8 * j;
-            bytes[offset_byte + 0] = (data[i][offset_data + 0] & (0xff));
-            bytes[offset_byte + 1] = ((data[i][offset_data + 0] >> 8) & 0x1f) |
-                                     ((data[i][offset_data + 1] & 0x07) << 5);
-            bytes[offset_byte + 2] = ((data[i][offset_data + 1] >> 3) & 0xff);
-            bytes[offset_byte + 3] = ((data[i][offset_data + 1] >> 11) & 0x03) |
-                                     ((data[i][offset_data + 2] & 0x3f) << 2);
-            bytes[offset_byte + 4] = ((data[i][offset_data + 2] >> 6) & 0x7f) |
-                                     ((data[i][offset_data + 3] & 0x01) << 7);
-            bytes[offset_byte + 5] = ((data[i][offset_data + 3] >> 1) & 0xff);
-            bytes[offset_byte + 6] = ((data[i][offset_data + 3] >> 9) & 0x0f) |
-                                     ((data[i][offset_data + 4] & 0x0f) << 4);
-            bytes[offset_byte + 7] = ((data[i][offset_data + 4] >> 4) & 0xff);
-            bytes[offset_byte + 8] = ((data[i][offset_data + 4] >> 12) & 0x01) |
-                                     ((data[i][offset_data + 5] & 0x7f) << 1);
-            bytes[offset_byte + 9] = ((data[i][offset_data + 5] >> 7) & 0x3f) |
-                                     ((data[i][offset_data + 6] & 0x03) << 6);
-            bytes[offset_byte + 10] = ((data[i][offset_data + 6] >> 2) & 0xff);
-            bytes[offset_byte + 11] = ((data[i][offset_data + 6] >> 10) & 0x07) |
-                                      ((data[i][offset_data + 7] & 0x1f) << 3);
-            bytes[offset_byte + 12] = ((data[i][offset_data + 7] >> 5) & 0xff);
-        }
+void PQCLEAN_FIRESABER_CLEAN_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLp2BS(bytes + i * SABER_POLYCOMPRESSEDBYTES, &data[i]);
     }
 }
 
-static void BS2POLVECq(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 13) / 8;
-        for (j = 0; j < SABER_N / 8; j++) {
-            offset_byte = offset_byte1 + 13 * j;
-            offset_data = 8 * j;
-            data[i][offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                       ((bytes[offset_byte + 1] & 0x1f) << 8);
-            data[i][offset_data + 1] = (bytes[offset_byte + 1] >> 5 & (0x07)) |
-                                       ((bytes[offset_byte + 2] & 0xff) << 3) |
-                                       ((bytes[offset_byte + 3] & 0x03) << 11);
-            data[i][offset_data + 2] = (bytes[offset_byte + 3] >> 2 & (0x3f)) |
-                                       ((bytes[offset_byte + 4] & 0x7f) << 6);
-            data[i][offset_data + 3] = (bytes[offset_byte + 4] >> 7 & (0x01)) |
-                                       ((bytes[offset_byte + 5] & 0xff) << 1) |
-                                       ((bytes[offset_byte + 6] & 0x0f) << 9);
-            data[i][offset_data + 4] = (bytes[offset_byte + 6] >> 4 & (0x0f)) |
-                                       ((bytes[offset_byte + 7] & 0xff) << 4) |
-                                       ((bytes[offset_byte + 8] & 0x01) << 12);
-            data[i][offset_data + 5] = (bytes[offset_byte + 8] >> 1 & (0x7f)) |
-                                       ((bytes[offset_byte + 9] & 0x3f) << 7);
-            data[i][offset_data + 6] = (bytes[offset_byte + 9] >> 6 & (0x03)) |
-                                       ((bytes[offset_byte + 10] & 0xff) << 2) |
-                                       ((bytes[offset_byte + 11] & 0x07) << 10);
-            data[i][offset_data + 7] = (bytes[offset_byte + 11] >> 3 & (0x1f)) |
-                                       ((bytes[offset_byte + 12] & 0xff) << 5);
-        }
+void PQCLEAN_FIRESABER_CLEAN_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLp(&data[i], bytes + i * SABER_POLYCOMPRESSEDBYTES);
     }
 }
 
-//only BS2POLq no BS2POLp
-void PQCLEAN_FIRESABER_CLEAN_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
-    for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 13 * j;
-        offset_data = 8 * j;
-        data[offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                ((bytes[offset_byte + 1] & 0x1f) << 8);
-        data[offset_data + 1] = (bytes[offset_byte + 1] >> 5 & (0x07)) |
-                                ((bytes[offset_byte + 2] & 0xff) << 3) |
-                                ((bytes[offset_byte + 3] & 0x03) << 11);
-        data[offset_data + 2] = (bytes[offset_byte + 3] >> 2 & (0x3f)) |
-                                ((bytes[offset_byte + 4] & 0x7f) << 6);
-        data[offset_data + 3] = (bytes[offset_byte + 4] >> 7 & (0x01)) |
-                                ((bytes[offset_byte + 5] & 0xff) << 1) |
-                                ((bytes[offset_byte + 6] & 0x0f) << 9);
-        data[offset_data + 4] = (bytes[offset_byte + 6] >> 4 & (0x0f)) |
-                                ((bytes[offset_byte + 7] & 0xff) << 4) |
-                                ((bytes[offset_byte + 8] & 0x01) << 12);
-        data[offset_data + 5] = (bytes[offset_byte + 8] >> 1 & (0x7f)) |
-                                ((bytes[offset_byte + 9] & 0x3f) << 7);
-        data[offset_data + 6] = (bytes[offset_byte + 9] >> 6 & (0x03)) |
-                                ((bytes[offset_byte + 10] & 0xff) << 2) |
-                                ((bytes[offset_byte + 11] & 0x07) << 10);
-        data[offset_data + 7] = (bytes[offset_byte + 11] >> 3 & (0x1f)) |
-                                ((bytes[offset_byte + 12] & 0xff) << 5);
+void PQCLEAN_FIRESABER_CLEAN_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]) {
+    size_t i, j;
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            data->coeffs[j * 8 + i] = ((bytes[j] >> i) & 0x01);
+        }
     }
 }
 
-void PQCLEAN_FIRESABER_CLEAN_POLVEC2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus) {
-    if (modulus == 1024) {
-        POLVECp2BS(bytes, data);
-    } else if (modulus == 8192) {
-        POLVECq2BS(bytes, data);
-    }
-}
+void PQCLEAN_FIRESABER_CLEAN_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data) {
+    size_t i, j;
+    memset(bytes, 0, SABER_KEYBYTES);
 
-void PQCLEAN_FIRESABER_CLEAN_BS2POLVEC(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus) {
-    if (modulus == 1024) {
-        BS2POLVECp(bytes, data);
-    } else if (modulus == 8192) {
-        BS2POLVECq(bytes, data);
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            bytes[j] = bytes[j] | ((data->coeffs[j * 8 + i] & 0x01) << i);
+        }
     }
 }

crypto_kem/firesaber/clean/pack_unpack.h

@@ -1,28 +1,28 @@
 #ifndef PACK_UNPACK_H
 #define PACK_UNPACK_H
-
 #include "SABER_params.h"
+#include "poly.h"
 #include <stdint.h>
 #include <stdio.h>
 
+void PQCLEAN_FIRESABER_CLEAN_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data);
+
+void PQCLEAN_FIRESABER_CLEAN_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]);
 
-void PQCLEAN_FIRESABER_CLEAN_pack_3bit(uint8_t *bytes, const uint16_t *data);
 
-void PQCLEAN_FIRESABER_CLEAN_un_pack3bit(const uint8_t *bytes, uint16_t *data);
+void PQCLEAN_FIRESABER_CLEAN_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]);
 
-void PQCLEAN_FIRESABER_CLEAN_pack_4bit(uint8_t *bytes, const uint16_t *data);
+void PQCLEAN_FIRESABER_CLEAN_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]);
 
-void PQCLEAN_FIRESABER_CLEAN_un_pack4bit(const unsigned char *bytes, uint16_t *ar);
 
-void PQCLEAN_FIRESABER_CLEAN_pack_6bit(uint8_t *bytes, const uint16_t *data);
+void PQCLEAN_FIRESABER_CLEAN_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]);
 
-void PQCLEAN_FIRESABER_CLEAN_un_pack6bit(const unsigned char *bytes, uint16_t *data);
+void PQCLEAN_FIRESABER_CLEAN_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]);
 
 
-void PQCLEAN_FIRESABER_CLEAN_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]);
+void PQCLEAN_FIRESABER_CLEAN_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]);
 
-void PQCLEAN_FIRESABER_CLEAN_POLVEC2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus);
+void PQCLEAN_FIRESABER_CLEAN_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data);
 
-void PQCLEAN_FIRESABER_CLEAN_BS2POLVEC(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus);
 
 #endif

crypto_kem/firesaber/clean/poly.c

@@ -1,21 +1,57 @@
-/*---------------------------------------------------------------------
-This file has been adapted from the implementation
-(available at, Public Domain https://github.com/pq-crystals/kyber)
-of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
-by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
-Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
-#include "SABER_params.h"
+#include "api.h"
 #include "cbd.h"
 #include "fips202.h"
+#include "pack_unpack.h"
 #include "poly.h"
+#include <stddef.h>
 
-void PQCLEAN_FIRESABER_CLEAN_GenSecret(uint16_t r[SABER_K][SABER_N], const unsigned char *seed) {
-    uint8_t buf[SABER_MU * SABER_N * SABER_K / 8];
+void PQCLEAN_FIRESABER_CLEAN_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const poly s[SABER_L], int16_t transpose) {
+    size_t i, j;
+
+    if (transpose) {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_FIRESABER_CLEAN_poly_mul(&c[i], &A[0][i], &s[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_FIRESABER_CLEAN_poly_mul(&c[i], &A[j][i], &s[j], 1);
+            }
+        }
+    } else {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_FIRESABER_CLEAN_poly_mul(&c[i], &A[i][0], &s[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_FIRESABER_CLEAN_poly_mul(&c[i], &A[i][j], &s[j], 1);
+            }
+        }
+    }
+}
+
+void PQCLEAN_FIRESABER_CLEAN_InnerProd(poly *c, const poly b[SABER_L], const poly s[SABER_L]) {
+    size_t i;
+
+    PQCLEAN_FIRESABER_CLEAN_poly_mul(c, &b[0], &s[0], 0);
+    for (i = 1; i < SABER_L; i++) {
+        PQCLEAN_FIRESABER_CLEAN_poly_mul(c, &b[i], &s[i], 1);
+    }
+}
+
+void PQCLEAN_FIRESABER_CLEAN_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYVECBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_FIRESABER_CLEAN_BS2POLVECq(A[i], buf + i * SABER_POLYVECBYTES);
+    }
+}
+
+void PQCLEAN_FIRESABER_CLEAN_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYCOINBYTES];
 
     shake128(buf, sizeof(buf), seed, SABER_NOISESEEDBYTES);
 
-    for (size_t i = 0; i < SABER_K; i++) {
-        PQCLEAN_FIRESABER_CLEAN_cbd(r[i], buf + i * SABER_MU * SABER_N / 8);
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_FIRESABER_CLEAN_cbd(s[i].coeffs, buf + i * SABER_POLYCOINBYTES);
     }
 }

crypto_kem/firesaber/clean/poly.h

@@ -1,26 +1,23 @@
 #ifndef POLY_H
 #define POLY_H
-
-/*---------------------------------------------------------------------
-This file has been adapted from the implementation
-(available at, Public Domain https://github.com/pq-crystals/kyber)
-of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
-by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
-Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
-
-
 #include "SABER_params.h"
 #include <stdint.h>
 
-typedef struct {
+typedef union {
     uint16_t coeffs[SABER_N];
 } poly;
 
-typedef struct {
-    poly vec[SABER_K];
-} polyvec;
 
-void PQCLEAN_FIRESABER_CLEAN_GenSecret(uint16_t r[SABER_K][SABER_N], const unsigned char *seed);
+void PQCLEAN_FIRESABER_CLEAN_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const poly s[SABER_L], int16_t transpose);
+
+void PQCLEAN_FIRESABER_CLEAN_InnerProd(poly *c, const poly b[SABER_L], const poly s[SABER_L]);
+
+void PQCLEAN_FIRESABER_CLEAN_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]);
+
+void PQCLEAN_FIRESABER_CLEAN_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]);
+
+
+void PQCLEAN_FIRESABER_CLEAN_poly_mul(poly *c, const poly *a, const poly *b, int accumulate);
+
 
 #endif

crypto_kem/firesaber/clean/poly_mul.c

@@ -1,4 +1,4 @@
-#include "poly_mul.h"
+#include "poly.h"
 #include <stdint.h>
 #include <string.h>
 
@@ -11,13 +11,13 @@
 #define OVERFLOWING_MUL(X, Y) ((uint16_t)((uint32_t)(X) * (uint32_t)(Y)))
 
 #define KARATSUBA_N 64
-static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t *result_final) {
+static void karatsuba_simple(uint16_t *result_final, const uint16_t *a_1, const uint16_t *b_1) {
     uint16_t d01[KARATSUBA_N / 2 - 1];
     uint16_t d0123[KARATSUBA_N / 2 - 1];
     uint16_t d23[KARATSUBA_N / 2 - 1];
     uint16_t result_d01[KARATSUBA_N - 1];
 
-    int32_t i, j;
+    size_t i, j;
 
     memset(result_d01, 0, (KARATSUBA_N - 1)*sizeof(uint16_t));
     memset(d01, 0, (KARATSUBA_N / 2 - 1)*sizeof(uint16_t));
@@ -110,7 +110,7 @@ static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t
 
 
 
-static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *result) {
+static void toom_cook_4way (uint16_t *result, const uint16_t *a1, const uint16_t *b1) {
     uint16_t inv3 = 43691, inv9 = 36409, inv15 = 61167;
 
     uint16_t aw1[N_SB], aw2[N_SB], aw3[N_SB], aw4[N_SB], aw5[N_SB], aw6[N_SB], aw7[N_SB];
@@ -181,13 +181,13 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 
     // MULTIPLICATION
 
-    karatsuba_simple(aw1, bw1, w1);
-    karatsuba_simple(aw2, bw2, w2);
-    karatsuba_simple(aw3, bw3, w3);
-    karatsuba_simple(aw4, bw4, w4);
-    karatsuba_simple(aw5, bw5, w5);
-    karatsuba_simple(aw6, bw6, w6);
-    karatsuba_simple(aw7, bw7, w7);
+    karatsuba_simple(w1, aw1, bw1);
+    karatsuba_simple(w2, aw2, bw2);
+    karatsuba_simple(w3, aw3, bw3);
+    karatsuba_simple(w4, aw4, bw4);
+    karatsuba_simple(w5, aw5, bw5);
+    karatsuba_simple(w6, aw6, bw6);
+    karatsuba_simple(w7, aw7, bw7);
 
     // INTERPOLATION
     for (i = 0; i < N_SB_RES; ++i) {
@@ -228,19 +228,21 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
     }
 }
 
-void PQCLEAN_FIRESABER_CLEAN_pol_mul(uint16_t *a, uint16_t *b, uint16_t *res, uint16_t p, uint32_t n) {
-    uint32_t i;
-    // normal multiplication
-    uint16_t c[512];
+/* res += a*b */
+void PQCLEAN_FIRESABER_CLEAN_poly_mul(poly *c, const poly *a, const poly *b, const int accumulate) {
+    uint16_t C[2 * SABER_N] = {0};
+    size_t i;
 
-    for (i = 0; i < 512; i++) {
-        c[i] = 0;
-    }
-
-    toom_cook_4way(a, b, c);
+    toom_cook_4way(C, a->coeffs, b->coeffs);
 
-    // reduction
-    for (i = n; i < 2 * n; i++) {
-        res[i - n] = (c[i - n] - c[i]) & (p - 1);
+    /* reduction */
+    if (accumulate == 0) {
+        for (i = SABER_N; i < 2 * SABER_N; i++) {
+            c->coeffs[i - SABER_N] = (C[i - SABER_N] - C[i]);
+        }
+    } else {
+        for (i = SABER_N; i < 2 * SABER_N; i++) {
+            c->coeffs[i - SABER_N] += (C[i - SABER_N] - C[i]);
+        }
     }
 }

crypto_kem/firesaber/clean/poly_mul.h

@@ -1,9 +1,3 @@
-#ifndef POLYMUL_H
-#define POLYMUL_H
 
-#include "SABER_params.h"
-#include <stdint.h>
 
-void PQCLEAN_FIRESABER_CLEAN_pol_mul(uint16_t *a, uint16_t *b, uint16_t *res, uint16_t p, uint32_t n);
 
-#endif

crypto_kem/firesaber/clean/verify.c

@@ -1,3 +1,5 @@
+#include "verify.h"
+
 /*-------------------------------------------------
 This file has been adapted from the implementation
 (available at https://github.com/pq-crystals/kyber) of
@@ -5,26 +7,25 @@ This file has been adapted from the implementation
  by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------*/
-#include "verify.h"
-#include <stdint.h>
+
 
 /* returns 0 for equal strings, 1 for non-equal strings */
-unsigned char PQCLEAN_FIRESABER_CLEAN_verify(const unsigned char *a, const unsigned char *b, size_t len) {
+uint8_t PQCLEAN_FIRESABER_CLEAN_verify(const uint8_t *a, const uint8_t *b, size_t len) {
     uint64_t r;
     size_t i;
-
     r = 0;
+
     for (i = 0; i < len; i++) {
         r |= a[i] ^ b[i];
     }
 
     r = (~r + 1); // Two's complement
     r >>= 63;
-    return (unsigned char)r;
+    return (uint8_t) r;
 }
 
 /* b = 1 means mov, b = 0 means don't mov*/
-void PQCLEAN_FIRESABER_CLEAN_cmov(unsigned char *r, const unsigned char *x, size_t len, unsigned char b) {
+void PQCLEAN_FIRESABER_CLEAN_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b) {
     size_t i;
 
     b = -b;

crypto_kem/firesaber/clean/verify.h

@@ -1,6 +1,5 @@
 #ifndef VERIFY_H
 #define VERIFY_H
-
 /*-------------------------------------------------
 This file has been adapted from the implementation
 (available at https://github.com/pq-crystals/kyber) of
@@ -13,9 +12,11 @@ Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 #include <stdint.h>
 
 /* returns 0 for equal strings, 1 for non-equal strings */
-unsigned char PQCLEAN_FIRESABER_CLEAN_verify(const unsigned char *a, const unsigned char *b, size_t len);
+uint8_t PQCLEAN_FIRESABER_CLEAN_verify(const uint8_t *a, const uint8_t *b, size_t len);
+
 
 /* b = 1 means mov, b = 0 means don't mov*/
-void PQCLEAN_FIRESABER_CLEAN_cmov(unsigned char *r, const unsigned char *x, size_t len, unsigned char b);
+void PQCLEAN_FIRESABER_CLEAN_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b);
+
 
 #endif

crypto_kem/lightsaber/META.yml

@@ -14,4 +14,13 @@ principal-submitters:
   - Frederik Vercauteren
 implementations:
     - name: clean
-      version: https://github.com/KULeuven-COSIC/SABER/commit/14ede83f1ff3bcc41f0464543542366c68b55871
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/1ae84c3c/saber
+    - name: avx2
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/1ae84c3c/saber
+      supported_platforms:
+          - architecture: x86_64
+            operating_systems:
+                - Linux
+                - Darwin
+            required_flags:
+                - avx2

crypto_kem/lightsaber/avx2/LICENSE

@@ -0,0 +1 @@
+Public Domain

crypto_kem/lightsaber/avx2/Makefile

@@ -0,0 +1,22 @@
+# This Makefile can be used with GNU Make or BSD Make
+
+LIB=liblightsaber_avx2.a
+HEADERS=api.h cbd.h pack_unpack.h poly.h SABER_indcpa.h SABER_params.h verify.h 
+OBJECTS=cbd.o kem.o pack_unpack.o poly.o poly_mul.o SABER_indcpa.o verify.o 
+
+CFLAGS=-O3 -mavx2 -Wall -Wextra -Wpedantic -Wvla -Werror -Wredundant-decls -Wmissing-prototypes -std=c99 -I../../../common $(EXTRAFLAGS)
+
+all: $(LIB)
+
+%.o: %.s $(HEADERS)
+	$(AS) -o $@ $<
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_kem/lightsaber/avx2/SABER_indcpa.c

@@ -0,0 +1,125 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "fips202.h"
+#include "pack_unpack.h"
+#include "poly.h"
+#include "randombytes.h"
+#include <stdint.h>
+#include <string.h>
+
+#define h1 (1 << (SABER_EQ - SABER_EP - 1))
+#define h2 ((1 << (SABER_EP - 2)) - (1 << (SABER_EP - SABER_ET - 1)) + (1 << (SABER_EQ - SABER_EP - 1)))
+
+void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]) {
+    size_t i, j;
+
+    poly A[SABER_L][SABER_L];
+    poly *skpv1 = A[0]; // use first row of A to hold sk temporarily
+    toom4_points skpv1_eval[SABER_L];
+    poly res[SABER_L];
+
+    uint8_t rand[SABER_NOISESEEDBYTES];
+    uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+
+    randombytes(seed_A, SABER_SEEDBYTES);
+    shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
+
+    randombytes(rand, SABER_NOISESEEDBYTES);
+    PQCLEAN_LIGHTSABER_AVX2_GenSecret(skpv1, rand);
+    PQCLEAN_LIGHTSABER_AVX2_POLVECq2BS(sk, skpv1); // pack secret key
+
+    for (j = 0; j < SABER_L; j++) {
+        PQCLEAN_LIGHTSABER_AVX2_toom4_eval(&skpv1_eval[j], &skpv1[j]);
+    }
+
+    PQCLEAN_LIGHTSABER_AVX2_GenMatrix(A, seed_A); // sample matrix A
+    PQCLEAN_LIGHTSABER_AVX2_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const toom4_points *)skpv1_eval, 1); // Matrix in transposed order
+
+    // rounding
+    for (i = 0; i < SABER_L; i++) {
+        for (j = 0; j < SABER_N; j++) {
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
+        }
+    }
+
+    PQCLEAN_LIGHTSABER_AVX2_POLVECp2BS(pk, res); // pack public key
+}
+
+
+void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]) {
+    size_t i, j;
+
+    poly A[SABER_L][SABER_L];
+    poly res[SABER_L];
+    toom4_points skpv1_eval[SABER_L];
+
+    poly *temp = A[0]; // re-use stack space
+    poly *vprime = &A[0][0];
+    poly *message = &A[0][1];
+
+    const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+    uint8_t *msk_c = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+
+    PQCLEAN_LIGHTSABER_AVX2_GenSecret(temp, noiseseed);
+    for (j = 0; j < SABER_L; j++) {
+        PQCLEAN_LIGHTSABER_AVX2_toom4_eval(&skpv1_eval[j], &temp[j]);
+    }
+
+    PQCLEAN_LIGHTSABER_AVX2_GenMatrix(A, seed_A);
+    PQCLEAN_LIGHTSABER_AVX2_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const toom4_points *)skpv1_eval, 0); // 0 => not transposed
+
+    // rounding
+    for (i = 0; i < SABER_L; i++) { //shift right EQ-EP bits
+        for (j = 0; j < SABER_N; j++) {
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
+        }
+    }
+    PQCLEAN_LIGHTSABER_AVX2_POLVECp2BS(ciphertext, res);
+
+    // vector-vector scalar multiplication with mod p
+    PQCLEAN_LIGHTSABER_AVX2_BS2POLVECp(temp, pk);
+    PQCLEAN_LIGHTSABER_AVX2_InnerProd(vprime, temp, skpv1_eval);
+    PQCLEAN_LIGHTSABER_AVX2_BS2POLmsg(message, m);
+
+    for (i = 0; i < SABER_N; i++) {
+        vprime->coeffs[i] += h1 - (message->coeffs[i] << (SABER_EP - 1));
+        vprime->coeffs[i] &= SABER_P - 1;
+        vprime->coeffs[i] >>= SABER_EP - SABER_ET;
+    }
+
+    PQCLEAN_LIGHTSABER_AVX2_POLT2BS(msk_c, vprime);
+}
+
+
+void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]) {
+    size_t i;
+
+    poly temp[SABER_L];
+    toom4_points sksv_eval[SABER_L];
+
+    const uint8_t *packed_cm = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+    poly *v = &temp[0];
+    poly *cm = &temp[1];
+
+    PQCLEAN_LIGHTSABER_AVX2_BS2POLVECq(temp, sk);
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_LIGHTSABER_AVX2_toom4_eval(&sksv_eval[i], &temp[i]);
+    }
+
+    PQCLEAN_LIGHTSABER_AVX2_BS2POLVECp(temp, ciphertext);
+    PQCLEAN_LIGHTSABER_AVX2_InnerProd(v, temp, sksv_eval);
+
+    PQCLEAN_LIGHTSABER_AVX2_BS2POLT(cm, packed_cm);
+
+    for (i = 0; i < SABER_N; i++) {
+        v->coeffs[i] += h2 - (cm->coeffs[i] << (SABER_EP - SABER_ET));
+        v->coeffs[i] &= SABER_P - 1;
+        v->coeffs[i] >>= SABER_EP - 1;
+    }
+
+    PQCLEAN_LIGHTSABER_AVX2_POLmsg2BS(m, v);
+}

crypto_kem/lightsaber/avx2/SABER_indcpa.h

@@ -0,0 +1,13 @@
+#ifndef INDCPA_H
+#define INDCPA_H
+#include "SABER_params.h"
+#include <stdint.h>
+
+void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]);
+
+void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]);
+
+void PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]);
+
+
+#endif

crypto_kem/lightsaber/avx2/SABER_params.h

@@ -0,0 +1,41 @@
+#ifndef PARAMS_H
+#define PARAMS_H
+
+
+/* Don't change anything below this line */
+#define SABER_L 2
+#define SABER_MU 10
+#define SABER_ET 3
+
+#define SABER_N 256
+
+#define SABER_EP 10
+#define SABER_P (1 << SABER_EP)
+
+#define SABER_EQ 13
+#define SABER_Q (1 << SABER_EQ)
+
+#define SABER_SEEDBYTES 32
+#define SABER_NOISESEEDBYTES 32
+#define SABER_KEYBYTES 32
+#define SABER_HASHBYTES 32
+
+#define SABER_POLYCOINBYTES (SABER_MU * SABER_N / 8)
+
+#define SABER_POLYBYTES (SABER_EQ * SABER_N / 8)
+#define SABER_POLYVECBYTES (SABER_L * SABER_POLYBYTES)
+
+#define SABER_POLYCOMPRESSEDBYTES (SABER_EP * SABER_N / 8)
+#define SABER_POLYVECCOMPRESSEDBYTES (SABER_L * SABER_POLYCOMPRESSEDBYTES)
+
+#define SABER_SCALEBYTES_KEM (SABER_ET * SABER_N / 8)
+
+#define SABER_INDCPA_PUBLICKEYBYTES (SABER_POLYVECCOMPRESSEDBYTES + SABER_SEEDBYTES)
+#define SABER_INDCPA_SECRETKEYBYTES (SABER_POLYVECBYTES)
+
+#define SABER_PUBLICKEYBYTES (SABER_INDCPA_PUBLICKEYBYTES)
+#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES + SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
+
+#define SABER_BYTES_CCA_DEC (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM)
+
+#endif

crypto_kem/lightsaber/avx2/api.h

@@ -0,0 +1,18 @@
+#ifndef PQCLEAN_LIGHTSABER_AVX2_API_H
+#define PQCLEAN_LIGHTSABER_AVX2_API_H
+
+
+#define PQCLEAN_LIGHTSABER_AVX2_CRYPTO_ALGNAME "LightSaber"
+#define PQCLEAN_LIGHTSABER_AVX2_CRYPTO_BYTES 32
+#define PQCLEAN_LIGHTSABER_AVX2_CRYPTO_CIPHERTEXTBYTES 736
+#define PQCLEAN_LIGHTSABER_AVX2_CRYPTO_PUBLICKEYBYTES 672
+#define PQCLEAN_LIGHTSABER_AVX2_CRYPTO_SECRETKEYBYTES 1568
+
+int PQCLEAN_LIGHTSABER_AVX2_crypto_kem_keypair(unsigned char *pk, unsigned char *sk);
+
+int PQCLEAN_LIGHTSABER_AVX2_crypto_kem_enc(unsigned char *ct, unsigned char *k, const unsigned char *pk);
+
+int PQCLEAN_LIGHTSABER_AVX2_crypto_kem_dec(unsigned char *k, const unsigned char *ct, const unsigned char *sk);
+
+
+#endif /* PQCLEAN_LIGHTSABER_AVX2_API_H */

crypto_kem/lightsaber/avx2/cbd.c

@@ -0,0 +1,48 @@
+#include "SABER_params.h"
+#include "api.h"
+#include "cbd.h"
+#include <stdint.h>
+/*---------------------------------------------------------------------
+This file has been adapted from the implementation
+(available at, Public Domain https://github.com/pq-crystals/kyber)
+of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------------------------*/
+
+
+static uint64_t load_littleendian(const uint8_t *x, int bytes) {
+    int i;
+    uint64_t r = x[0];
+    for (i = 1; i < bytes; i++) {
+        r |= (uint64_t)x[i] << (8 * i);
+    }
+    return r;
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]) {
+    uint64_t t, d, a[4], b[4];
+    int i, j;
+
+    for (i = 0; i < SABER_N / 4; i++) {
+        t = load_littleendian(buf + 5 * i, 5);
+        d = 0;
+        for (j = 0; j < 5; j++) {
+            d += (t >> j) & 0x0842108421UL;
+        }
+
+        a[0] = d & 0x1f;
+        b[0] = (d >> 5) & 0x1f;
+        a[1] = (d >> 10) & 0x1f;
+        b[1] = (d >> 15) & 0x1f;
+        a[2] = (d >> 20) & 0x1f;
+        b[2] = (d >> 25) & 0x1f;
+        a[3] = (d >> 30) & 0x1f;
+        b[3] = (d >> 35);
+
+        s[4 * i + 0] = (uint16_t)(a[0] - b[0]);
+        s[4 * i + 1] = (uint16_t)(a[1] - b[1]);
+        s[4 * i + 2] = (uint16_t)(a[2] - b[2]);
+        s[4 * i + 3] = (uint16_t)(a[3] - b[3]);
+    }
+}

crypto_kem/lightsaber/avx2/cbd.h

@@ -0,0 +1,16 @@
+#ifndef CBD_H
+#define CBD_H
+/*---------------------------------------------------------------------
+This file has been adapted from the implementation
+(available at, Public Domain https://github.com/pq-crystals/kyber)
+of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------------------------*/
+#include "SABER_params.h"
+#include <stdint.h>
+
+void PQCLEAN_LIGHTSABER_AVX2_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]);
+
+
+#endif

crypto_kem/lightsaber/avx2/kem.c

@@ -0,0 +1,77 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "api.h"
+#include "fips202.h"
+#include "randombytes.h"
+#include "verify.h"
+#include <stddef.h>
+#include <stdint.h>
+
+
+int PQCLEAN_LIGHTSABER_AVX2_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
+    size_t i;
+
+    PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
+    for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
+        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
+    }
+
+    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES); // Then hash(pk) is appended.
+
+    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES); // Remaining part of sk contains a pseudo-random number.
+    // This is output when check in PQCLEAN_LIGHTSABER_AVX2_crypto_kem_dec() fails.
+    return (0);
+}
+
+int PQCLEAN_LIGHTSABER_AVX2_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t *pk) {
+
+    uint8_t kr[64]; // Will contain key, coins
+    uint8_t buf[64];
+
+    randombytes(buf, 32);
+
+    sha3_256(buf, buf, 32); // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
+
+    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES); // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
+
+    sha3_512(kr, buf, 64);               // kr[0:63] <-- Hash(buf[0:63]);
+    // K^ <-- kr[0:31]
+    // noiseseed (r) <-- kr[32:63];
+    PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_enc(c, buf, kr + 32, pk); // buf[0:31] contains message; kr[32:63] contains randomness r;
+
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC);
+
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}
+
+int PQCLEAN_LIGHTSABER_AVX2_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
+    size_t i;
+    uint8_t fail;
+    uint8_t cmp[SABER_BYTES_CCA_DEC];
+    uint8_t buf[64];
+    uint8_t kr[64]; // Will contain key, coins
+    const uint8_t *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
+
+    PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_dec(buf, sk, c); // buf[0:31] <-- message
+
+    // Multitarget countermeasure for coins + contributory KEM
+    for (i = 0; i < 32; i++) { // Save hash by storing h(pk) in sk
+        buf[32 + i] = sk[SABER_SECRETKEYBYTES - 64 + i];
+    }
+
+    sha3_512(kr, buf, 64);
+
+    PQCLEAN_LIGHTSABER_AVX2_indcpa_kem_enc(cmp, buf, kr + 32, pk);
+
+    fail = PQCLEAN_LIGHTSABER_AVX2_verify(c, cmp, SABER_BYTES_CCA_DEC);
+
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC); // overwrite coins in kr with h(c)
+
+    PQCLEAN_LIGHTSABER_AVX2_cmov(kr, sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES, fail);
+
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}

crypto_kem/lightsaber/avx2/pack_unpack.c

@@ -0,0 +1,155 @@
+#include "SABER_params.h"
+#include "pack_unpack.h"
+#include "poly.h"
+#include <string.h>
+
+void PQCLEAN_LIGHTSABER_AVX2_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (uint8_t) ((in[0] & 0x7) | ((in[1] & 0x7) << 3) | (in[2] << 6));
+        out[1] = (uint8_t) (((in[2] >> 2) & 0x01) | ((in[3] & 0x7) << 1) | ((in[4] & 0x7) << 4) | (in[5] << 7));
+        out[2] = (uint8_t) (((in[5] >> 1) & 0x03) | ((in[6] & 0x7) << 2) | (in[7] << 5));
+        in += 8;
+        out += 3;
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]) {
+    /* This function does not reduce its output mod T */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = in[0];
+        out[1] = in[0] >> 3;
+        out[2] = (in[0] >> 6) | (in[1] << 2);
+        out[3] = in[1] >> 1;
+        out[4] = in[1] >> 4;
+        out[5] = (in[1] >> 7) | (in[2] << 1);
+        out[6] = in[2] >> 2;
+        out[7] = in[2] >> 5;
+        in += 3;
+        out += 8;
+    }
+}
+
+static void POLq2BS(uint8_t bytes[SABER_POLYBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x1f) | (in[1] << 5));
+        out[2] = (uint8_t) (in[1] >> 3);
+        out[3] = (uint8_t) (((in[1] >> 11) & 0x03) | (in[2] << 2));
+        out[4] = (uint8_t) (((in[2] >> 6) & 0x7f) | (in[3] << 7));
+        out[5] = (uint8_t) (in[3] >> 1);
+        out[6] = (uint8_t) (((in[3] >> 9) & 0x0f) | (in[4] << 4));
+        out[7] = (uint8_t) (in[4] >> 4);
+        out[8] = (uint8_t) (((in[4] >> 12) & 0x01) | (in[5] << 1));
+        out[9] = (uint8_t) (((in[5] >> 7) & 0x3f) | (in[6] << 6));
+        out[10] = (uint8_t) (in[6] >> 2);
+        out[11] = (uint8_t) (((in[6] >> 10) & 0x07) | (in[7] << 3));
+        out[12] = (uint8_t) (in[7] >> 5);
+        in += 8;
+        out += 13;
+    }
+}
+
+static void BS2POLq(poly *data, const uint8_t bytes[SABER_POLYBYTES]) {
+    /* This function does not reduce its output mod Q */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (in[0]) | (in[1] << 8);
+        out[1] = (in[1] >> 5) | (in[2] << 3) | (in[3] << 11);
+        out[2] = (in[3] >> 2) | (in[4] << 6);
+        out[3] = (in[4] >> 7) | (in[5] << 1) | (in[6] << 9);
+        out[4] = (in[6] >> 4) | (in[7] << 4) | (in[8] << 12);
+        out[5] = (in[8] >> 1) | (in[9] << 7);
+        out[6] = (in[9] >> 6) | (in[10] << 2) | (in[11] << 10);
+        out[7] = (in[11] >> 3) | (in[12] << 5);
+        in += 13;
+        out += 8;
+    }
+}
+
+static void POLp2BS(uint8_t bytes[SABER_POLYCOMPRESSEDBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x03) | (in[1] << 2));
+        out[2] = (uint8_t) (((in[1] >> 6) & 0x0f) | (in[2] << 4));
+        out[3] = (uint8_t) (((in[2] >> 4) & 0x3f) | (in[3] << 6));
+        out[4] = (uint8_t) (in[3] >> 2);
+        in += 4;
+        out += 5;
+    }
+}
+
+static void BS2POLp(poly *data, const uint8_t bytes[SABER_POLYCOMPRESSEDBYTES]) {
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = in[0] | (in[1] << 8);
+        out[1] = (in[1] >> 2) | (in[2] << 6);
+        out[2] = (in[2] >> 4) | (in[3] << 4);
+        out[3] = (in[3] >> 6) | (in[4] << 2);
+        in += 5;
+        out += 4;
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLq2BS(bytes + i * SABER_POLYBYTES, &data[i]);
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLq(&data[i], bytes + i * SABER_POLYBYTES);
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLp2BS(bytes + i * SABER_POLYCOMPRESSEDBYTES, &data[i]);
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLp(&data[i], bytes + i * SABER_POLYCOMPRESSEDBYTES);
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]) {
+    size_t i, j;
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            data->coeffs[j * 8 + i] = ((bytes[j] >> i) & 0x01);
+        }
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data) {
+    size_t i, j;
+    memset(bytes, 0, SABER_KEYBYTES);
+
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            bytes[j] = bytes[j] | ((data->coeffs[j * 8 + i] & 0x01) << i);
+        }
+    }
+}

crypto_kem/lightsaber/avx2/pack_unpack.h

@@ -0,0 +1,28 @@
+#ifndef PACK_UNPACK_H
+#define PACK_UNPACK_H
+#include "SABER_params.h"
+#include "poly.h"
+#include <stdint.h>
+#include <stdio.h>
+
+void PQCLEAN_LIGHTSABER_AVX2_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data);
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]);
+
+
+void PQCLEAN_LIGHTSABER_AVX2_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]);
+
+void PQCLEAN_LIGHTSABER_AVX2_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]);
+
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]);
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]);
+
+
+void PQCLEAN_LIGHTSABER_AVX2_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]);
+
+void PQCLEAN_LIGHTSABER_AVX2_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data);
+
+
+#endif

crypto_kem/lightsaber/avx2/poly.c

@@ -0,0 +1,62 @@
+#include "cbd.h"
+#include "fips202.h"
+#include "pack_unpack.h"
+#include "poly.h"
+
+
+void PQCLEAN_LIGHTSABER_AVX2_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const toom4_points s_eval[SABER_L], int transpose) {
+    size_t i, j;
+    toom4_points_product c_eval;
+
+    if (transpose) {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_LIGHTSABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[0][i], &s_eval[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_LIGHTSABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[j][i], &s_eval[j], 1);
+            }
+            PQCLEAN_LIGHTSABER_AVX2_toom4_interp(&c[i], &c_eval);
+        }
+    } else {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_LIGHTSABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[i][0], &s_eval[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_LIGHTSABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[i][j], &s_eval[j], 1);
+            }
+            PQCLEAN_LIGHTSABER_AVX2_toom4_interp(&c[i], &c_eval);
+        }
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_InnerProd(poly *c, const poly b[SABER_L], const toom4_points s_eval[SABER_L]) {
+    size_t i;
+    toom4_points_product c_eval; //Holds results for 9 Karatsuba at a time
+
+    PQCLEAN_LIGHTSABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &b[0], &s_eval[0], 0);
+    for (i = 1; i < SABER_L; i++) {
+        PQCLEAN_LIGHTSABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &b[i], &s_eval[i], 1);
+    }
+
+    PQCLEAN_LIGHTSABER_AVX2_toom4_interp(c, &c_eval);
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYVECBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_LIGHTSABER_AVX2_BS2POLVECq(A[i], buf + i * SABER_POLYVECBYTES);
+    }
+}
+
+void PQCLEAN_LIGHTSABER_AVX2_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYCOINBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_NOISESEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_LIGHTSABER_AVX2_cbd(s[i].coeffs, buf + i * SABER_POLYCOINBYTES);
+    }
+}

crypto_kem/lightsaber/avx2/poly.h

@@ -0,0 +1,38 @@
+#ifndef POLY_H
+#define POLY_H
+#include "SABER_params.h"
+#include <immintrin.h>
+#include <stdint.h>
+
+typedef union {
+    uint16_t coeffs[SABER_N];
+    __m256i dummy;
+} poly;
+
+typedef union {
+    uint16_t coeffs[4 * SABER_N];
+    __m256i dummy;
+} toom4_points;
+
+typedef union {
+    uint16_t coeffs[8 * SABER_N];
+    __m256i dummy;
+} toom4_points_product;
+
+void PQCLEAN_LIGHTSABER_AVX2_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const toom4_points s_eval[SABER_L], int transpose);
+
+void PQCLEAN_LIGHTSABER_AVX2_InnerProd(poly *c, const poly b[SABER_L], const toom4_points s_eval[SABER_L]);
+
+void PQCLEAN_LIGHTSABER_AVX2_GenMatrix(poly a[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]);
+
+void PQCLEAN_LIGHTSABER_AVX2_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]);
+
+
+void PQCLEAN_LIGHTSABER_AVX2_toom4_interp(poly *res_avx, const toom4_points_product *c_eval);
+
+void PQCLEAN_LIGHTSABER_AVX2_toom4_eval(toom4_points *b_eval, const poly *b);
+
+void PQCLEAN_LIGHTSABER_AVX2_toom4_mul_A_by_B_eval(toom4_points_product *c_eval, const poly *a_avx, const toom4_points *b_eval, int accumulate);
+
+
+#endif

crypto_kem/lightsaber/avx2/verify.c

@@ -0,0 +1,35 @@
+#include "verify.h"
+
+/*-------------------------------------------------
+This file has been adapted from the implementation
+(available at https://github.com/pq-crystals/kyber) of
+"CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------*/
+
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+uint8_t PQCLEAN_LIGHTSABER_AVX2_verify(const uint8_t *a, const uint8_t *b, size_t len) {
+    uint64_t r;
+    size_t i;
+    r = 0;
+
+    for (i = 0; i < len; i++) {
+        r |= a[i] ^ b[i];
+    }
+
+    r = (~r + 1); // Two's complement
+    r >>= 63;
+    return (uint8_t) r;
+}
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_LIGHTSABER_AVX2_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b) {
+    size_t i;
+
+    b = -b;
+    for (i = 0; i < len; i++) {
+        r[i] ^= b & (x[i] ^ r[i]);
+    }
+}

crypto_kem/lightsaber/avx2/verify.h

@@ -0,0 +1,22 @@
+#ifndef VERIFY_H
+#define VERIFY_H
+/*-------------------------------------------------
+This file has been adapted from the implementation
+(available at https://github.com/pq-crystals/kyber) of
+"CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------*/
+
+#include <stddef.h>
+#include <stdint.h>
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+uint8_t PQCLEAN_LIGHTSABER_AVX2_verify(const uint8_t *a, const uint8_t *b, size_t len);
+
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_LIGHTSABER_AVX2_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b);
+
+
+#endif

crypto_kem/lightsaber/clean/LICENSE

@@ -1,8 +1 @@
-SABER_v1.1
-
-Public domain
-
-Authors: Jan-Pieter D'Anvers, Angshuman Karmakar, Sujoy Sinha Roy,
-Frederik Vercauteren
+Public Domain

crypto_kem/lightsaber/clean/Makefile

@@ -1,10 +1,10 @@
 # This Makefile can be used with GNU Make or BSD Make
 
 LIB=liblightsaber_clean.a
-HEADERS=api.h cbd.h poly.h poly_mul.h SABER_indcpa.h SABER_params.h verify.h pack_unpack.h 
+HEADERS=api.h cbd.h pack_unpack.h poly.h poly_mul.h SABER_indcpa.h SABER_params.h verify.h 
 OBJECTS=cbd.o kem.o pack_unpack.o poly.o poly_mul.o SABER_indcpa.o verify.o 
 
-CFLAGS=-O3 -Wall -Wextra -Wpedantic -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common $(EXTRAFLAGS)
+CFLAGS=-O3 -Wall -Wextra -Wpedantic -Wvla -Werror -Wredundant-decls -Wmissing-prototypes -std=c99 -I../../../common $(EXTRAFLAGS)
 
 all: $(LIB)
 

crypto_kem/lightsaber/clean/SABER_indcpa.c

@@ -3,296 +3,111 @@
 #include "fips202.h"
 #include "pack_unpack.h"
 #include "poly.h"
-#include "poly_mul.h"
 #include "randombytes.h"
 #include <stdint.h>
 #include <string.h>
 
+#define h1 (1 << (SABER_EQ - SABER_EP - 1))
+#define h2 ((1 << (SABER_EP - 2)) - (1 << (SABER_EP - SABER_ET - 1)) + (1 << (SABER_EQ - SABER_EP - 1)))
 
+void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]) {
+    size_t i, j;
 
-/*-----------------------------------------------------------------------------------
-    This routine generates a=[Matrix K x K] of 256-coefficient polynomials
+    poly A[SABER_L][SABER_L];
+    poly s[SABER_L];
+    poly res[SABER_L];
 
-#define h1 4 //2^(EQ-EP-1)
+    uint8_t rand[SABER_NOISESEEDBYTES];
+    uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
 
-#define h2 ( (1<<(SABER_EP-2)) - (1<<(SABER_EP-SABER_ET-1)) + (1<<(SABER_EQ-SABER_EP-1)) )
+    randombytes(seed_A, SABER_SEEDBYTES);
+    shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
 
-static void InnerProd(uint16_t pkcl[SABER_K][SABER_N], uint16_t skpv[SABER_K][SABER_N], uint16_t mod, uint16_t res[SABER_N]);
-static void MatrixVectorMul(polyvec *a, uint16_t skpv[SABER_K][SABER_N], uint16_t res[SABER_K][SABER_N], uint16_t mod, int16_t transpose);
+    randombytes(rand, SABER_NOISESEEDBYTES);
+    PQCLEAN_LIGHTSABER_CLEAN_GenSecret(s, rand);
+    PQCLEAN_LIGHTSABER_CLEAN_POLVECq2BS(sk, s);
 
-static void POL2MSG(const uint16_t *message_dec_unpacked, unsigned char *message_dec);
+    PQCLEAN_LIGHTSABER_CLEAN_GenMatrix(A, seed_A); // sample matrix A
+    PQCLEAN_LIGHTSABER_CLEAN_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const poly *)s, 1); // Matrix in transposed order
 
-static void GenMatrix(polyvec *a, const unsigned char *seed) {
-    unsigned char buf[SABER_K * SABER_K * (13 * SABER_N / 8)];
 
-    uint16_t temp_ar[SABER_N];
-
-    int i, j, k;
-    uint16_t mod = (SABER_Q - 1);
-
-    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
-
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_K; j++) {
-            PQCLEAN_LIGHTSABER_CLEAN_BS2POL(buf + (i * SABER_K + j) * (13 * SABER_N / 8), temp_ar);
-            for (k = 0; k < SABER_N; k++) {
-                a[i].vec[j].coeffs[k] = (temp_ar[k])& mod ;
-            }
-        }
-    }
-}
-
-
-void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(unsigned char *pk, unsigned char *sk) {
-    polyvec a[SABER_K];
-
-    uint16_t skpv[SABER_K][SABER_N];
-
-    unsigned char seed[SABER_SEEDBYTES];
-    unsigned char noiseseed[SABER_COINBYTES];
-    int32_t i, j;
-    uint16_t mod_q = SABER_Q - 1;
-
-
-    uint16_t res[SABER_K][SABER_N];
-
-    randombytes(seed, SABER_SEEDBYTES);
-
-    // for not revealing system RNG state
-    shake128(seed, SABER_SEEDBYTES, seed, SABER_SEEDBYTES);
-    randombytes(noiseseed, SABER_COINBYTES);
-
-    GenMatrix(a, seed);   //sample matrix A
-
-    // generate secret from constant-time binomial distribution
-    PQCLEAN_LIGHTSABER_CLEAN_GenSecret(skpv, noiseseed);
-
-    // do the matrix vector multiplication and rounding
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = 0;
-        }
-    }
-    MatrixVectorMul(a, skpv, res, SABER_Q - 1, 1);
-
-    // now rounding
-    for (i = 0; i < SABER_K; i++) {
+    // rounding
+    for (i = 0; i < SABER_L; i++) {
         for (j = 0; j < SABER_N; j++) {
-            // shift right 3 bits
-            res[i][j] = (res[i][j] + h1) & (mod_q);
-            res[i][j] = (res[i][j] >> (SABER_EQ - SABER_EP));
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
         }
     }
 
-    // unload and pack sk=3 x (256 coefficients of 14 bits)
-    PQCLEAN_LIGHTSABER_CLEAN_POLVEC2BS(sk, skpv, SABER_Q);
-
-    // unload and pack pk=256 bits seed and 3 x (256 coefficients of 11 bits)
-    // load the public-key coefficients
-    PQCLEAN_LIGHTSABER_CLEAN_POLVEC2BS(pk, res, SABER_P);
-
-
-    // now load the seedbytes in PK. Easy since seed bytes are kept in byte format.
-    for (i = 0; i < SABER_SEEDBYTES; i++) {
-        pk[SABER_POLYVECCOMPRESSEDBYTES + i] = seed[i];
-    }
-
+    PQCLEAN_LIGHTSABER_CLEAN_POLVECp2BS(pk, res); // pack public key
 }
 
 
-void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(const unsigned char *message_received, unsigned char *noiseseed, const unsigned char *pk, unsigned char *ciphertext) {
-    uint32_t i, j, k;
-    polyvec a[SABER_K];
-    unsigned char seed[SABER_SEEDBYTES];
-    // public key of received by the client
-    uint16_t pkcl[SABER_K][SABER_N];
-    uint16_t skpv1[SABER_K][SABER_N];
-    uint16_t message[SABER_KEYBYTES * 8];
-    uint16_t res[SABER_K][SABER_N];
-    uint16_t mod_p = SABER_P - 1;
-    uint16_t mod_q = SABER_Q - 1;
-    uint16_t vprime[SABER_N];
-    unsigned char msk_c[SABER_SCALEBYTES_KEM];
+void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]) {
+    size_t i, j;
 
-    // extract the seedbytes from Public Key.
-    for (i = 0; i < SABER_SEEDBYTES; i++) {
-        seed[i] = pk[ SABER_POLYVECCOMPRESSEDBYTES + i];
-    }
+    poly A[SABER_L][SABER_L];
+    poly res[SABER_L];
+    poly s[SABER_L];
+    poly *temp = A[0]; // re-use stack space
+    poly *vprime = &A[0][0];
+    poly *message = &A[0][1];
 
-    GenMatrix(a, seed);
+    const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+    uint8_t *msk_c = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
 
-    // generate secret from constant-time binomial distribution
-    PQCLEAN_LIGHTSABER_CLEAN_GenSecret(skpv1, noiseseed);
+    PQCLEAN_LIGHTSABER_CLEAN_GenSecret(s, noiseseed);
+    PQCLEAN_LIGHTSABER_CLEAN_GenMatrix(A, seed_A);
+    PQCLEAN_LIGHTSABER_CLEAN_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const poly *)s, 0); // 0 => not transposed
 
-    // matrix-vector multiplication and rounding
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = 0;
-        }
-    }
-    MatrixVectorMul(a, skpv1, res, SABER_Q - 1, 0);
-
-    // now rounding
-    //shift right 3 bits
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = ( res[i][j] + h1 ) & mod_q;
-            res[i][j] = (res[i][j] >> (SABER_EQ - SABER_EP) );
-        }
-    }
 
-    PQCLEAN_LIGHTSABER_CLEAN_POLVEC2BS(ciphertext, res, SABER_P);
-
-    // ************client matrix-vector multiplication ends************
-
-    // now calculate the v'
-    // unpack the public_key
-    // pkcl is the b in the protocol
-    PQCLEAN_LIGHTSABER_CLEAN_BS2POLVEC(pk, pkcl, SABER_P);
-    for (i = 0; i < SABER_N; i++) {
-        vprime[i] = 0;
-    }
-    for (i = 0; i < SABER_K; i++) {
+    // rounding
+    for (i = 0; i < SABER_L; i++) { //shift right EQ-EP bits
         for (j = 0; j < SABER_N; j++) {
-            skpv1[i][j] = skpv1[i][j] & (mod_p);
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
         }
     }
+    PQCLEAN_LIGHTSABER_CLEAN_POLVECp2BS(ciphertext, res);
 
     // vector-vector scalar multiplication with mod p
-    InnerProd(pkcl, skpv1, mod_p, vprime);
+    PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECp(temp, pk);
+    PQCLEAN_LIGHTSABER_CLEAN_InnerProd(vprime, temp, s);
+    PQCLEAN_LIGHTSABER_CLEAN_BS2POLmsg(message, m);
 
-    // addition of h1 to vprime
     for (i = 0; i < SABER_N; i++) {
-        vprime[i] = vprime[i] + h1;
+        vprime->coeffs[i] += h1 - (message->coeffs[i] << (SABER_EP - 1));
+        vprime->coeffs[i] &= SABER_P - 1;
+        vprime->coeffs[i] >>= SABER_EP - SABER_ET;
     }
 
-    // unpack message_received;
-    for (j = 0; j < SABER_KEYBYTES; j++) {
-        for (i = 0; i < 8; i++) {
-            message[8 * j + i] = ((message_received[j] >> i) & 0x01);
-        }
-    }
-
-    // message encoding
-    for (i = 0; i < SABER_N; i++) {
-        message[i] = (message[i] << (SABER_EP - 1));
-    }
-
-    for (k = 0; k < SABER_N; k++) {
-        vprime[k] = ( (vprime[k] - message[k]) & (mod_p) ) >> (SABER_EP - SABER_ET);
-    }
-
-
-    PQCLEAN_LIGHTSABER_CLEAN_pack_3bit(msk_c, vprime);
-
-    for (j = 0; j < SABER_SCALEBYTES_KEM; j++) {
-        ciphertext[SABER_POLYVECCOMPRESSEDBYTES + j] = msk_c[j];
-    }
+    PQCLEAN_LIGHTSABER_CLEAN_POLT2BS(msk_c, vprime);
 }
 
 
-void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_dec(const unsigned char *sk, const unsigned char *ciphertext, unsigned char message_dec[]) {
-    uint32_t i, j;
-    // secret key of the server
-    uint16_t sksv[SABER_K][SABER_N];
-    uint16_t pksv[SABER_K][SABER_N];
-    uint8_t scale_ar[SABER_SCALEBYTES_KEM];
-    uint16_t mod_p = SABER_P - 1;
-    uint16_t v[SABER_N];
-    uint16_t op[SABER_N];
+void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]) {
+    size_t i;
 
-    // sksv is the secret-key
-    PQCLEAN_LIGHTSABER_CLEAN_BS2POLVEC(sk, sksv, SABER_Q);
-    // pksv is the ciphertext
-    PQCLEAN_LIGHTSABER_CLEAN_BS2POLVEC(ciphertext, pksv, SABER_P);
+    poly temp[SABER_L];
+    poly s[SABER_L];
 
-    // vector-vector scalar multiplication with mod p
-    for (i = 0; i < SABER_N; i++) {
-        v[i] = 0;
-    }
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            sksv[i][j] = sksv[i][j] & (mod_p);
-        }
-    }
-    InnerProd(pksv, sksv, mod_p, v);
+    const uint8_t *packed_cm = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+    poly *v = &temp[0];
+    poly *cm = &temp[1];
 
-    //Extraction
-    for (i = 0; i < SABER_SCALEBYTES_KEM; i++) {
-        scale_ar[i] = ciphertext[SABER_POLYVECCOMPRESSEDBYTES + i];
-    }
+    PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECq(s, sk);
+    PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECp(temp, ciphertext);
+    PQCLEAN_LIGHTSABER_CLEAN_InnerProd(&temp[0], temp, s);
 
-    PQCLEAN_LIGHTSABER_CLEAN_un_pack3bit(scale_ar, op);
+    PQCLEAN_LIGHTSABER_CLEAN_BS2POLT(cm, packed_cm);
 
-    //addition of h1
     for (i = 0; i < SABER_N; i++) {
-        v[i] = ( ( v[i] + h2 - (op[i] << (SABER_EP - SABER_ET)) ) & (mod_p) ) >> (SABER_EP - 1);
+        v->coeffs[i] += h2 - (cm->coeffs[i] << (SABER_EP - SABER_ET));
+        v->coeffs[i] &= SABER_P - 1;
+        v->coeffs[i] >>= SABER_EP - 1;
     }
 
-    // pack decrypted message
-    POL2MSG(v, message_dec);
-}
-static void MatrixVectorMul(polyvec *a, uint16_t skpv[SABER_K][SABER_N], uint16_t res[SABER_K][SABER_N], uint16_t mod, int16_t transpose) {
-    uint16_t acc[SABER_N];
-    int32_t i, j, k;
-
-    if (transpose == 1) {
-        for (i = 0; i < SABER_K; i++) {
-            for (j = 0; j < SABER_K; j++) {
-                PQCLEAN_LIGHTSABER_CLEAN_pol_mul((uint16_t *)&a[j].vec[i], skpv[j], acc, SABER_Q, SABER_N);
-
-                for (k = 0; k < SABER_N; k++) {
-                    res[i][k] = res[i][k] + acc[k];
-                    //reduction mod p
-                    res[i][k] = (res[i][k] & mod);
-                    //clear the accumulator
-                    acc[k] = 0;
-                }
-            }
-        }
-    } else {
-        for (i = 0; i < SABER_K; i++) {
-            for (j = 0; j < SABER_K; j++) {
-                PQCLEAN_LIGHTSABER_CLEAN_pol_mul((uint16_t *)&a[i].vec[j], skpv[j], acc, SABER_Q, SABER_N);
-                for (k = 0; k < SABER_N; k++) {
-                    res[i][k] = res[i][k] + acc[k];
-                    // reduction
-                    res[i][k] = res[i][k] & mod;
-                    // clear the accumulator
-                    acc[k] = 0;
-                }
-            }
-        }
-    }
-}
-
-static void POL2MSG(const uint16_t *message_dec_unpacked, unsigned char *message_dec) {
-    int32_t i, j;
-
-    for (j = 0; j < SABER_KEYBYTES; j++) {
-        message_dec[j] = 0;
-        for (i = 0; i < 8; i++) {
-            message_dec[j] = message_dec[j] | (uint8_t) (message_dec_unpacked[j * 8 + i] << i);
-        }
-    }
-}
-
-
-static void InnerProd(uint16_t pkcl[SABER_K][SABER_N], uint16_t skpv[SABER_K][SABER_N], uint16_t mod, uint16_t res[SABER_N]) {
-    uint32_t j, k;
-    uint16_t acc[SABER_N];
-
-    // vector-vector scalar multiplication with mod p
-    for (j = 0; j < SABER_K; j++) {
-        PQCLEAN_LIGHTSABER_CLEAN_pol_mul(pkcl[j], skpv[j], acc, SABER_P, SABER_N);
-
-        for (k = 0; k < SABER_N; k++) {
-            res[k] = res[k] + acc[k];
-            // reduction
-            res[k] = res[k] & mod;
-            // clear the accumulator
-            acc[k] = 0;
-        }
-    }
+    PQCLEAN_LIGHTSABER_CLEAN_POLmsg2BS(m, v);
 }

crypto_kem/lightsaber/clean/SABER_indcpa.h

@@ -1,9 +1,13 @@
 #ifndef INDCPA_H
 #define INDCPA_H
+#include "SABER_params.h"
+#include <stdint.h>
 
-void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(unsigned char *pk, unsigned char *sk);
-void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(const unsigned char *message, unsigned char *noiseseed, const unsigned char *pk, unsigned char *ciphertext);
-void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_dec(const unsigned char *sk, const unsigned char *ciphertext, unsigned char *message_dec);
+void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]);
 
-#endif
+void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]);
+
+void PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]);
 
+
+#endif

crypto_kem/lightsaber/clean/SABER_params.h

@@ -1,50 +1,41 @@
 #ifndef PARAMS_H
 #define PARAMS_H
 
-#include "api.h"
 
-#define SABER_K 2
+/* Don't change anything below this line */
+#define SABER_L 2
 #define SABER_MU 10
 #define SABER_ET 3
 
-
-#define SABER_EQ 13
-#define SABER_EP 10
-
 #define SABER_N 256
-#define SABER_Q 8192
-#define SABER_P 1024
-
-#define SABER_SEEDBYTES       32
-#define SABER_NOISESEEDBYTES  32
-#define SABER_COINBYTES       32
-#define SABER_KEYBYTES        32
 
-#define SABER_HASHBYTES       32
+#define SABER_EP 10
+#define SABER_P (1 << SABER_EP)
 
-#define SABER_POLYBYTES       416 //13*256/8 
+#define SABER_EQ 13
+#define SABER_Q (1 << SABER_EQ)
 
-#define SABER_POLYVECBYTES    (SABER_K * SABER_POLYBYTES)
+#define SABER_SEEDBYTES 32
+#define SABER_NOISESEEDBYTES 32
+#define SABER_KEYBYTES 32
+#define SABER_HASHBYTES 32
 
-#define SABER_POLYVECCOMPRESSEDBYTES (SABER_K * 320) //10*256/8 NOTE : changed till here due to parameter adaptation
+#define SABER_POLYCOINBYTES (SABER_MU * SABER_N / 8)
 
-#define SABER_CIPHERTEXTBYTES (SABER_POLYVECCOMPRESSEDBYTES)
+#define SABER_POLYBYTES (SABER_EQ * SABER_N / 8)
+#define SABER_POLYVECBYTES (SABER_L * SABER_POLYBYTES)
 
-#define SABER_SCALEBYTES (SABER_DELTA*SABER_N/8)
+#define SABER_POLYCOMPRESSEDBYTES (SABER_EP * SABER_N / 8)
+#define SABER_POLYVECCOMPRESSEDBYTES (SABER_L * SABER_POLYCOMPRESSEDBYTES)
 
-#define SABER_SCALEBYTES_KEM ((SABER_ET)*SABER_N/8)
+#define SABER_SCALEBYTES_KEM (SABER_ET * SABER_N / 8)
 
 #define SABER_INDCPA_PUBLICKEYBYTES (SABER_POLYVECCOMPRESSEDBYTES + SABER_SEEDBYTES)
 #define SABER_INDCPA_SECRETKEYBYTES (SABER_POLYVECBYTES)
 
 #define SABER_PUBLICKEYBYTES (SABER_INDCPA_PUBLICKEYBYTES)
+#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES + SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
 
-#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES +  SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
-
-#define SABER_BYTES_CCA_DEC   (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM) /* Second part is for Targhi-Unruh */
-
-
-
+#define SABER_BYTES_CCA_DEC (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM)
 
 #endif
-

crypto_kem/lightsaber/clean/api.h

@@ -1,14 +1,18 @@
 #ifndef PQCLEAN_LIGHTSABER_CLEAN_API_H
 #define PQCLEAN_LIGHTSABER_CLEAN_API_H
 
+
 #define PQCLEAN_LIGHTSABER_CLEAN_CRYPTO_ALGNAME "LightSaber"
-#define PQCLEAN_LIGHTSABER_CLEAN_CRYPTO_SECRETKEYBYTES 1568
-#define PQCLEAN_LIGHTSABER_CLEAN_CRYPTO_PUBLICKEYBYTES (2*320+32)
 #define PQCLEAN_LIGHTSABER_CLEAN_CRYPTO_BYTES 32
 #define PQCLEAN_LIGHTSABER_CLEAN_CRYPTO_CIPHERTEXTBYTES 736
+#define PQCLEAN_LIGHTSABER_CLEAN_CRYPTO_PUBLICKEYBYTES 672
+#define PQCLEAN_LIGHTSABER_CLEAN_CRYPTO_SECRETKEYBYTES 1568
 
 int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_keypair(unsigned char *pk, unsigned char *sk);
-int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *ss, const unsigned char *pk);
-int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_dec(unsigned char *ss, const unsigned char *ct, const unsigned char *sk);
 
-#endif /* api_h */
+int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *k, const unsigned char *pk);
+
+int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_dec(unsigned char *k, const unsigned char *ct, const unsigned char *sk);
+
+
+#endif /* PQCLEAN_LIGHTSABER_CLEAN_API_H */

crypto_kem/lightsaber/clean/cbd.c

@@ -1,3 +1,7 @@
+#include "SABER_params.h"
+#include "api.h"
+#include "cbd.h"
+#include <stdint.h>
 /*---------------------------------------------------------------------
 This file has been adapted from the implementation
 (available at, Public Domain https://github.com/pq-crystals/kyber)
@@ -6,12 +10,8 @@ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------------------------*/
 
-#include "SABER_params.h"
-#include "api.h"
-#include "cbd.h"
-#include <stdint.h>
 
-static uint64_t load_littleendian(const unsigned char *x, int bytes) {
+static uint64_t load_littleendian(const uint8_t *x, int bytes) {
     int i;
     uint64_t r = x[0];
     for (i = 1; i < bytes; i++) {
@@ -20,10 +20,7 @@ static uint64_t load_littleendian(const unsigned char *x, int bytes) {
     return r;
 }
 
-
-void PQCLEAN_LIGHTSABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
-    uint16_t Qmod_minus1 = SABER_Q - 1;
-
+void PQCLEAN_LIGHTSABER_CLEAN_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]) {
     uint64_t t, d, a[4], b[4];
     int i, j;
 
@@ -34,8 +31,8 @@ void PQCLEAN_LIGHTSABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
             d += (t >> j) & 0x0842108421UL;
         }
 
-        a[0] =  d & 0x1f;
-        b[0] = (d >>  5) & 0x1f;
+        a[0] = d & 0x1f;
+        b[0] = (d >> 5) & 0x1f;
         a[1] = (d >> 10) & 0x1f;
         b[1] = (d >> 15) & 0x1f;
         a[2] = (d >> 20) & 0x1f;
@@ -43,9 +40,9 @@ void PQCLEAN_LIGHTSABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
         a[3] = (d >> 30) & 0x1f;
         b[3] = (d >> 35);
 
-        r[4 * i + 0] = (uint16_t)(a[0]  - b[0]) & Qmod_minus1;
-        r[4 * i + 1] = (uint16_t)(a[1]  - b[1]) & Qmod_minus1;
-        r[4 * i + 2] = (uint16_t)(a[2]  - b[2]) & Qmod_minus1;
-        r[4 * i + 3] = (uint16_t)(a[3]  - b[3]) & Qmod_minus1;
+        s[4 * i + 0] = (uint16_t)(a[0] - b[0]);
+        s[4 * i + 1] = (uint16_t)(a[1] - b[1]);
+        s[4 * i + 2] = (uint16_t)(a[2] - b[2]);
+        s[4 * i + 3] = (uint16_t)(a[3] - b[3]);
     }
 }

crypto_kem/lightsaber/clean/cbd.h

@@ -1,6 +1,5 @@
 #ifndef CBD_H
 #define CBD_H
-
 /*---------------------------------------------------------------------
 This file has been adapted from the implementation
 (available at, Public Domain https://github.com/pq-crystals/kyber)
@@ -8,10 +7,10 @@ of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
 by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------------------------*/
-
-#include "poly.h"
+#include "SABER_params.h"
 #include <stdint.h>
 
-void PQCLEAN_LIGHTSABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf);
+void PQCLEAN_LIGHTSABER_CLEAN_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]);
+
 
 #endif

crypto_kem/lightsaber/clean/kem.c

@@ -1,96 +1,77 @@
 #include "SABER_indcpa.h"
 #include "SABER_params.h"
+#include "api.h"
 #include "fips202.h"
 #include "randombytes.h"
 #include "verify.h"
+#include <stddef.h>
 #include <stdint.h>
-#include <stdio.h>
-#include <string.h>
 
-int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_keypair(unsigned char *pk, unsigned char *sk) {
-    int i;
 
-    // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
-    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(pk, sk);
+int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
+    size_t i;
 
-    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
+    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
     for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
-        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];
+        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
     }
 
-    // Then hash(pk) is appended.
-    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES);
+    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES); // Then hash(pk) is appended.
 
-    // Remaining part of sk contains a pseudo-random number.
-    // This is output when check in crypto_kem_dec() fails.
-    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES );
+    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES); // Remaining part of sk contains a pseudo-random number.
+    // This is output when check in PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_dec() fails.
     return (0);
 }
 
-int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *ss, const unsigned char *pk) {
-    // Will contain key, coins
-    unsigned char kr[64];
-    unsigned char buf[64];
+int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t *pk) {
 
-    randombytes(buf, 32);
+    uint8_t kr[64]; // Will contain key, coins
+    uint8_t buf[64];
 
-    // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
-    sha3_256(buf, buf, 32);
+    randombytes(buf, 32);
 
-    // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
-    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES);
+    sha3_256(buf, buf, 32); // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
 
-    // kr[0:63] <-- Hash(buf[0:63]);
-    sha3_512(kr, buf, 64);
+    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES); // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
 
+    sha3_512(kr, buf, 64);               // kr[0:63] <-- Hash(buf[0:63]);
     // K^ <-- kr[0:31]
     // noiseseed (r) <-- kr[32:63];
-    // buf[0:31] contains message; kr[32:63] contains randomness r;
-    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(buf, kr + 32, pk,  ct);
+    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(c, buf, kr + 32, pk); // buf[0:31] contains message; kr[32:63] contains randomness r;
 
-    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC);
 
-    // hash concatenation of pre-k and h(c) to k
-    sha3_256(ss, kr, 64);
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
 
     return (0);
 }
 
+int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
+    size_t i;
+    uint8_t fail;
+    uint8_t cmp[SABER_BYTES_CCA_DEC];
+    uint8_t buf[64];
+    uint8_t kr[64]; // Will contain key, coins
+    const uint8_t *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
 
-int PQCLEAN_LIGHTSABER_CLEAN_crypto_kem_dec(unsigned char *ss, const unsigned char *ct, const unsigned char *sk) {
-    int i;
-    unsigned char fail;
-    unsigned char cmp[SABER_BYTES_CCA_DEC];
-    unsigned char buf[64];
-
-    // Will contain key, coins
-    unsigned char kr[64];
-    const unsigned char *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
-
-    // buf[0:31] <-- message
-    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_dec(sk, ct, buf);
-
+    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_dec(buf, sk, c); // buf[0:31] <-- message
 
     // Multitarget countermeasure for coins + contributory KEM
-    // Save hash by storing h(pk) in sk
-    for (i = 0; i < 32; i++) {
+    for (i = 0; i < 32; i++) { // Save hash by storing h(pk) in sk
         buf[32 + i] = sk[SABER_SECRETKEYBYTES - 64 + i];
     }
 
     sha3_512(kr, buf, 64);
 
-    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(buf, kr + 32, pk, cmp);
-
+    PQCLEAN_LIGHTSABER_CLEAN_indcpa_kem_enc(cmp, buf, kr + 32, pk);
 
-    fail = PQCLEAN_LIGHTSABER_CLEAN_verify(ct, cmp, SABER_BYTES_CCA_DEC);
+    fail = PQCLEAN_LIGHTSABER_CLEAN_verify(c, cmp, SABER_BYTES_CCA_DEC);
 
-    // overwrite coins in kr with h(c)
-    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC); // overwrite coins in kr with h(c)
 
     PQCLEAN_LIGHTSABER_CLEAN_cmov(kr, sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES, fail);
 
-    // hash concatenation of pre-k and h(c) to k
-    sha3_256(ss, kr, 64);
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
 
     return (0);
 }

crypto_kem/lightsaber/clean/pack_unpack.c

@@ -1,254 +1,155 @@
+#include "SABER_params.h"
 #include "pack_unpack.h"
+#include "poly.h"
+#include <string.h>
 
-void PQCLEAN_LIGHTSABER_CLEAN_pack_3bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+void PQCLEAN_LIGHTSABER_CLEAN_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
     for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 3 * j;
-        offset_data = 8 * j;
-        bytes[offset_byte + 0] = (data[offset_data + 0] & 0x7) |
-                                 ((data[offset_data + 1] & 0x7) << 3) |
-                                 ((data[offset_data + 2] & 0x3) << 6);
-        bytes[offset_byte + 1] = ((data[offset_data + 2] >> 2 ) & 0x01)  |
-                                 ((data[offset_data + 3] & 0x7) << 1) |
-                                 ((data[offset_data + 4] & 0x7) << 4) |
-                                 (((data[offset_data + 5]) & 0x01) << 7);
-        bytes[offset_byte + 2] = ((data[offset_data + 5] >> 1 ) & 0x03) |
-                                 ((data[offset_data + 6] & 0x7) << 2) |
-                                 ((data[offset_data + 7] & 0x7) << 5);
+        out[0] = (uint8_t) ((in[0] & 0x7) | ((in[1] & 0x7) << 3) | (in[2] << 6));
+        out[1] = (uint8_t) (((in[2] >> 2) & 0x01) | ((in[3] & 0x7) << 1) | ((in[4] & 0x7) << 4) | (in[5] << 7));
+        out[2] = (uint8_t) (((in[5] >> 1) & 0x03) | ((in[6] & 0x7) << 2) | (in[7] << 5));
+        in += 8;
+        out += 3;
     }
 }
 
-void PQCLEAN_LIGHTSABER_CLEAN_un_pack3bit(const uint8_t *bytes, uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]) {
+    /* This function does not reduce its output mod T */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
     for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 3 * j;
-        offset_data = 8 * j;
-        data[offset_data + 0] = (bytes[offset_byte + 0]) & 0x07;
-        data[offset_data + 1] = ((bytes[offset_byte + 0]) >> 3 ) & 0x07;
-        data[offset_data + 2] = (((bytes[offset_byte + 0]) >> 6 ) & 0x03) |
-                                (((bytes[offset_byte + 1]) & 0x01) << 2);
-        data[offset_data + 3] = ((bytes[offset_byte + 1]) >> 1 ) & 0x07;
-        data[offset_data + 4] = ((bytes[offset_byte + 1]) >> 4 ) & 0x07;
-        data[offset_data + 5] = (((bytes[offset_byte + 1]) >> 7 ) & 0x01) |
-                                (((bytes[offset_byte + 2]) & 0x03) << 1);
-        data[offset_data + 6] = ((bytes[offset_byte + 2] >> 2) & 0x07);
-        data[offset_data + 7] = ((bytes[offset_byte + 2] >> 5) & 0x07);
+        out[0] = in[0];
+        out[1] = in[0] >> 3;
+        out[2] = (in[0] >> 6) | (in[1] << 2);
+        out[3] = in[1] >> 1;
+        out[4] = in[1] >> 4;
+        out[5] = (in[1] >> 7) | (in[2] << 1);
+        out[6] = in[2] >> 2;
+        out[7] = in[2] >> 5;
+        in += 3;
+        out += 8;
     }
 }
 
-void PQCLEAN_LIGHTSABER_CLEAN_pack_4bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data;
-
-    for (j = 0; j < SABER_N / 2; j++) {
-        offset_data = 2 * j;
-        bytes[j] = (data[offset_data] & 0x0f) |
-                   ((data[offset_data + 1] & 0x0f) << 4);
+static void POLq2BS(uint8_t bytes[SABER_POLYBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x1f) | (in[1] << 5));
+        out[2] = (uint8_t) (in[1] >> 3);
+        out[3] = (uint8_t) (((in[1] >> 11) & 0x03) | (in[2] << 2));
+        out[4] = (uint8_t) (((in[2] >> 6) & 0x7f) | (in[3] << 7));
+        out[5] = (uint8_t) (in[3] >> 1);
+        out[6] = (uint8_t) (((in[3] >> 9) & 0x0f) | (in[4] << 4));
+        out[7] = (uint8_t) (in[4] >> 4);
+        out[8] = (uint8_t) (((in[4] >> 12) & 0x01) | (in[5] << 1));
+        out[9] = (uint8_t) (((in[5] >> 7) & 0x3f) | (in[6] << 6));
+        out[10] = (uint8_t) (in[6] >> 2);
+        out[11] = (uint8_t) (((in[6] >> 10) & 0x07) | (in[7] << 3));
+        out[12] = (uint8_t) (in[7] >> 5);
+        in += 8;
+        out += 13;
     }
 }
 
-void PQCLEAN_LIGHTSABER_CLEAN_un_pack4bit(const unsigned char *bytes, uint16_t *ar) {
-    uint32_t j;
-    uint32_t offset_data;
-
-    for (j = 0; j < SABER_N / 2; j++) {
-        offset_data = 2 * j;
-        ar[offset_data] = bytes[j] & 0x0f;
-        ar[offset_data + 1] = (bytes[j] >> 4) & 0x0f;
+static void BS2POLq(poly *data, const uint8_t bytes[SABER_POLYBYTES]) {
+    /* This function does not reduce its output mod Q */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (in[0]) | (in[1] << 8);
+        out[1] = (in[1] >> 5) | (in[2] << 3) | (in[3] << 11);
+        out[2] = (in[3] >> 2) | (in[4] << 6);
+        out[3] = (in[4] >> 7) | (in[5] << 1) | (in[6] << 9);
+        out[4] = (in[6] >> 4) | (in[7] << 4) | (in[8] << 12);
+        out[5] = (in[8] >> 1) | (in[9] << 7);
+        out[6] = (in[9] >> 6) | (in[10] << 2) | (in[11] << 10);
+        out[7] = (in[11] >> 3) | (in[12] << 5);
+        in += 13;
+        out += 8;
     }
 }
 
-void PQCLEAN_LIGHTSABER_CLEAN_pack_6bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+static void POLp2BS(uint8_t bytes[SABER_POLYCOMPRESSEDBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
     for (j = 0; j < SABER_N / 4; j++) {
-        offset_byte = 3 * j;
-        offset_data = 4 * j;
-        bytes[offset_byte + 0] = (data[offset_data + 0] & 0x3f) |
-                                 ((data[offset_data + 1] & 0x03) << 6);
-        bytes[offset_byte + 1] = ((data[offset_data + 1] >> 2) & 0x0f) |
-                                 ((data[offset_data + 2] & 0x0f) << 4);
-        bytes[offset_byte + 2] = ((data[offset_data + 2] >> 4) & 0x03) |
-                                 ((data[offset_data + 3] & 0x3f) << 2);
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x03) | (in[1] << 2));
+        out[2] = (uint8_t) (((in[1] >> 6) & 0x0f) | (in[2] << 4));
+        out[3] = (uint8_t) (((in[2] >> 4) & 0x3f) | (in[3] << 6));
+        out[4] = (uint8_t) (in[3] >> 2);
+        in += 4;
+        out += 5;
     }
 }
 
-
-void PQCLEAN_LIGHTSABER_CLEAN_un_pack6bit(const unsigned char *bytes, uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+static void BS2POLp(poly *data, const uint8_t bytes[SABER_POLYCOMPRESSEDBYTES]) {
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
     for (j = 0; j < SABER_N / 4; j++) {
-        offset_byte = 3 * j;
-        offset_data = 4 * j;
-        data[offset_data + 0] = bytes[offset_byte + 0] & 0x3f;
-        data[offset_data + 1] = ((bytes[offset_byte + 0] >> 6) & 0x03) |
-                                ((bytes[offset_byte + 1] & 0x0f) << 2);
-        data[offset_data + 2] = ((bytes[offset_byte + 1] & 0xff) >> 4) |
-                                ((bytes[offset_byte + 2] & 0x03) << 4);
-        data[offset_data + 3] = ((bytes[offset_byte + 2] & 0xff) >> 2);
+        out[0] = in[0] | (in[1] << 8);
+        out[1] = (in[1] >> 2) | (in[2] << 6);
+        out[2] = (in[2] >> 4) | (in[3] << 4);
+        out[3] = (in[3] >> 6) | (in[4] << 2);
+        in += 5;
+        out += 4;
     }
 }
 
-
-static void POLVECp2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 10) / 8;
-        for (j = 0; j < SABER_N / 4; j++) {
-            offset_byte = offset_byte1 + 5 * j;
-            offset_data = 4 * j;
-            bytes[offset_byte + 0] = (data[i][offset_data + 0] & (0xff));
-            bytes[offset_byte + 1] = ((data[i][offset_data + 0] >> 8) & 0x03) |
-                                     ((data[i][offset_data + 1] & 0x3f) << 2);
-            bytes[offset_byte + 2] = ((data[i][offset_data + 1] >> 6) & 0x0f) |
-                                     ((data[i][offset_data + 2] & 0x0f) << 4);
-            bytes[offset_byte + 3] = ((data[i][offset_data + 2] >> 4) & 0x3f) |
-                                     ((data[i][offset_data + 3] & 0x03) << 6);
-            bytes[offset_byte + 4] = ((data[i][offset_data + 3] >> 2) & 0xff);
-        }
+void PQCLEAN_LIGHTSABER_CLEAN_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLq2BS(bytes + i * SABER_POLYBYTES, &data[i]);
     }
 }
 
-static void BS2POLVECp(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 10) / 8;
-        for (j = 0; j < SABER_N / 4; j++) {
-            offset_byte = offset_byte1 + 5 * j;
-            offset_data = 4 * j;
-            data[i][offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                       ((bytes[offset_byte + 1] & 0x03) << 8);
-            data[i][offset_data + 1] = ((bytes[offset_byte + 1] >> 2) & (0x3f)) |
-                                       ((bytes[offset_byte + 2] & 0x0f) << 6);
-            data[i][offset_data + 2] = ((bytes[offset_byte + 2] >> 4) & (0x0f)) |
-                                       ((bytes[offset_byte + 3] & 0x3f) << 4);
-            data[i][offset_data + 3] = ((bytes[offset_byte + 3] >> 6) & (0x03)) |
-                                       ((bytes[offset_byte + 4] & 0xff) << 2);
-        }
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLq(&data[i], bytes + i * SABER_POLYBYTES);
     }
 }
 
-
-
-static void POLVECq2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 13) / 8;
-        for (j = 0; j < SABER_N / 8; j++) {
-            offset_byte = offset_byte1 + 13 * j;
-            offset_data = 8 * j;
-            bytes[offset_byte + 0] = (data[i][offset_data + 0] & (0xff));
-            bytes[offset_byte + 1] = ((data[i][offset_data + 0] >> 8) & 0x1f) |
-                                     ((data[i][offset_data + 1] & 0x07) << 5);
-            bytes[offset_byte + 2] = ((data[i][offset_data + 1] >> 3) & 0xff);
-            bytes[offset_byte + 3] = ((data[i][offset_data + 1] >> 11) & 0x03) |
-                                     ((data[i][offset_data + 2] & 0x3f) << 2);
-            bytes[offset_byte + 4] = ((data[i][offset_data + 2] >> 6) & 0x7f) |
-                                     ((data[i][offset_data + 3] & 0x01) << 7);
-            bytes[offset_byte + 5] = ((data[i][offset_data + 3] >> 1) & 0xff);
-            bytes[offset_byte + 6] = ((data[i][offset_data + 3] >> 9) & 0x0f) |
-                                     ((data[i][offset_data + 4] & 0x0f) << 4);
-            bytes[offset_byte + 7] = ((data[i][offset_data + 4] >> 4) & 0xff);
-            bytes[offset_byte + 8] = ((data[i][offset_data + 4] >> 12) & 0x01) |
-                                     ((data[i][offset_data + 5] & 0x7f) << 1);
-            bytes[offset_byte + 9] = ((data[i][offset_data + 5] >> 7) & 0x3f) |
-                                     ((data[i][offset_data + 6] & 0x03) << 6);
-            bytes[offset_byte + 10] = ((data[i][offset_data + 6] >> 2) & 0xff);
-            bytes[offset_byte + 11] = ((data[i][offset_data + 6] >> 10) & 0x07) |
-                                      ((data[i][offset_data + 7] & 0x1f) << 3);
-            bytes[offset_byte + 12] = ((data[i][offset_data + 7] >> 5) & 0xff);
-        }
+void PQCLEAN_LIGHTSABER_CLEAN_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLp2BS(bytes + i * SABER_POLYCOMPRESSEDBYTES, &data[i]);
     }
 }
 
-static void BS2POLVECq(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 13) / 8;
-        for (j = 0; j < SABER_N / 8; j++) {
-            offset_byte = offset_byte1 + 13 * j;
-            offset_data = 8 * j;
-            data[i][offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                       ((bytes[offset_byte + 1] & 0x1f) << 8);
-            data[i][offset_data + 1] = (bytes[offset_byte + 1] >> 5 & (0x07)) |
-                                       ((bytes[offset_byte + 2] & 0xff) << 3) |
-                                       ((bytes[offset_byte + 3] & 0x03) << 11);
-            data[i][offset_data + 2] = (bytes[offset_byte + 3] >> 2 & (0x3f)) |
-                                       ((bytes[offset_byte + 4] & 0x7f) << 6);
-            data[i][offset_data + 3] = (bytes[offset_byte + 4] >> 7 & (0x01)) |
-                                       ((bytes[offset_byte + 5] & 0xff) << 1) |
-                                       ((bytes[offset_byte + 6] & 0x0f) << 9);
-            data[i][offset_data + 4] = (bytes[offset_byte + 6] >> 4 & (0x0f)) |
-                                       ((bytes[offset_byte + 7] & 0xff) << 4) |
-                                       ((bytes[offset_byte + 8] & 0x01) << 12);
-            data[i][offset_data + 5] = (bytes[offset_byte + 8] >> 1 & (0x7f)) |
-                                       ((bytes[offset_byte + 9] & 0x3f) << 7);
-            data[i][offset_data + 6] = (bytes[offset_byte + 9] >> 6 & (0x03)) |
-                                       ((bytes[offset_byte + 10] & 0xff) << 2) |
-                                       ((bytes[offset_byte + 11] & 0x07) << 10);
-            data[i][offset_data + 7] = (bytes[offset_byte + 11] >> 3 & (0x1f)) |
-                                       ((bytes[offset_byte + 12] & 0xff) << 5);
-        }
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLp(&data[i], bytes + i * SABER_POLYCOMPRESSEDBYTES);
     }
 }
 
-//only BS2POLq no BS2POLp
-void PQCLEAN_LIGHTSABER_CLEAN_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
-    for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 13 * j;
-        offset_data = 8 * j;
-        data[offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                ((bytes[offset_byte + 1] & 0x1f) << 8);
-        data[offset_data + 1] = (bytes[offset_byte + 1] >> 5 & (0x07)) |
-                                ((bytes[offset_byte + 2] & 0xff) << 3) |
-                                ((bytes[offset_byte + 3] & 0x03) << 11);
-        data[offset_data + 2] = (bytes[offset_byte + 3] >> 2 & (0x3f)) |
-                                ((bytes[offset_byte + 4] & 0x7f) << 6);
-        data[offset_data + 3] = (bytes[offset_byte + 4] >> 7 & (0x01)) |
-                                ((bytes[offset_byte + 5] & 0xff) << 1) |
-                                ((bytes[offset_byte + 6] & 0x0f) << 9);
-        data[offset_data + 4] = (bytes[offset_byte + 6] >> 4 & (0x0f)) |
-                                ((bytes[offset_byte + 7] & 0xff) << 4) |
-                                ((bytes[offset_byte + 8] & 0x01) << 12);
-        data[offset_data + 5] = (bytes[offset_byte + 8] >> 1 & (0x7f)) |
-                                ((bytes[offset_byte + 9] & 0x3f) << 7);
-        data[offset_data + 6] = (bytes[offset_byte + 9] >> 6 & (0x03)) |
-                                ((bytes[offset_byte + 10] & 0xff) << 2) |
-                                ((bytes[offset_byte + 11] & 0x07) << 10);
-        data[offset_data + 7] = (bytes[offset_byte + 11] >> 3 & (0x1f)) |
-                                ((bytes[offset_byte + 12] & 0xff) << 5);
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]) {
+    size_t i, j;
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            data->coeffs[j * 8 + i] = ((bytes[j] >> i) & 0x01);
+        }
     }
 }
 
-void PQCLEAN_LIGHTSABER_CLEAN_POLVEC2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus) {
-    if (modulus == 1024) {
-        POLVECp2BS(bytes, data);
-    } else if (modulus == 8192) {
-        POLVECq2BS(bytes, data);
-    }
-}
+void PQCLEAN_LIGHTSABER_CLEAN_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data) {
+    size_t i, j;
+    memset(bytes, 0, SABER_KEYBYTES);
 
-void PQCLEAN_LIGHTSABER_CLEAN_BS2POLVEC(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus) {
-    if (modulus == 1024) {
-        BS2POLVECp(bytes, data);
-    } else if (modulus == 8192) {
-        BS2POLVECq(bytes, data);
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            bytes[j] = bytes[j] | ((data->coeffs[j * 8 + i] & 0x01) << i);
+        }
     }
 }

crypto_kem/lightsaber/clean/pack_unpack.h

@@ -1,28 +1,28 @@
 #ifndef PACK_UNPACK_H
 #define PACK_UNPACK_H
-
 #include "SABER_params.h"
+#include "poly.h"
 #include <stdint.h>
 #include <stdio.h>
 
+void PQCLEAN_LIGHTSABER_CLEAN_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data);
+
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]);
 
-void PQCLEAN_LIGHTSABER_CLEAN_pack_3bit(uint8_t *bytes, const uint16_t *data);
 
-void PQCLEAN_LIGHTSABER_CLEAN_un_pack3bit(const uint8_t *bytes, uint16_t *data);
+void PQCLEAN_LIGHTSABER_CLEAN_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]);
 
-void PQCLEAN_LIGHTSABER_CLEAN_pack_4bit(uint8_t *bytes, const uint16_t *data);
+void PQCLEAN_LIGHTSABER_CLEAN_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]);
 
-void PQCLEAN_LIGHTSABER_CLEAN_un_pack4bit(const unsigned char *bytes, uint16_t *ar);
 
-void PQCLEAN_LIGHTSABER_CLEAN_pack_6bit(uint8_t *bytes, const uint16_t *data);
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]);
 
-void PQCLEAN_LIGHTSABER_CLEAN_un_pack6bit(const unsigned char *bytes, uint16_t *data);
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]);
 
 
-void PQCLEAN_LIGHTSABER_CLEAN_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]);
+void PQCLEAN_LIGHTSABER_CLEAN_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]);
 
-void PQCLEAN_LIGHTSABER_CLEAN_POLVEC2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus);
+void PQCLEAN_LIGHTSABER_CLEAN_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data);
 
-void PQCLEAN_LIGHTSABER_CLEAN_BS2POLVEC(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus);
 
 #endif

crypto_kem/lightsaber/clean/poly.c

@@ -1,21 +1,57 @@
-/*---------------------------------------------------------------------
-This file has been adapted from the implementation
-(available at, Public Domain https://github.com/pq-crystals/kyber)
-of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
-by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
-Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
-#include "SABER_params.h"
+#include "api.h"
 #include "cbd.h"
 #include "fips202.h"
+#include "pack_unpack.h"
 #include "poly.h"
+#include <stddef.h>
 
-void PQCLEAN_LIGHTSABER_CLEAN_GenSecret(uint16_t r[SABER_K][SABER_N], const unsigned char *seed) {
-    uint8_t buf[SABER_MU * SABER_N * SABER_K / 8];
+void PQCLEAN_LIGHTSABER_CLEAN_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const poly s[SABER_L], int16_t transpose) {
+    size_t i, j;
+
+    if (transpose) {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_LIGHTSABER_CLEAN_poly_mul(&c[i], &A[0][i], &s[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_LIGHTSABER_CLEAN_poly_mul(&c[i], &A[j][i], &s[j], 1);
+            }
+        }
+    } else {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_LIGHTSABER_CLEAN_poly_mul(&c[i], &A[i][0], &s[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_LIGHTSABER_CLEAN_poly_mul(&c[i], &A[i][j], &s[j], 1);
+            }
+        }
+    }
+}
+
+void PQCLEAN_LIGHTSABER_CLEAN_InnerProd(poly *c, const poly b[SABER_L], const poly s[SABER_L]) {
+    size_t i;
+
+    PQCLEAN_LIGHTSABER_CLEAN_poly_mul(c, &b[0], &s[0], 0);
+    for (i = 1; i < SABER_L; i++) {
+        PQCLEAN_LIGHTSABER_CLEAN_poly_mul(c, &b[i], &s[i], 1);
+    }
+}
+
+void PQCLEAN_LIGHTSABER_CLEAN_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYVECBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_LIGHTSABER_CLEAN_BS2POLVECq(A[i], buf + i * SABER_POLYVECBYTES);
+    }
+}
+
+void PQCLEAN_LIGHTSABER_CLEAN_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYCOINBYTES];
 
     shake128(buf, sizeof(buf), seed, SABER_NOISESEEDBYTES);
 
-    for (size_t i = 0; i < SABER_K; i++) {
-        PQCLEAN_LIGHTSABER_CLEAN_cbd(r[i], buf + i * SABER_MU * SABER_N / 8);
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_LIGHTSABER_CLEAN_cbd(s[i].coeffs, buf + i * SABER_POLYCOINBYTES);
     }
 }

crypto_kem/lightsaber/clean/poly.h

@@ -1,26 +1,23 @@
 #ifndef POLY_H
 #define POLY_H
-
-/*---------------------------------------------------------------------
-This file has been adapted from the implementation
-(available at, Public Domain https://github.com/pq-crystals/kyber)
-of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
-by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
-Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
-
-
 #include "SABER_params.h"
 #include <stdint.h>
 
-typedef struct {
+typedef union {
     uint16_t coeffs[SABER_N];
 } poly;
 
-typedef struct {
-    poly vec[SABER_K];
-} polyvec;
 
-void PQCLEAN_LIGHTSABER_CLEAN_GenSecret(uint16_t r[SABER_K][SABER_N], const unsigned char *seed);
+void PQCLEAN_LIGHTSABER_CLEAN_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const poly s[SABER_L], int16_t transpose);
+
+void PQCLEAN_LIGHTSABER_CLEAN_InnerProd(poly *c, const poly b[SABER_L], const poly s[SABER_L]);
+
+void PQCLEAN_LIGHTSABER_CLEAN_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]);
+
+void PQCLEAN_LIGHTSABER_CLEAN_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]);
+
+
+void PQCLEAN_LIGHTSABER_CLEAN_poly_mul(poly *c, const poly *a, const poly *b, int accumulate);
+
 
 #endif

crypto_kem/lightsaber/clean/poly_mul.c

@@ -1,4 +1,4 @@
-#include "poly_mul.h"
+#include "poly.h"
 #include <stdint.h>
 #include <string.h>
 
@@ -11,13 +11,13 @@
 #define OVERFLOWING_MUL(X, Y) ((uint16_t)((uint32_t)(X) * (uint32_t)(Y)))
 
 #define KARATSUBA_N 64
-static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t *result_final) {
+static void karatsuba_simple(uint16_t *result_final, const uint16_t *a_1, const uint16_t *b_1) {
     uint16_t d01[KARATSUBA_N / 2 - 1];
     uint16_t d0123[KARATSUBA_N / 2 - 1];
     uint16_t d23[KARATSUBA_N / 2 - 1];
     uint16_t result_d01[KARATSUBA_N - 1];
 
-    int32_t i, j;
+    size_t i, j;
 
     memset(result_d01, 0, (KARATSUBA_N - 1)*sizeof(uint16_t));
     memset(d01, 0, (KARATSUBA_N / 2 - 1)*sizeof(uint16_t));
@@ -110,7 +110,7 @@ static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t
 
 
 
-static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *result) {
+static void toom_cook_4way (uint16_t *result, const uint16_t *a1, const uint16_t *b1) {
     uint16_t inv3 = 43691, inv9 = 36409, inv15 = 61167;
 
     uint16_t aw1[N_SB], aw2[N_SB], aw3[N_SB], aw4[N_SB], aw5[N_SB], aw6[N_SB], aw7[N_SB];
@@ -181,13 +181,13 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 
     // MULTIPLICATION
 
-    karatsuba_simple(aw1, bw1, w1);
-    karatsuba_simple(aw2, bw2, w2);
-    karatsuba_simple(aw3, bw3, w3);
-    karatsuba_simple(aw4, bw4, w4);
-    karatsuba_simple(aw5, bw5, w5);
-    karatsuba_simple(aw6, bw6, w6);
-    karatsuba_simple(aw7, bw7, w7);
+    karatsuba_simple(w1, aw1, bw1);
+    karatsuba_simple(w2, aw2, bw2);
+    karatsuba_simple(w3, aw3, bw3);
+    karatsuba_simple(w4, aw4, bw4);
+    karatsuba_simple(w5, aw5, bw5);
+    karatsuba_simple(w6, aw6, bw6);
+    karatsuba_simple(w7, aw7, bw7);
 
     // INTERPOLATION
     for (i = 0; i < N_SB_RES; ++i) {
@@ -228,19 +228,21 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
     }
 }
 
-void PQCLEAN_LIGHTSABER_CLEAN_pol_mul(uint16_t *a, uint16_t *b, uint16_t *res, uint16_t p, uint32_t n) {
-    uint32_t i;
-    // normal multiplication
-    uint16_t c[512];
+/* res += a*b */
+void PQCLEAN_LIGHTSABER_CLEAN_poly_mul(poly *c, const poly *a, const poly *b, const int accumulate) {
+    uint16_t C[2 * SABER_N] = {0};
+    size_t i;
 
-    for (i = 0; i < 512; i++) {
-        c[i] = 0;
-    }
-
-    toom_cook_4way(a, b, c);
+    toom_cook_4way(C, a->coeffs, b->coeffs);
 
-    // reduction
-    for (i = n; i < 2 * n; i++) {
-        res[i - n] = (c[i - n] - c[i]) & (p - 1);
+    /* reduction */
+    if (accumulate == 0) {
+        for (i = SABER_N; i < 2 * SABER_N; i++) {
+            c->coeffs[i - SABER_N] = (C[i - SABER_N] - C[i]);
+        }
+    } else {
+        for (i = SABER_N; i < 2 * SABER_N; i++) {
+            c->coeffs[i - SABER_N] += (C[i - SABER_N] - C[i]);
+        }
     }
 }

crypto_kem/lightsaber/clean/poly_mul.h

@@ -1,9 +1,3 @@
-#ifndef POLYMUL_H
-#define POLYMUL_H
 
-#include "SABER_params.h"
-#include <stdint.h>
 
-void PQCLEAN_LIGHTSABER_CLEAN_pol_mul(uint16_t *a, uint16_t *b, uint16_t *res, uint16_t p, uint32_t n);
 
-#endif

crypto_kem/lightsaber/clean/verify.c

@@ -1,3 +1,5 @@
+#include "verify.h"
+
 /*-------------------------------------------------
 This file has been adapted from the implementation
 (available at https://github.com/pq-crystals/kyber) of
@@ -5,26 +7,25 @@ This file has been adapted from the implementation
  by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------*/
-#include "verify.h"
-#include <stdint.h>
+
 
 /* returns 0 for equal strings, 1 for non-equal strings */
-unsigned char PQCLEAN_LIGHTSABER_CLEAN_verify(const unsigned char *a, const unsigned char *b, size_t len) {
+uint8_t PQCLEAN_LIGHTSABER_CLEAN_verify(const uint8_t *a, const uint8_t *b, size_t len) {
     uint64_t r;
     size_t i;
-
     r = 0;
+
     for (i = 0; i < len; i++) {
         r |= a[i] ^ b[i];
     }
 
     r = (~r + 1); // Two's complement
     r >>= 63;
-    return (unsigned char)r;
+    return (uint8_t) r;
 }
 
 /* b = 1 means mov, b = 0 means don't mov*/
-void PQCLEAN_LIGHTSABER_CLEAN_cmov(unsigned char *r, const unsigned char *x, size_t len, unsigned char b) {
+void PQCLEAN_LIGHTSABER_CLEAN_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b) {
     size_t i;
 
     b = -b;

crypto_kem/lightsaber/clean/verify.h

@@ -1,6 +1,5 @@
 #ifndef VERIFY_H
 #define VERIFY_H
-
 /*-------------------------------------------------
 This file has been adapted from the implementation
 (available at https://github.com/pq-crystals/kyber) of
@@ -13,9 +12,11 @@ Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 #include <stdint.h>
 
 /* returns 0 for equal strings, 1 for non-equal strings */
-unsigned char PQCLEAN_LIGHTSABER_CLEAN_verify(const unsigned char *a, const unsigned char *b, size_t len);
+uint8_t PQCLEAN_LIGHTSABER_CLEAN_verify(const uint8_t *a, const uint8_t *b, size_t len);
+
 
 /* b = 1 means mov, b = 0 means don't mov*/
-void PQCLEAN_LIGHTSABER_CLEAN_cmov(unsigned char *r, const unsigned char *x, size_t len, unsigned char b);
+void PQCLEAN_LIGHTSABER_CLEAN_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b);
+
 
 #endif

crypto_kem/saber/META.yml

@@ -14,4 +14,13 @@ principal-submitters:
   - Frederik Vercauteren
 implementations:
     - name: clean
-      version: https://github.com/KULeuven-COSIC/SABER/commit/14ede83f1ff3bcc41f0464543542366c68b55871
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/1ae84c3c/saber
+    - name: avx2
+      version: https://github.com/KULeuven-COSIC/SABER/tree/509cc5ec3a7e12a751ccdd2ef5bd6e54e00bd350 via https://github.com/jschanck/package-pqclean/tree/1ae84c3c/saber
+      supported_platforms:
+          - architecture: x86_64
+            operating_systems:
+                - Linux
+                - Darwin
+            required_flags:
+                - avx2

crypto_kem/saber/avx2/LICENSE

@@ -0,0 +1 @@
+Public Domain

crypto_kem/saber/avx2/Makefile

@@ -0,0 +1,22 @@
+# This Makefile can be used with GNU Make or BSD Make
+
+LIB=libsaber_avx2.a
+HEADERS=api.h cbd.h pack_unpack.h poly.h SABER_indcpa.h SABER_params.h verify.h 
+OBJECTS=cbd.o kem.o pack_unpack.o poly.o poly_mul.o SABER_indcpa.o verify.o 
+
+CFLAGS=-O3 -mavx2 -Wall -Wextra -Wpedantic -Wvla -Werror -Wredundant-decls -Wmissing-prototypes -std=c99 -I../../../common $(EXTRAFLAGS)
+
+all: $(LIB)
+
+%.o: %.s $(HEADERS)
+	$(AS) -o $@ $<
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_kem/saber/avx2/SABER_indcpa.c

@@ -0,0 +1,125 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "fips202.h"
+#include "pack_unpack.h"
+#include "poly.h"
+#include "randombytes.h"
+#include <stdint.h>
+#include <string.h>
+
+#define h1 (1 << (SABER_EQ - SABER_EP - 1))
+#define h2 ((1 << (SABER_EP - 2)) - (1 << (SABER_EP - SABER_ET - 1)) + (1 << (SABER_EQ - SABER_EP - 1)))
+
+void PQCLEAN_SABER_AVX2_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]) {
+    size_t i, j;
+
+    poly A[SABER_L][SABER_L];
+    poly *skpv1 = A[0]; // use first row of A to hold sk temporarily
+    toom4_points skpv1_eval[SABER_L];
+    poly res[SABER_L];
+
+    uint8_t rand[SABER_NOISESEEDBYTES];
+    uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+
+    randombytes(seed_A, SABER_SEEDBYTES);
+    shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
+
+    randombytes(rand, SABER_NOISESEEDBYTES);
+    PQCLEAN_SABER_AVX2_GenSecret(skpv1, rand);
+    PQCLEAN_SABER_AVX2_POLVECq2BS(sk, skpv1); // pack secret key
+
+    for (j = 0; j < SABER_L; j++) {
+        PQCLEAN_SABER_AVX2_toom4_eval(&skpv1_eval[j], &skpv1[j]);
+    }
+
+    PQCLEAN_SABER_AVX2_GenMatrix(A, seed_A); // sample matrix A
+    PQCLEAN_SABER_AVX2_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const toom4_points *)skpv1_eval, 1); // Matrix in transposed order
+
+    // rounding
+    for (i = 0; i < SABER_L; i++) {
+        for (j = 0; j < SABER_N; j++) {
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
+        }
+    }
+
+    PQCLEAN_SABER_AVX2_POLVECp2BS(pk, res); // pack public key
+}
+
+
+void PQCLEAN_SABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]) {
+    size_t i, j;
+
+    poly A[SABER_L][SABER_L];
+    poly res[SABER_L];
+    toom4_points skpv1_eval[SABER_L];
+
+    poly *temp = A[0]; // re-use stack space
+    poly *vprime = &A[0][0];
+    poly *message = &A[0][1];
+
+    const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+    uint8_t *msk_c = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+
+    PQCLEAN_SABER_AVX2_GenSecret(temp, noiseseed);
+    for (j = 0; j < SABER_L; j++) {
+        PQCLEAN_SABER_AVX2_toom4_eval(&skpv1_eval[j], &temp[j]);
+    }
+
+    PQCLEAN_SABER_AVX2_GenMatrix(A, seed_A);
+    PQCLEAN_SABER_AVX2_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const toom4_points *)skpv1_eval, 0); // 0 => not transposed
+
+    // rounding
+    for (i = 0; i < SABER_L; i++) { //shift right EQ-EP bits
+        for (j = 0; j < SABER_N; j++) {
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
+        }
+    }
+    PQCLEAN_SABER_AVX2_POLVECp2BS(ciphertext, res);
+
+    // vector-vector scalar multiplication with mod p
+    PQCLEAN_SABER_AVX2_BS2POLVECp(temp, pk);
+    PQCLEAN_SABER_AVX2_InnerProd(vprime, temp, skpv1_eval);
+    PQCLEAN_SABER_AVX2_BS2POLmsg(message, m);
+
+    for (i = 0; i < SABER_N; i++) {
+        vprime->coeffs[i] += h1 - (message->coeffs[i] << (SABER_EP - 1));
+        vprime->coeffs[i] &= SABER_P - 1;
+        vprime->coeffs[i] >>= SABER_EP - SABER_ET;
+    }
+
+    PQCLEAN_SABER_AVX2_POLT2BS(msk_c, vprime);
+}
+
+
+void PQCLEAN_SABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]) {
+    size_t i;
+
+    poly temp[SABER_L];
+    toom4_points sksv_eval[SABER_L];
+
+    const uint8_t *packed_cm = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+    poly *v = &temp[0];
+    poly *cm = &temp[1];
+
+    PQCLEAN_SABER_AVX2_BS2POLVECq(temp, sk);
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_SABER_AVX2_toom4_eval(&sksv_eval[i], &temp[i]);
+    }
+
+    PQCLEAN_SABER_AVX2_BS2POLVECp(temp, ciphertext);
+    PQCLEAN_SABER_AVX2_InnerProd(v, temp, sksv_eval);
+
+    PQCLEAN_SABER_AVX2_BS2POLT(cm, packed_cm);
+
+    for (i = 0; i < SABER_N; i++) {
+        v->coeffs[i] += h2 - (cm->coeffs[i] << (SABER_EP - SABER_ET));
+        v->coeffs[i] &= SABER_P - 1;
+        v->coeffs[i] >>= SABER_EP - 1;
+    }
+
+    PQCLEAN_SABER_AVX2_POLmsg2BS(m, v);
+}

crypto_kem/saber/avx2/SABER_indcpa.h

@@ -0,0 +1,13 @@
+#ifndef INDCPA_H
+#define INDCPA_H
+#include "SABER_params.h"
+#include <stdint.h>
+
+void PQCLEAN_SABER_AVX2_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]);
+
+void PQCLEAN_SABER_AVX2_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]);
+
+void PQCLEAN_SABER_AVX2_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]);
+
+
+#endif

crypto_kem/saber/avx2/SABER_params.h

@@ -0,0 +1,41 @@
+#ifndef PARAMS_H
+#define PARAMS_H
+
+
+/* Don't change anything below this line */
+#define SABER_L 3
+#define SABER_MU 8
+#define SABER_ET 4
+
+#define SABER_N 256
+
+#define SABER_EP 10
+#define SABER_P (1 << SABER_EP)
+
+#define SABER_EQ 13
+#define SABER_Q (1 << SABER_EQ)
+
+#define SABER_SEEDBYTES 32
+#define SABER_NOISESEEDBYTES 32
+#define SABER_KEYBYTES 32
+#define SABER_HASHBYTES 32
+
+#define SABER_POLYCOINBYTES (SABER_MU * SABER_N / 8)
+
+#define SABER_POLYBYTES (SABER_EQ * SABER_N / 8)
+#define SABER_POLYVECBYTES (SABER_L * SABER_POLYBYTES)
+
+#define SABER_POLYCOMPRESSEDBYTES (SABER_EP * SABER_N / 8)
+#define SABER_POLYVECCOMPRESSEDBYTES (SABER_L * SABER_POLYCOMPRESSEDBYTES)
+
+#define SABER_SCALEBYTES_KEM (SABER_ET * SABER_N / 8)
+
+#define SABER_INDCPA_PUBLICKEYBYTES (SABER_POLYVECCOMPRESSEDBYTES + SABER_SEEDBYTES)
+#define SABER_INDCPA_SECRETKEYBYTES (SABER_POLYVECBYTES)
+
+#define SABER_PUBLICKEYBYTES (SABER_INDCPA_PUBLICKEYBYTES)
+#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES + SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
+
+#define SABER_BYTES_CCA_DEC (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM)
+
+#endif

crypto_kem/saber/avx2/api.h

@@ -0,0 +1,18 @@
+#ifndef PQCLEAN_SABER_AVX2_API_H
+#define PQCLEAN_SABER_AVX2_API_H
+
+
+#define PQCLEAN_SABER_AVX2_CRYPTO_ALGNAME "Saber"
+#define PQCLEAN_SABER_AVX2_CRYPTO_BYTES 32
+#define PQCLEAN_SABER_AVX2_CRYPTO_CIPHERTEXTBYTES 1088
+#define PQCLEAN_SABER_AVX2_CRYPTO_PUBLICKEYBYTES 992
+#define PQCLEAN_SABER_AVX2_CRYPTO_SECRETKEYBYTES 2304
+
+int PQCLEAN_SABER_AVX2_crypto_kem_keypair(unsigned char *pk, unsigned char *sk);
+
+int PQCLEAN_SABER_AVX2_crypto_kem_enc(unsigned char *ct, unsigned char *k, const unsigned char *pk);
+
+int PQCLEAN_SABER_AVX2_crypto_kem_dec(unsigned char *k, const unsigned char *ct, const unsigned char *sk);
+
+
+#endif /* PQCLEAN_SABER_AVX2_API_H */

crypto_kem/saber/avx2/cbd.c

@@ -0,0 +1,48 @@
+#include "SABER_params.h"
+#include "api.h"
+#include "cbd.h"
+#include <stdint.h>
+/*---------------------------------------------------------------------
+This file has been adapted from the implementation
+(available at, Public Domain https://github.com/pq-crystals/kyber)
+of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------------------------*/
+
+
+static uint64_t load_littleendian(const uint8_t *x, int bytes) {
+    int i;
+    uint64_t r = x[0];
+    for (i = 1; i < bytes; i++) {
+        r |= (uint64_t)x[i] << (8 * i);
+    }
+    return r;
+}
+
+void PQCLEAN_SABER_AVX2_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]) {
+    uint32_t t, d, a[4], b[4];
+    int i, j;
+
+    for (i = 0; i < SABER_N / 4; i++) {
+        t = (uint32_t) load_littleendian(buf + 4 * i, 4);
+        d = 0;
+        for (j = 0; j < 4; j++) {
+            d += (t >> j) & 0x11111111;
+        }
+
+        a[0] = d & 0xf;
+        b[0] = (d >> 4) & 0xf;
+        a[1] = (d >> 8) & 0xf;
+        b[1] = (d >> 12) & 0xf;
+        a[2] = (d >> 16) & 0xf;
+        b[2] = (d >> 20) & 0xf;
+        a[3] = (d >> 24) & 0xf;
+        b[3] = (d >> 28);
+
+        s[4 * i + 0] = (uint16_t)(a[0] - b[0]);
+        s[4 * i + 1] = (uint16_t)(a[1] - b[1]);
+        s[4 * i + 2] = (uint16_t)(a[2] - b[2]);
+        s[4 * i + 3] = (uint16_t)(a[3] - b[3]);
+    }
+}

crypto_kem/saber/avx2/cbd.h

@@ -0,0 +1,16 @@
+#ifndef CBD_H
+#define CBD_H
+/*---------------------------------------------------------------------
+This file has been adapted from the implementation
+(available at, Public Domain https://github.com/pq-crystals/kyber)
+of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------------------------*/
+#include "SABER_params.h"
+#include <stdint.h>
+
+void PQCLEAN_SABER_AVX2_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]);
+
+
+#endif

crypto_kem/saber/avx2/kem.c

@@ -0,0 +1,77 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "api.h"
+#include "fips202.h"
+#include "randombytes.h"
+#include "verify.h"
+#include <stddef.h>
+#include <stdint.h>
+
+
+int PQCLEAN_SABER_AVX2_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
+    size_t i;
+
+    PQCLEAN_SABER_AVX2_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
+    for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
+        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
+    }
+
+    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES); // Then hash(pk) is appended.
+
+    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES); // Remaining part of sk contains a pseudo-random number.
+    // This is output when check in PQCLEAN_SABER_AVX2_crypto_kem_dec() fails.
+    return (0);
+}
+
+int PQCLEAN_SABER_AVX2_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t *pk) {
+
+    uint8_t kr[64]; // Will contain key, coins
+    uint8_t buf[64];
+
+    randombytes(buf, 32);
+
+    sha3_256(buf, buf, 32); // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
+
+    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES); // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
+
+    sha3_512(kr, buf, 64);               // kr[0:63] <-- Hash(buf[0:63]);
+    // K^ <-- kr[0:31]
+    // noiseseed (r) <-- kr[32:63];
+    PQCLEAN_SABER_AVX2_indcpa_kem_enc(c, buf, kr + 32, pk); // buf[0:31] contains message; kr[32:63] contains randomness r;
+
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC);
+
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}
+
+int PQCLEAN_SABER_AVX2_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
+    size_t i;
+    uint8_t fail;
+    uint8_t cmp[SABER_BYTES_CCA_DEC];
+    uint8_t buf[64];
+    uint8_t kr[64]; // Will contain key, coins
+    const uint8_t *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
+
+    PQCLEAN_SABER_AVX2_indcpa_kem_dec(buf, sk, c); // buf[0:31] <-- message
+
+    // Multitarget countermeasure for coins + contributory KEM
+    for (i = 0; i < 32; i++) { // Save hash by storing h(pk) in sk
+        buf[32 + i] = sk[SABER_SECRETKEYBYTES - 64 + i];
+    }
+
+    sha3_512(kr, buf, 64);
+
+    PQCLEAN_SABER_AVX2_indcpa_kem_enc(cmp, buf, kr + 32, pk);
+
+    fail = PQCLEAN_SABER_AVX2_verify(c, cmp, SABER_BYTES_CCA_DEC);
+
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC); // overwrite coins in kr with h(c)
+
+    PQCLEAN_SABER_AVX2_cmov(kr, sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES, fail);
+
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}

crypto_kem/saber/avx2/pack_unpack.c

@@ -0,0 +1,147 @@
+#include "SABER_params.h"
+#include "pack_unpack.h"
+#include "poly.h"
+#include <string.h>
+
+void PQCLEAN_SABER_AVX2_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 2; j++) {
+        out[0] = (uint8_t) ((in[0] & 0x0f) | (in[1] << 4));
+        in += 2;
+        out += 1;
+    }
+}
+
+void PQCLEAN_SABER_AVX2_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]) {
+    /* This function does not reduce its output mod T */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 2; j++) {
+        out[0] = in[0];
+        out[1] = in[0] >> 4;
+        in += 1;
+        out += 2;
+    }
+}
+
+static void POLq2BS(uint8_t bytes[SABER_POLYBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x1f) | (in[1] << 5));
+        out[2] = (uint8_t) (in[1] >> 3);
+        out[3] = (uint8_t) (((in[1] >> 11) & 0x03) | (in[2] << 2));
+        out[4] = (uint8_t) (((in[2] >> 6) & 0x7f) | (in[3] << 7));
+        out[5] = (uint8_t) (in[3] >> 1);
+        out[6] = (uint8_t) (((in[3] >> 9) & 0x0f) | (in[4] << 4));
+        out[7] = (uint8_t) (in[4] >> 4);
+        out[8] = (uint8_t) (((in[4] >> 12) & 0x01) | (in[5] << 1));
+        out[9] = (uint8_t) (((in[5] >> 7) & 0x3f) | (in[6] << 6));
+        out[10] = (uint8_t) (in[6] >> 2);
+        out[11] = (uint8_t) (((in[6] >> 10) & 0x07) | (in[7] << 3));
+        out[12] = (uint8_t) (in[7] >> 5);
+        in += 8;
+        out += 13;
+    }
+}
+
+static void BS2POLq(poly *data, const uint8_t bytes[SABER_POLYBYTES]) {
+    /* This function does not reduce its output mod Q */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (in[0]) | (in[1] << 8);
+        out[1] = (in[1] >> 5) | (in[2] << 3) | (in[3] << 11);
+        out[2] = (in[3] >> 2) | (in[4] << 6);
+        out[3] = (in[4] >> 7) | (in[5] << 1) | (in[6] << 9);
+        out[4] = (in[6] >> 4) | (in[7] << 4) | (in[8] << 12);
+        out[5] = (in[8] >> 1) | (in[9] << 7);
+        out[6] = (in[9] >> 6) | (in[10] << 2) | (in[11] << 10);
+        out[7] = (in[11] >> 3) | (in[12] << 5);
+        in += 13;
+        out += 8;
+    }
+}
+
+static void POLp2BS(uint8_t bytes[SABER_POLYCOMPRESSEDBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x03) | (in[1] << 2));
+        out[2] = (uint8_t) (((in[1] >> 6) & 0x0f) | (in[2] << 4));
+        out[3] = (uint8_t) (((in[2] >> 4) & 0x3f) | (in[3] << 6));
+        out[4] = (uint8_t) (in[3] >> 2);
+        in += 4;
+        out += 5;
+    }
+}
+
+static void BS2POLp(poly *data, const uint8_t bytes[SABER_POLYCOMPRESSEDBYTES]) {
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 4; j++) {
+        out[0] = in[0] | (in[1] << 8);
+        out[1] = (in[1] >> 2) | (in[2] << 6);
+        out[2] = (in[2] >> 4) | (in[3] << 4);
+        out[3] = (in[3] >> 6) | (in[4] << 2);
+        in += 5;
+        out += 4;
+    }
+}
+
+void PQCLEAN_SABER_AVX2_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLq2BS(bytes + i * SABER_POLYBYTES, &data[i]);
+    }
+}
+
+void PQCLEAN_SABER_AVX2_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLq(&data[i], bytes + i * SABER_POLYBYTES);
+    }
+}
+
+void PQCLEAN_SABER_AVX2_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLp2BS(bytes + i * SABER_POLYCOMPRESSEDBYTES, &data[i]);
+    }
+}
+
+void PQCLEAN_SABER_AVX2_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLp(&data[i], bytes + i * SABER_POLYCOMPRESSEDBYTES);
+    }
+}
+
+void PQCLEAN_SABER_AVX2_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]) {
+    size_t i, j;
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            data->coeffs[j * 8 + i] = ((bytes[j] >> i) & 0x01);
+        }
+    }
+}
+
+void PQCLEAN_SABER_AVX2_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data) {
+    size_t i, j;
+    memset(bytes, 0, SABER_KEYBYTES);
+
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            bytes[j] = bytes[j] | ((data->coeffs[j * 8 + i] & 0x01) << i);
+        }
+    }
+}

crypto_kem/saber/avx2/pack_unpack.h

@@ -0,0 +1,28 @@
+#ifndef PACK_UNPACK_H
+#define PACK_UNPACK_H
+#include "SABER_params.h"
+#include "poly.h"
+#include <stdint.h>
+#include <stdio.h>
+
+void PQCLEAN_SABER_AVX2_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data);
+
+void PQCLEAN_SABER_AVX2_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]);
+
+
+void PQCLEAN_SABER_AVX2_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]);
+
+void PQCLEAN_SABER_AVX2_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]);
+
+
+void PQCLEAN_SABER_AVX2_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]);
+
+void PQCLEAN_SABER_AVX2_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]);
+
+
+void PQCLEAN_SABER_AVX2_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]);
+
+void PQCLEAN_SABER_AVX2_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data);
+
+
+#endif

crypto_kem/saber/avx2/poly.c

@@ -0,0 +1,62 @@
+#include "cbd.h"
+#include "fips202.h"
+#include "pack_unpack.h"
+#include "poly.h"
+
+
+void PQCLEAN_SABER_AVX2_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const toom4_points s_eval[SABER_L], int transpose) {
+    size_t i, j;
+    toom4_points_product c_eval;
+
+    if (transpose) {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_SABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[0][i], &s_eval[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_SABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[j][i], &s_eval[j], 1);
+            }
+            PQCLEAN_SABER_AVX2_toom4_interp(&c[i], &c_eval);
+        }
+    } else {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_SABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[i][0], &s_eval[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_SABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &A[i][j], &s_eval[j], 1);
+            }
+            PQCLEAN_SABER_AVX2_toom4_interp(&c[i], &c_eval);
+        }
+    }
+}
+
+void PQCLEAN_SABER_AVX2_InnerProd(poly *c, const poly b[SABER_L], const toom4_points s_eval[SABER_L]) {
+    size_t i;
+    toom4_points_product c_eval; //Holds results for 9 Karatsuba at a time
+
+    PQCLEAN_SABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &b[0], &s_eval[0], 0);
+    for (i = 1; i < SABER_L; i++) {
+        PQCLEAN_SABER_AVX2_toom4_mul_A_by_B_eval(&c_eval, &b[i], &s_eval[i], 1);
+    }
+
+    PQCLEAN_SABER_AVX2_toom4_interp(c, &c_eval);
+}
+
+void PQCLEAN_SABER_AVX2_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYVECBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_SABER_AVX2_BS2POLVECq(A[i], buf + i * SABER_POLYVECBYTES);
+    }
+}
+
+void PQCLEAN_SABER_AVX2_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYCOINBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_NOISESEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_SABER_AVX2_cbd(s[i].coeffs, buf + i * SABER_POLYCOINBYTES);
+    }
+}

crypto_kem/saber/avx2/poly.h

@@ -0,0 +1,38 @@
+#ifndef POLY_H
+#define POLY_H
+#include "SABER_params.h"
+#include <immintrin.h>
+#include <stdint.h>
+
+typedef union {
+    uint16_t coeffs[SABER_N];
+    __m256i dummy;
+} poly;
+
+typedef union {
+    uint16_t coeffs[4 * SABER_N];
+    __m256i dummy;
+} toom4_points;
+
+typedef union {
+    uint16_t coeffs[8 * SABER_N];
+    __m256i dummy;
+} toom4_points_product;
+
+void PQCLEAN_SABER_AVX2_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const toom4_points s_eval[SABER_L], int transpose);
+
+void PQCLEAN_SABER_AVX2_InnerProd(poly *c, const poly b[SABER_L], const toom4_points s_eval[SABER_L]);
+
+void PQCLEAN_SABER_AVX2_GenMatrix(poly a[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]);
+
+void PQCLEAN_SABER_AVX2_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]);
+
+
+void PQCLEAN_SABER_AVX2_toom4_interp(poly *res_avx, const toom4_points_product *c_eval);
+
+void PQCLEAN_SABER_AVX2_toom4_eval(toom4_points *b_eval, const poly *b);
+
+void PQCLEAN_SABER_AVX2_toom4_mul_A_by_B_eval(toom4_points_product *c_eval, const poly *a_avx, const toom4_points *b_eval, int accumulate);
+
+
+#endif

crypto_kem/saber/avx2/verify.c

@@ -0,0 +1,35 @@
+#include "verify.h"
+
+/*-------------------------------------------------
+This file has been adapted from the implementation
+(available at https://github.com/pq-crystals/kyber) of
+"CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------*/
+
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+uint8_t PQCLEAN_SABER_AVX2_verify(const uint8_t *a, const uint8_t *b, size_t len) {
+    uint64_t r;
+    size_t i;
+    r = 0;
+
+    for (i = 0; i < len; i++) {
+        r |= a[i] ^ b[i];
+    }
+
+    r = (~r + 1); // Two's complement
+    r >>= 63;
+    return (uint8_t) r;
+}
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_SABER_AVX2_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b) {
+    size_t i;
+
+    b = -b;
+    for (i = 0; i < len; i++) {
+        r[i] ^= b & (x[i] ^ r[i]);
+    }
+}

crypto_kem/saber/avx2/verify.h

@@ -0,0 +1,22 @@
+#ifndef VERIFY_H
+#define VERIFY_H
+/*-------------------------------------------------
+This file has been adapted from the implementation
+(available at https://github.com/pq-crystals/kyber) of
+"CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
+ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
+Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
+----------------------------------------------------*/
+
+#include <stddef.h>
+#include <stdint.h>
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+uint8_t PQCLEAN_SABER_AVX2_verify(const uint8_t *a, const uint8_t *b, size_t len);
+
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_SABER_AVX2_cmov(uint8_t *r, const uint8_t *x, size_t len, uint8_t b);
+
+
+#endif

crypto_kem/saber/clean/LICENSE

@@ -1,8 +1 @@
-SABER_v1.1
-
-Public domain
-
-Authors: Jan-Pieter D'Anvers, Angshuman Karmakar, Sujoy Sinha Roy,
-Frederik Vercauteren
+Public Domain

crypto_kem/saber/clean/Makefile

@@ -1,10 +1,10 @@
 # This Makefile can be used with GNU Make or BSD Make
 
 LIB=libsaber_clean.a
-HEADERS=api.h cbd.h poly.h poly_mul.h SABER_indcpa.h SABER_params.h verify.h pack_unpack.h 
+HEADERS=api.h cbd.h pack_unpack.h poly.h poly_mul.h SABER_indcpa.h SABER_params.h verify.h 
 OBJECTS=cbd.o kem.o pack_unpack.o poly.o poly_mul.o SABER_indcpa.o verify.o 
 
-CFLAGS=-O3 -Wall -Wextra -Wpedantic -Werror -Wmissing-prototypes -Wredundant-decls -std=c99 -I../../../common $(EXTRAFLAGS)
+CFLAGS=-O3 -Wall -Wextra -Wpedantic -Wvla -Werror -Wredundant-decls -Wmissing-prototypes -std=c99 -I../../../common $(EXTRAFLAGS)
 
 all: $(LIB)
 

crypto_kem/saber/clean/SABER_indcpa.c

@@ -3,296 +3,111 @@
 #include "fips202.h"
 #include "pack_unpack.h"
 #include "poly.h"
-#include "poly_mul.h"
 #include "randombytes.h"
 #include <stdint.h>
 #include <string.h>
 
+#define h1 (1 << (SABER_EQ - SABER_EP - 1))
+#define h2 ((1 << (SABER_EP - 2)) - (1 << (SABER_EP - SABER_ET - 1)) + (1 << (SABER_EQ - SABER_EP - 1)))
 
+void PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]) {
+    size_t i, j;
 
-/*-----------------------------------------------------------------------------------
-    This routine generates a=[Matrix K x K] of 256-coefficient polynomials
+    poly A[SABER_L][SABER_L];
+    poly s[SABER_L];
+    poly res[SABER_L];
 
-#define h1 4 //2^(EQ-EP-1)
+    uint8_t rand[SABER_NOISESEEDBYTES];
+    uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
 
-#define h2 ( (1<<(SABER_EP-2)) - (1<<(SABER_EP-SABER_ET-1)) + (1<<(SABER_EQ-SABER_EP-1)) )
+    randombytes(seed_A, SABER_SEEDBYTES);
+    shake128(seed_A, SABER_SEEDBYTES, seed_A, SABER_SEEDBYTES); // for not revealing system RNG state
 
-static void InnerProd(uint16_t pkcl[SABER_K][SABER_N], uint16_t skpv[SABER_K][SABER_N], uint16_t mod, uint16_t res[SABER_N]);
-static void MatrixVectorMul(polyvec *a, uint16_t skpv[SABER_K][SABER_N], uint16_t res[SABER_K][SABER_N], uint16_t mod, int16_t transpose);
+    randombytes(rand, SABER_NOISESEEDBYTES);
+    PQCLEAN_SABER_CLEAN_GenSecret(s, rand);
+    PQCLEAN_SABER_CLEAN_POLVECq2BS(sk, s);
 
-static void POL2MSG(const uint16_t *message_dec_unpacked, unsigned char *message_dec);
+    PQCLEAN_SABER_CLEAN_GenMatrix(A, seed_A); // sample matrix A
+    PQCLEAN_SABER_CLEAN_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const poly *)s, 1); // Matrix in transposed order
 
-static void GenMatrix(polyvec *a, const unsigned char *seed) {
-    unsigned char buf[SABER_K * SABER_K * (13 * SABER_N / 8)];
 
-    uint16_t temp_ar[SABER_N];
-
-    int i, j, k;
-    uint16_t mod = (SABER_Q - 1);
-
-    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
-
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_K; j++) {
-            PQCLEAN_SABER_CLEAN_BS2POL(buf + (i * SABER_K + j) * (13 * SABER_N / 8), temp_ar);
-            for (k = 0; k < SABER_N; k++) {
-                a[i].vec[j].coeffs[k] = (temp_ar[k])& mod ;
-            }
-        }
-    }
-}
-
-
-void PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(unsigned char *pk, unsigned char *sk) {
-    polyvec a[SABER_K];
-
-    uint16_t skpv[SABER_K][SABER_N];
-
-    unsigned char seed[SABER_SEEDBYTES];
-    unsigned char noiseseed[SABER_COINBYTES];
-    int32_t i, j;
-    uint16_t mod_q = SABER_Q - 1;
-
-
-    uint16_t res[SABER_K][SABER_N];
-
-    randombytes(seed, SABER_SEEDBYTES);
-
-    // for not revealing system RNG state
-    shake128(seed, SABER_SEEDBYTES, seed, SABER_SEEDBYTES);
-    randombytes(noiseseed, SABER_COINBYTES);
-
-    GenMatrix(a, seed);   //sample matrix A
-
-    // generate secret from constant-time binomial distribution
-    PQCLEAN_SABER_CLEAN_GenSecret(skpv, noiseseed);
-
-    // do the matrix vector multiplication and rounding
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = 0;
-        }
-    }
-    MatrixVectorMul(a, skpv, res, SABER_Q - 1, 1);
-
-    // now rounding
-    for (i = 0; i < SABER_K; i++) {
+    // rounding
+    for (i = 0; i < SABER_L; i++) {
         for (j = 0; j < SABER_N; j++) {
-            // shift right 3 bits
-            res[i][j] = (res[i][j] + h1) & (mod_q);
-            res[i][j] = (res[i][j] >> (SABER_EQ - SABER_EP));
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
         }
     }
 
-    // unload and pack sk=3 x (256 coefficients of 14 bits)
-    PQCLEAN_SABER_CLEAN_POLVEC2BS(sk, skpv, SABER_Q);
-
-    // unload and pack pk=256 bits seed and 3 x (256 coefficients of 11 bits)
-    // load the public-key coefficients
-    PQCLEAN_SABER_CLEAN_POLVEC2BS(pk, res, SABER_P);
-
-
-    // now load the seedbytes in PK. Easy since seed bytes are kept in byte format.
-    for (i = 0; i < SABER_SEEDBYTES; i++) {
-        pk[SABER_POLYVECCOMPRESSEDBYTES + i] = seed[i];
-    }
-
+    PQCLEAN_SABER_CLEAN_POLVECp2BS(pk, res); // pack public key
 }
 
 
-void PQCLEAN_SABER_CLEAN_indcpa_kem_enc(const unsigned char *message_received, unsigned char *noiseseed, const unsigned char *pk, unsigned char *ciphertext) {
-    uint32_t i, j, k;
-    polyvec a[SABER_K];
-    unsigned char seed[SABER_SEEDBYTES];
-    // public key of received by the client
-    uint16_t pkcl[SABER_K][SABER_N];
-    uint16_t skpv1[SABER_K][SABER_N];
-    uint16_t message[SABER_KEYBYTES * 8];
-    uint16_t res[SABER_K][SABER_N];
-    uint16_t mod_p = SABER_P - 1;
-    uint16_t mod_q = SABER_Q - 1;
-    uint16_t vprime[SABER_N];
-    unsigned char msk_c[SABER_SCALEBYTES_KEM];
+void PQCLEAN_SABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]) {
+    size_t i, j;
 
-    // extract the seedbytes from Public Key.
-    for (i = 0; i < SABER_SEEDBYTES; i++) {
-        seed[i] = pk[ SABER_POLYVECCOMPRESSEDBYTES + i];
-    }
+    poly A[SABER_L][SABER_L];
+    poly res[SABER_L];
+    poly s[SABER_L];
+    poly *temp = A[0]; // re-use stack space
+    poly *vprime = &A[0][0];
+    poly *message = &A[0][1];
 
-    GenMatrix(a, seed);
+    const uint8_t *seed_A = pk + SABER_POLYVECCOMPRESSEDBYTES;
+    uint8_t *msk_c = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
 
-    // generate secret from constant-time binomial distribution
-    PQCLEAN_SABER_CLEAN_GenSecret(skpv1, noiseseed);
+    PQCLEAN_SABER_CLEAN_GenSecret(s, noiseseed);
+    PQCLEAN_SABER_CLEAN_GenMatrix(A, seed_A);
+    PQCLEAN_SABER_CLEAN_MatrixVectorMul(res, (const poly (*)[SABER_L])A, (const poly *)s, 0); // 0 => not transposed
 
-    // matrix-vector multiplication and rounding
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = 0;
-        }
-    }
-    MatrixVectorMul(a, skpv1, res, SABER_Q - 1, 0);
-
-    // now rounding
-    //shift right 3 bits
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            res[i][j] = ( res[i][j] + h1 ) & mod_q;
-            res[i][j] = (res[i][j] >> (SABER_EQ - SABER_EP) );
-        }
-    }
 
-    PQCLEAN_SABER_CLEAN_POLVEC2BS(ciphertext, res, SABER_P);
-
-    // ************client matrix-vector multiplication ends************
-
-    // now calculate the v'
-    // unpack the public_key
-    // pkcl is the b in the protocol
-    PQCLEAN_SABER_CLEAN_BS2POLVEC(pk, pkcl, SABER_P);
-    for (i = 0; i < SABER_N; i++) {
-        vprime[i] = 0;
-    }
-    for (i = 0; i < SABER_K; i++) {
+    // rounding
+    for (i = 0; i < SABER_L; i++) { //shift right EQ-EP bits
         for (j = 0; j < SABER_N; j++) {
-            skpv1[i][j] = skpv1[i][j] & (mod_p);
+            res[i].coeffs[j] += h1;
+            res[i].coeffs[j] >>= SABER_EQ - SABER_EP;
+            res[i].coeffs[j] &= SABER_Q - 1;
         }
     }
+    PQCLEAN_SABER_CLEAN_POLVECp2BS(ciphertext, res);
 
     // vector-vector scalar multiplication with mod p
-    InnerProd(pkcl, skpv1, mod_p, vprime);
+    PQCLEAN_SABER_CLEAN_BS2POLVECp(temp, pk);
+    PQCLEAN_SABER_CLEAN_InnerProd(vprime, temp, s);
+    PQCLEAN_SABER_CLEAN_BS2POLmsg(message, m);
 
-    // addition of h1 to vprime
     for (i = 0; i < SABER_N; i++) {
-        vprime[i] = vprime[i] + h1;
+        vprime->coeffs[i] += h1 - (message->coeffs[i] << (SABER_EP - 1));
+        vprime->coeffs[i] &= SABER_P - 1;
+        vprime->coeffs[i] >>= SABER_EP - SABER_ET;
     }
 
-    // unpack message_received;
-    for (j = 0; j < SABER_KEYBYTES; j++) {
-        for (i = 0; i < 8; i++) {
-            message[8 * j + i] = ((message_received[j] >> i) & 0x01);
-        }
-    }
-
-    // message encoding
-    for (i = 0; i < SABER_N; i++) {
-        message[i] = (message[i] << (SABER_EP - 1));
-    }
-
-    for (k = 0; k < SABER_N; k++) {
-        vprime[k] = ( (vprime[k] - message[k]) & (mod_p) ) >> (SABER_EP - SABER_ET);
-    }
-
-
-    PQCLEAN_SABER_CLEAN_pack_4bit(msk_c, vprime);
-
-    for (j = 0; j < SABER_SCALEBYTES_KEM; j++) {
-        ciphertext[SABER_POLYVECCOMPRESSEDBYTES + j] = msk_c[j];
-    }
+    PQCLEAN_SABER_CLEAN_POLT2BS(msk_c, vprime);
 }
 
 
-void PQCLEAN_SABER_CLEAN_indcpa_kem_dec(const unsigned char *sk, const unsigned char *ciphertext, unsigned char message_dec[]) {
-    uint32_t i, j;
-    // secret key of the server
-    uint16_t sksv[SABER_K][SABER_N];
-    uint16_t pksv[SABER_K][SABER_N];
-    uint8_t scale_ar[SABER_SCALEBYTES_KEM];
-    uint16_t mod_p = SABER_P - 1;
-    uint16_t v[SABER_N];
-    uint16_t op[SABER_N];
+void PQCLEAN_SABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]) {
+    size_t i;
 
-    // sksv is the secret-key
-    PQCLEAN_SABER_CLEAN_BS2POLVEC(sk, sksv, SABER_Q);
-    // pksv is the ciphertext
-    PQCLEAN_SABER_CLEAN_BS2POLVEC(ciphertext, pksv, SABER_P);
+    poly temp[SABER_L];
+    poly s[SABER_L];
 
-    // vector-vector scalar multiplication with mod p
-    for (i = 0; i < SABER_N; i++) {
-        v[i] = 0;
-    }
-    for (i = 0; i < SABER_K; i++) {
-        for (j = 0; j < SABER_N; j++) {
-            sksv[i][j] = sksv[i][j] & (mod_p);
-        }
-    }
-    InnerProd(pksv, sksv, mod_p, v);
+    const uint8_t *packed_cm = ciphertext + SABER_POLYVECCOMPRESSEDBYTES;
+    poly *v = &temp[0];
+    poly *cm = &temp[1];
 
-    //Extraction
-    for (i = 0; i < SABER_SCALEBYTES_KEM; i++) {
-        scale_ar[i] = ciphertext[SABER_POLYVECCOMPRESSEDBYTES + i];
-    }
+    PQCLEAN_SABER_CLEAN_BS2POLVECq(s, sk);
+    PQCLEAN_SABER_CLEAN_BS2POLVECp(temp, ciphertext);
+    PQCLEAN_SABER_CLEAN_InnerProd(&temp[0], temp, s);
 
-    PQCLEAN_SABER_CLEAN_un_pack4bit(scale_ar, op);
+    PQCLEAN_SABER_CLEAN_BS2POLT(cm, packed_cm);
 
-    //addition of h1
     for (i = 0; i < SABER_N; i++) {
-        v[i] = ( ( v[i] + h2 - (op[i] << (SABER_EP - SABER_ET)) ) & (mod_p) ) >> (SABER_EP - 1);
+        v->coeffs[i] += h2 - (cm->coeffs[i] << (SABER_EP - SABER_ET));
+        v->coeffs[i] &= SABER_P - 1;
+        v->coeffs[i] >>= SABER_EP - 1;
     }
 
-    // pack decrypted message
-    POL2MSG(v, message_dec);
-}
-static void MatrixVectorMul(polyvec *a, uint16_t skpv[SABER_K][SABER_N], uint16_t res[SABER_K][SABER_N], uint16_t mod, int16_t transpose) {
-    uint16_t acc[SABER_N];
-    int32_t i, j, k;
-
-    if (transpose == 1) {
-        for (i = 0; i < SABER_K; i++) {
-            for (j = 0; j < SABER_K; j++) {
-                PQCLEAN_SABER_CLEAN_pol_mul((uint16_t *)&a[j].vec[i], skpv[j], acc, SABER_Q, SABER_N);
-
-                for (k = 0; k < SABER_N; k++) {
-                    res[i][k] = res[i][k] + acc[k];
-                    //reduction mod p
-                    res[i][k] = (res[i][k] & mod);
-                    //clear the accumulator
-                    acc[k] = 0;
-                }
-            }
-        }
-    } else {
-        for (i = 0; i < SABER_K; i++) {
-            for (j = 0; j < SABER_K; j++) {
-                PQCLEAN_SABER_CLEAN_pol_mul((uint16_t *)&a[i].vec[j], skpv[j], acc, SABER_Q, SABER_N);
-                for (k = 0; k < SABER_N; k++) {
-                    res[i][k] = res[i][k] + acc[k];
-                    // reduction
-                    res[i][k] = res[i][k] & mod;
-                    // clear the accumulator
-                    acc[k] = 0;
-                }
-            }
-        }
-    }
-}
-
-static void POL2MSG(const uint16_t *message_dec_unpacked, unsigned char *message_dec) {
-    int32_t i, j;
-
-    for (j = 0; j < SABER_KEYBYTES; j++) {
-        message_dec[j] = 0;
-        for (i = 0; i < 8; i++) {
-            message_dec[j] = message_dec[j] | (uint8_t) (message_dec_unpacked[j * 8 + i] << i);
-        }
-    }
-}
-
-
-static void InnerProd(uint16_t pkcl[SABER_K][SABER_N], uint16_t skpv[SABER_K][SABER_N], uint16_t mod, uint16_t res[SABER_N]) {
-    uint32_t j, k;
-    uint16_t acc[SABER_N];
-
-    // vector-vector scalar multiplication with mod p
-    for (j = 0; j < SABER_K; j++) {
-        PQCLEAN_SABER_CLEAN_pol_mul(pkcl[j], skpv[j], acc, SABER_P, SABER_N);
-
-        for (k = 0; k < SABER_N; k++) {
-            res[k] = res[k] + acc[k];
-            // reduction
-            res[k] = res[k] & mod;
-            // clear the accumulator
-            acc[k] = 0;
-        }
-    }
+    PQCLEAN_SABER_CLEAN_POLmsg2BS(m, v);
 }

crypto_kem/saber/clean/SABER_indcpa.h

@@ -1,9 +1,13 @@
 #ifndef INDCPA_H
 #define INDCPA_H
+#include "SABER_params.h"
+#include <stdint.h>
 
-void PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(unsigned char *pk, unsigned char *sk);
-void PQCLEAN_SABER_CLEAN_indcpa_kem_enc(const unsigned char *message, unsigned char *noiseseed, const unsigned char *pk, unsigned char *ciphertext);
-void PQCLEAN_SABER_CLEAN_indcpa_kem_dec(const unsigned char *sk, const unsigned char *ciphertext, unsigned char *message_dec);
+void PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES], uint8_t sk[SABER_INDCPA_SECRETKEYBYTES]);
 
-#endif
+void PQCLEAN_SABER_CLEAN_indcpa_kem_enc(uint8_t ciphertext[SABER_BYTES_CCA_DEC], const uint8_t m[SABER_KEYBYTES], const uint8_t noiseseed[SABER_NOISESEEDBYTES], const uint8_t pk[SABER_INDCPA_PUBLICKEYBYTES]);
+
+void PQCLEAN_SABER_CLEAN_indcpa_kem_dec(uint8_t m[SABER_KEYBYTES], const uint8_t sk[SABER_INDCPA_SECRETKEYBYTES], const uint8_t ciphertext[SABER_BYTES_CCA_DEC]);
 
+
+#endif

crypto_kem/saber/clean/SABER_params.h

@@ -1,50 +1,41 @@
 #ifndef PARAMS_H
 #define PARAMS_H
 
-#include "api.h"
 
-#define SABER_K 3
+/* Don't change anything below this line */
+#define SABER_L 3
 #define SABER_MU 8
 #define SABER_ET 4
 
-
-#define SABER_EQ 13
-#define SABER_EP 10
-
 #define SABER_N 256
-#define SABER_Q 8192
-#define SABER_P 1024
-
-#define SABER_SEEDBYTES       32
-#define SABER_NOISESEEDBYTES  32
-#define SABER_COINBYTES       32
-#define SABER_KEYBYTES        32
 
-#define SABER_HASHBYTES       32
+#define SABER_EP 10
+#define SABER_P (1 << SABER_EP)
 
-#define SABER_POLYBYTES       416 //13*256/8 
+#define SABER_EQ 13
+#define SABER_Q (1 << SABER_EQ)
 
-#define SABER_POLYVECBYTES    (SABER_K * SABER_POLYBYTES)
+#define SABER_SEEDBYTES 32
+#define SABER_NOISESEEDBYTES 32
+#define SABER_KEYBYTES 32
+#define SABER_HASHBYTES 32
 
-#define SABER_POLYVECCOMPRESSEDBYTES (SABER_K * 320) //10*256/8 NOTE : changed till here due to parameter adaptation
+#define SABER_POLYCOINBYTES (SABER_MU * SABER_N / 8)
 
-#define SABER_CIPHERTEXTBYTES (SABER_POLYVECCOMPRESSEDBYTES)
+#define SABER_POLYBYTES (SABER_EQ * SABER_N / 8)
+#define SABER_POLYVECBYTES (SABER_L * SABER_POLYBYTES)
 
-#define SABER_SCALEBYTES (SABER_DELTA*SABER_N/8)
+#define SABER_POLYCOMPRESSEDBYTES (SABER_EP * SABER_N / 8)
+#define SABER_POLYVECCOMPRESSEDBYTES (SABER_L * SABER_POLYCOMPRESSEDBYTES)
 
-#define SABER_SCALEBYTES_KEM ((SABER_ET)*SABER_N/8)
+#define SABER_SCALEBYTES_KEM (SABER_ET * SABER_N / 8)
 
 #define SABER_INDCPA_PUBLICKEYBYTES (SABER_POLYVECCOMPRESSEDBYTES + SABER_SEEDBYTES)
 #define SABER_INDCPA_SECRETKEYBYTES (SABER_POLYVECBYTES)
 
 #define SABER_PUBLICKEYBYTES (SABER_INDCPA_PUBLICKEYBYTES)
+#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES + SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
 
-#define SABER_SECRETKEYBYTES (SABER_INDCPA_SECRETKEYBYTES +  SABER_INDCPA_PUBLICKEYBYTES + SABER_HASHBYTES + SABER_KEYBYTES)
-
-#define SABER_BYTES_CCA_DEC   (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM) /* Second part is for Targhi-Unruh */
-
-
-
+#define SABER_BYTES_CCA_DEC (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM)
 
 #endif
-

crypto_kem/saber/clean/api.h

@@ -1,14 +1,18 @@
 #ifndef PQCLEAN_SABER_CLEAN_API_H
 #define PQCLEAN_SABER_CLEAN_API_H
 
+
 #define PQCLEAN_SABER_CLEAN_CRYPTO_ALGNAME "Saber"
-#define PQCLEAN_SABER_CLEAN_CRYPTO_SECRETKEYBYTES 2304
-#define PQCLEAN_SABER_CLEAN_CRYPTO_PUBLICKEYBYTES (3*320+32)
 #define PQCLEAN_SABER_CLEAN_CRYPTO_BYTES 32
 #define PQCLEAN_SABER_CLEAN_CRYPTO_CIPHERTEXTBYTES 1088
+#define PQCLEAN_SABER_CLEAN_CRYPTO_PUBLICKEYBYTES 992
+#define PQCLEAN_SABER_CLEAN_CRYPTO_SECRETKEYBYTES 2304
 
 int PQCLEAN_SABER_CLEAN_crypto_kem_keypair(unsigned char *pk, unsigned char *sk);
-int PQCLEAN_SABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *ss, const unsigned char *pk);
-int PQCLEAN_SABER_CLEAN_crypto_kem_dec(unsigned char *ss, const unsigned char *ct, const unsigned char *sk);
 
-#endif /* api_h */
+int PQCLEAN_SABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *k, const unsigned char *pk);
+
+int PQCLEAN_SABER_CLEAN_crypto_kem_dec(unsigned char *k, const unsigned char *ct, const unsigned char *sk);
+
+
+#endif /* PQCLEAN_SABER_CLEAN_API_H */

crypto_kem/saber/clean/cbd.c

@@ -1,3 +1,7 @@
+#include "SABER_params.h"
+#include "api.h"
+#include "cbd.h"
+#include <stdint.h>
 /*---------------------------------------------------------------------
 This file has been adapted from the implementation
 (available at, Public Domain https://github.com/pq-crystals/kyber)
@@ -6,12 +10,8 @@ by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------------------------*/
 
-#include "SABER_params.h"
-#include "api.h"
-#include "cbd.h"
-#include <stdint.h>
 
-static uint64_t load_littleendian(const unsigned char *x, int bytes) {
+static uint64_t load_littleendian(const uint8_t *x, int bytes) {
     int i;
     uint64_t r = x[0];
     for (i = 1; i < bytes; i++) {
@@ -20,10 +20,7 @@ static uint64_t load_littleendian(const unsigned char *x, int bytes) {
     return r;
 }
 
-
-void PQCLEAN_SABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
-    uint16_t Qmod_minus1 = SABER_Q - 1;
-
+void PQCLEAN_SABER_CLEAN_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]) {
     uint32_t t, d, a[4], b[4];
     int i, j;
 
@@ -34,18 +31,18 @@ void PQCLEAN_SABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
             d += (t >> j) & 0x11111111;
         }
 
-        a[0] =  d & 0xf;
-        b[0] = (d >>  4) & 0xf;
-        a[1] = (d >>  8) & 0xf;
+        a[0] = d & 0xf;
+        b[0] = (d >> 4) & 0xf;
+        a[1] = (d >> 8) & 0xf;
         b[1] = (d >> 12) & 0xf;
         a[2] = (d >> 16) & 0xf;
         b[2] = (d >> 20) & 0xf;
         a[3] = (d >> 24) & 0xf;
         b[3] = (d >> 28);
 
-        r[4 * i + 0] = (uint16_t)(a[0]  - b[0]) & Qmod_minus1;
-        r[4 * i + 1] = (uint16_t)(a[1]  - b[1]) & Qmod_minus1;
-        r[4 * i + 2] = (uint16_t)(a[2]  - b[2]) & Qmod_minus1;
-        r[4 * i + 3] = (uint16_t)(a[3]  - b[3]) & Qmod_minus1;
+        s[4 * i + 0] = (uint16_t)(a[0] - b[0]);
+        s[4 * i + 1] = (uint16_t)(a[1] - b[1]);
+        s[4 * i + 2] = (uint16_t)(a[2] - b[2]);
+        s[4 * i + 3] = (uint16_t)(a[3] - b[3]);
     }
 }

crypto_kem/saber/clean/cbd.h

@@ -1,6 +1,5 @@
 #ifndef CBD_H
 #define CBD_H
-
 /*---------------------------------------------------------------------
 This file has been adapted from the implementation
 (available at, Public Domain https://github.com/pq-crystals/kyber)
@@ -8,10 +7,10 @@ of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
 by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
 Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
 ----------------------------------------------------------------------*/
-
-#include "poly.h"
+#include "SABER_params.h"
 #include <stdint.h>
 
-void PQCLEAN_SABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf);
+void PQCLEAN_SABER_CLEAN_cbd(uint16_t s[SABER_N], const uint8_t buf[SABER_POLYCOINBYTES]);
+
 
 #endif

crypto_kem/saber/clean/kem.c

@@ -1,96 +1,77 @@
 #include "SABER_indcpa.h"
 #include "SABER_params.h"
+#include "api.h"
 #include "fips202.h"
 #include "randombytes.h"
 #include "verify.h"
+#include <stddef.h>
 #include <stdint.h>
-#include <stdio.h>
-#include <string.h>
 
-int PQCLEAN_SABER_CLEAN_crypto_kem_keypair(unsigned char *pk, unsigned char *sk) {
-    int i;
 
-    // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
-    PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(pk, sk);
+int PQCLEAN_SABER_CLEAN_crypto_kem_keypair(uint8_t *pk, uint8_t *sk) {
+    size_t i;
 
-    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
+    PQCLEAN_SABER_CLEAN_indcpa_kem_keypair(pk, sk); // sk[0:SABER_INDCPA_SECRETKEYBYTES-1] <-- sk
     for (i = 0; i < SABER_INDCPA_PUBLICKEYBYTES; i++) {
-        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];
+        sk[i + SABER_INDCPA_SECRETKEYBYTES] = pk[i];    // sk[SABER_INDCPA_SECRETKEYBYTES:SABER_INDCPA_SECRETKEYBYTES+SABER_INDCPA_SECRETKEYBYTES-1] <-- pk
     }
 
-    // Then hash(pk) is appended.
-    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES);
+    sha3_256(sk + SABER_SECRETKEYBYTES - 64, pk, SABER_INDCPA_PUBLICKEYBYTES); // Then hash(pk) is appended.
 
-    // Remaining part of sk contains a pseudo-random number.
-    // This is output when check in crypto_kem_dec() fails.
-    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES );
+    randombytes(sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES); // Remaining part of sk contains a pseudo-random number.
+    // This is output when check in PQCLEAN_SABER_CLEAN_crypto_kem_dec() fails.
     return (0);
 }
 
-int PQCLEAN_SABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *ss, const unsigned char *pk) {
-    // Will contain key, coins
-    unsigned char kr[64];
-    unsigned char buf[64];
+int PQCLEAN_SABER_CLEAN_crypto_kem_enc(uint8_t *c, uint8_t *k, const uint8_t *pk) {
 
-    randombytes(buf, 32);
+    uint8_t kr[64]; // Will contain key, coins
+    uint8_t buf[64];
 
-    // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
-    sha3_256(buf, buf, 32);
+    randombytes(buf, 32);
 
-    // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
-    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES);
+    sha3_256(buf, buf, 32); // BUF[0:31] <-- random message (will be used as the key for client) Note: hash doesnot release system RNG output
 
-    // kr[0:63] <-- Hash(buf[0:63]);
-    sha3_512(kr, buf, 64);
+    sha3_256(buf + 32, pk, SABER_INDCPA_PUBLICKEYBYTES); // BUF[32:63] <-- Hash(public key);  Multitarget countermeasure for coins + contributory KEM
 
+    sha3_512(kr, buf, 64);               // kr[0:63] <-- Hash(buf[0:63]);
     // K^ <-- kr[0:31]
     // noiseseed (r) <-- kr[32:63];
-    // buf[0:31] contains message; kr[32:63] contains randomness r;
-    PQCLEAN_SABER_CLEAN_indcpa_kem_enc(buf, kr + 32, pk,  ct);
+    PQCLEAN_SABER_CLEAN_indcpa_kem_enc(c, buf, kr + 32, pk); // buf[0:31] contains message; kr[32:63] contains randomness r;
 
-    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC);
 
-    // hash concatenation of pre-k and h(c) to k
-    sha3_256(ss, kr, 64);
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
 
     return (0);
 }
 
+int PQCLEAN_SABER_CLEAN_crypto_kem_dec(uint8_t *k, const uint8_t *c, const uint8_t *sk) {
+    size_t i;
+    uint8_t fail;
+    uint8_t cmp[SABER_BYTES_CCA_DEC];
+    uint8_t buf[64];
+    uint8_t kr[64]; // Will contain key, coins
+    const uint8_t *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
 
-int PQCLEAN_SABER_CLEAN_crypto_kem_dec(unsigned char *ss, const unsigned char *ct, const unsigned char *sk) {
-    int i;
-    unsigned char fail;
-    unsigned char cmp[SABER_BYTES_CCA_DEC];
-    unsigned char buf[64];
-
-    // Will contain key, coins
-    unsigned char kr[64];
-    const unsigned char *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
-
-    // buf[0:31] <-- message
-    PQCLEAN_SABER_CLEAN_indcpa_kem_dec(sk, ct, buf);
-
+    PQCLEAN_SABER_CLEAN_indcpa_kem_dec(buf, sk, c); // buf[0:31] <-- message
 
     // Multitarget countermeasure for coins + contributory KEM
-    // Save hash by storing h(pk) in sk
-    for (i = 0; i < 32; i++) {
+    for (i = 0; i < 32; i++) { // Save hash by storing h(pk) in sk
         buf[32 + i] = sk[SABER_SECRETKEYBYTES - 64 + i];
     }
 
     sha3_512(kr, buf, 64);
 
-    PQCLEAN_SABER_CLEAN_indcpa_kem_enc(buf, kr + 32, pk, cmp);
-
+    PQCLEAN_SABER_CLEAN_indcpa_kem_enc(cmp, buf, kr + 32, pk);
 
-    fail = PQCLEAN_SABER_CLEAN_verify(ct, cmp, SABER_BYTES_CCA_DEC);
+    fail = PQCLEAN_SABER_CLEAN_verify(c, cmp, SABER_BYTES_CCA_DEC);
 
-    // overwrite coins in kr with h(c)
-    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);
+    sha3_256(kr + 32, c, SABER_BYTES_CCA_DEC); // overwrite coins in kr with h(c)
 
     PQCLEAN_SABER_CLEAN_cmov(kr, sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES, fail);
 
-    // hash concatenation of pre-k and h(c) to k
-    sha3_256(ss, kr, 64);
+    sha3_256(k, kr, 64); // hash concatenation of pre-k and h(c) to k
 
     return (0);
 }

crypto_kem/saber/clean/pack_unpack.c

@@ -1,254 +1,147 @@
+#include "SABER_params.h"
 #include "pack_unpack.h"
+#include "poly.h"
+#include <string.h>
 
-void PQCLEAN_SABER_CLEAN_pack_3bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
-    for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 3 * j;
-        offset_data = 8 * j;
-        bytes[offset_byte + 0] = (data[offset_data + 0] & 0x7) |
-                                 ((data[offset_data + 1] & 0x7) << 3) |
-                                 ((data[offset_data + 2] & 0x3) << 6);
-        bytes[offset_byte + 1] = ((data[offset_data + 2] >> 2 ) & 0x01)  |
-                                 ((data[offset_data + 3] & 0x7) << 1) |
-                                 ((data[offset_data + 4] & 0x7) << 4) |
-                                 (((data[offset_data + 5]) & 0x01) << 7);
-        bytes[offset_byte + 2] = ((data[offset_data + 5] >> 1 ) & 0x03) |
-                                 ((data[offset_data + 6] & 0x7) << 2) |
-                                 ((data[offset_data + 7] & 0x7) << 5);
+void PQCLEAN_SABER_CLEAN_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 2; j++) {
+        out[0] = (uint8_t) ((in[0] & 0x0f) | (in[1] << 4));
+        in += 2;
+        out += 1;
     }
 }
 
-void PQCLEAN_SABER_CLEAN_un_pack3bit(const uint8_t *bytes, uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
-    for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 3 * j;
-        offset_data = 8 * j;
-        data[offset_data + 0] = (bytes[offset_byte + 0]) & 0x07;
-        data[offset_data + 1] = ((bytes[offset_byte + 0]) >> 3 ) & 0x07;
-        data[offset_data + 2] = (((bytes[offset_byte + 0]) >> 6 ) & 0x03) |
-                                (((bytes[offset_byte + 1]) & 0x01) << 2);
-        data[offset_data + 3] = ((bytes[offset_byte + 1]) >> 1 ) & 0x07;
-        data[offset_data + 4] = ((bytes[offset_byte + 1]) >> 4 ) & 0x07;
-        data[offset_data + 5] = (((bytes[offset_byte + 1]) >> 7 ) & 0x01) |
-                                (((bytes[offset_byte + 2]) & 0x03) << 1);
-        data[offset_data + 6] = ((bytes[offset_byte + 2] >> 2) & 0x07);
-        data[offset_data + 7] = ((bytes[offset_byte + 2] >> 5) & 0x07);
+void PQCLEAN_SABER_CLEAN_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]) {
+    /* This function does not reduce its output mod T */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 2; j++) {
+        out[0] = in[0];
+        out[1] = in[0] >> 4;
+        in += 1;
+        out += 2;
     }
 }
 
-void PQCLEAN_SABER_CLEAN_pack_4bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data;
-
-    for (j = 0; j < SABER_N / 2; j++) {
-        offset_data = 2 * j;
-        bytes[j] = (data[offset_data] & 0x0f) |
-                   ((data[offset_data + 1] & 0x0f) << 4);
+static void POLq2BS(uint8_t bytes[SABER_POLYBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x1f) | (in[1] << 5));
+        out[2] = (uint8_t) (in[1] >> 3);
+        out[3] = (uint8_t) (((in[1] >> 11) & 0x03) | (in[2] << 2));
+        out[4] = (uint8_t) (((in[2] >> 6) & 0x7f) | (in[3] << 7));
+        out[5] = (uint8_t) (in[3] >> 1);
+        out[6] = (uint8_t) (((in[3] >> 9) & 0x0f) | (in[4] << 4));
+        out[7] = (uint8_t) (in[4] >> 4);
+        out[8] = (uint8_t) (((in[4] >> 12) & 0x01) | (in[5] << 1));
+        out[9] = (uint8_t) (((in[5] >> 7) & 0x3f) | (in[6] << 6));
+        out[10] = (uint8_t) (in[6] >> 2);
+        out[11] = (uint8_t) (((in[6] >> 10) & 0x07) | (in[7] << 3));
+        out[12] = (uint8_t) (in[7] >> 5);
+        in += 8;
+        out += 13;
     }
 }
 
-void PQCLEAN_SABER_CLEAN_un_pack4bit(const unsigned char *bytes, uint16_t *ar) {
-    uint32_t j;
-    uint32_t offset_data;
-
-    for (j = 0; j < SABER_N / 2; j++) {
-        offset_data = 2 * j;
-        ar[offset_data] = bytes[j] & 0x0f;
-        ar[offset_data + 1] = (bytes[j] >> 4) & 0x0f;
+static void BS2POLq(poly *data, const uint8_t bytes[SABER_POLYBYTES]) {
+    /* This function does not reduce its output mod Q */
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
+    for (j = 0; j < SABER_N / 8; j++) {
+        out[0] = (in[0]) | (in[1] << 8);
+        out[1] = (in[1] >> 5) | (in[2] << 3) | (in[3] << 11);
+        out[2] = (in[3] >> 2) | (in[4] << 6);
+        out[3] = (in[4] >> 7) | (in[5] << 1) | (in[6] << 9);
+        out[4] = (in[6] >> 4) | (in[7] << 4) | (in[8] << 12);
+        out[5] = (in[8] >> 1) | (in[9] << 7);
+        out[6] = (in[9] >> 6) | (in[10] << 2) | (in[11] << 10);
+        out[7] = (in[11] >> 3) | (in[12] << 5);
+        in += 13;
+        out += 8;
     }
 }
 
-void PQCLEAN_SABER_CLEAN_pack_6bit(uint8_t *bytes, const uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+static void POLp2BS(uint8_t bytes[SABER_POLYCOMPRESSEDBYTES], const poly *data) {
+    size_t j;
+    const uint16_t *in = data->coeffs;
+    uint8_t *out = bytes;
     for (j = 0; j < SABER_N / 4; j++) {
-        offset_byte = 3 * j;
-        offset_data = 4 * j;
-        bytes[offset_byte + 0] = (data[offset_data + 0] & 0x3f) |
-                                 ((data[offset_data + 1] & 0x03) << 6);
-        bytes[offset_byte + 1] = ((data[offset_data + 1] >> 2) & 0x0f) |
-                                 ((data[offset_data + 2] & 0x0f) << 4);
-        bytes[offset_byte + 2] = ((data[offset_data + 2] >> 4) & 0x03) |
-                                 ((data[offset_data + 3] & 0x3f) << 2);
+        out[0] = (uint8_t) (in[0]);
+        out[1] = (uint8_t) (((in[0] >> 8) & 0x03) | (in[1] << 2));
+        out[2] = (uint8_t) (((in[1] >> 6) & 0x0f) | (in[2] << 4));
+        out[3] = (uint8_t) (((in[2] >> 4) & 0x3f) | (in[3] << 6));
+        out[4] = (uint8_t) (in[3] >> 2);
+        in += 4;
+        out += 5;
     }
 }
 
-
-void PQCLEAN_SABER_CLEAN_un_pack6bit(const unsigned char *bytes, uint16_t *data) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
+static void BS2POLp(poly *data, const uint8_t bytes[SABER_POLYCOMPRESSEDBYTES]) {
+    size_t j;
+    const uint8_t *in = bytes;
+    uint16_t *out = data->coeffs;
     for (j = 0; j < SABER_N / 4; j++) {
-        offset_byte = 3 * j;
-        offset_data = 4 * j;
-        data[offset_data + 0] = bytes[offset_byte + 0] & 0x3f;
-        data[offset_data + 1] = ((bytes[offset_byte + 0] >> 6) & 0x03) |
-                                ((bytes[offset_byte + 1] & 0x0f) << 2);
-        data[offset_data + 2] = ((bytes[offset_byte + 1] & 0xff) >> 4) |
-                                ((bytes[offset_byte + 2] & 0x03) << 4);
-        data[offset_data + 3] = ((bytes[offset_byte + 2] & 0xff) >> 2);
+        out[0] = in[0] | (in[1] << 8);
+        out[1] = (in[1] >> 2) | (in[2] << 6);
+        out[2] = (in[2] >> 4) | (in[3] << 4);
+        out[3] = (in[3] >> 6) | (in[4] << 2);
+        in += 5;
+        out += 4;
     }
 }
 
-
-static void POLVECp2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 10) / 8;
-        for (j = 0; j < SABER_N / 4; j++) {
-            offset_byte = offset_byte1 + 5 * j;
-            offset_data = 4 * j;
-            bytes[offset_byte + 0] = (data[i][offset_data + 0] & (0xff));
-            bytes[offset_byte + 1] = ((data[i][offset_data + 0] >> 8) & 0x03) |
-                                     ((data[i][offset_data + 1] & 0x3f) << 2);
-            bytes[offset_byte + 2] = ((data[i][offset_data + 1] >> 6) & 0x0f) |
-                                     ((data[i][offset_data + 2] & 0x0f) << 4);
-            bytes[offset_byte + 3] = ((data[i][offset_data + 2] >> 4) & 0x3f) |
-                                     ((data[i][offset_data + 3] & 0x03) << 6);
-            bytes[offset_byte + 4] = ((data[i][offset_data + 3] >> 2) & 0xff);
-        }
+void PQCLEAN_SABER_CLEAN_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLq2BS(bytes + i * SABER_POLYBYTES, &data[i]);
     }
 }
 
-static void BS2POLVECp(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 10) / 8;
-        for (j = 0; j < SABER_N / 4; j++) {
-            offset_byte = offset_byte1 + 5 * j;
-            offset_data = 4 * j;
-            data[i][offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                       ((bytes[offset_byte + 1] & 0x03) << 8);
-            data[i][offset_data + 1] = ((bytes[offset_byte + 1] >> 2) & (0x3f)) |
-                                       ((bytes[offset_byte + 2] & 0x0f) << 6);
-            data[i][offset_data + 2] = ((bytes[offset_byte + 2] >> 4) & (0x0f)) |
-                                       ((bytes[offset_byte + 3] & 0x3f) << 4);
-            data[i][offset_data + 3] = ((bytes[offset_byte + 3] >> 6) & (0x03)) |
-                                       ((bytes[offset_byte + 4] & 0xff) << 2);
-        }
+void PQCLEAN_SABER_CLEAN_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLq(&data[i], bytes + i * SABER_POLYBYTES);
     }
 }
 
-
-
-static void POLVECq2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 13) / 8;
-        for (j = 0; j < SABER_N / 8; j++) {
-            offset_byte = offset_byte1 + 13 * j;
-            offset_data = 8 * j;
-            bytes[offset_byte + 0] = (data[i][offset_data + 0] & (0xff));
-            bytes[offset_byte + 1] = ((data[i][offset_data + 0] >> 8) & 0x1f) |
-                                     ((data[i][offset_data + 1] & 0x07) << 5);
-            bytes[offset_byte + 2] = ((data[i][offset_data + 1] >> 3) & 0xff);
-            bytes[offset_byte + 3] = ((data[i][offset_data + 1] >> 11) & 0x03) |
-                                     ((data[i][offset_data + 2] & 0x3f) << 2);
-            bytes[offset_byte + 4] = ((data[i][offset_data + 2] >> 6) & 0x7f) |
-                                     ((data[i][offset_data + 3] & 0x01) << 7);
-            bytes[offset_byte + 5] = ((data[i][offset_data + 3] >> 1) & 0xff);
-            bytes[offset_byte + 6] = ((data[i][offset_data + 3] >> 9) & 0x0f) |
-                                     ((data[i][offset_data + 4] & 0x0f) << 4);
-            bytes[offset_byte + 7] = ((data[i][offset_data + 4] >> 4) & 0xff);
-            bytes[offset_byte + 8] = ((data[i][offset_data + 4] >> 12) & 0x01) |
-                                     ((data[i][offset_data + 5] & 0x7f) << 1);
-            bytes[offset_byte + 9] = ((data[i][offset_data + 5] >> 7) & 0x3f) |
-                                     ((data[i][offset_data + 6] & 0x03) << 6);
-            bytes[offset_byte + 10] = ((data[i][offset_data + 6] >> 2) & 0xff);
-            bytes[offset_byte + 11] = ((data[i][offset_data + 6] >> 10) & 0x07) |
-                                      ((data[i][offset_data + 7] & 0x1f) << 3);
-            bytes[offset_byte + 12] = ((data[i][offset_data + 7] >> 5) & 0xff);
-        }
+void PQCLEAN_SABER_CLEAN_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        POLp2BS(bytes + i * SABER_POLYCOMPRESSEDBYTES, &data[i]);
     }
 }
 
-static void BS2POLVECq(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N]) {
-    uint32_t i, j;
-    uint32_t offset_data, offset_byte, offset_byte1;
-
-    for (i = 0; i < SABER_K; i++) {
-        offset_byte1 = i * (SABER_N * 13) / 8;
-        for (j = 0; j < SABER_N / 8; j++) {
-            offset_byte = offset_byte1 + 13 * j;
-            offset_data = 8 * j;
-            data[i][offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                       ((bytes[offset_byte + 1] & 0x1f) << 8);
-            data[i][offset_data + 1] = (bytes[offset_byte + 1] >> 5 & (0x07)) |
-                                       ((bytes[offset_byte + 2] & 0xff) << 3) |
-                                       ((bytes[offset_byte + 3] & 0x03) << 11);
-            data[i][offset_data + 2] = (bytes[offset_byte + 3] >> 2 & (0x3f)) |
-                                       ((bytes[offset_byte + 4] & 0x7f) << 6);
-            data[i][offset_data + 3] = (bytes[offset_byte + 4] >> 7 & (0x01)) |
-                                       ((bytes[offset_byte + 5] & 0xff) << 1) |
-                                       ((bytes[offset_byte + 6] & 0x0f) << 9);
-            data[i][offset_data + 4] = (bytes[offset_byte + 6] >> 4 & (0x0f)) |
-                                       ((bytes[offset_byte + 7] & 0xff) << 4) |
-                                       ((bytes[offset_byte + 8] & 0x01) << 12);
-            data[i][offset_data + 5] = (bytes[offset_byte + 8] >> 1 & (0x7f)) |
-                                       ((bytes[offset_byte + 9] & 0x3f) << 7);
-            data[i][offset_data + 6] = (bytes[offset_byte + 9] >> 6 & (0x03)) |
-                                       ((bytes[offset_byte + 10] & 0xff) << 2) |
-                                       ((bytes[offset_byte + 11] & 0x07) << 10);
-            data[i][offset_data + 7] = (bytes[offset_byte + 11] >> 3 & (0x1f)) |
-                                       ((bytes[offset_byte + 12] & 0xff) << 5);
-        }
+void PQCLEAN_SABER_CLEAN_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]) {
+    size_t i;
+    for (i = 0; i < SABER_L; i++) {
+        BS2POLp(&data[i], bytes + i * SABER_POLYCOMPRESSEDBYTES);
     }
 }
 
-//only BS2POLq no BS2POLp
-void PQCLEAN_SABER_CLEAN_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]) {
-    uint32_t j;
-    uint32_t offset_data, offset_byte;
-
-    for (j = 0; j < SABER_N / 8; j++) {
-        offset_byte = 13 * j;
-        offset_data = 8 * j;
-        data[offset_data + 0] = (bytes[offset_byte + 0] & (0xff)) |
-                                ((bytes[offset_byte + 1] & 0x1f) << 8);
-        data[offset_data + 1] = (bytes[offset_byte + 1] >> 5 & (0x07)) |
-                                ((bytes[offset_byte + 2] & 0xff) << 3) |
-                                ((bytes[offset_byte + 3] & 0x03) << 11);
-        data[offset_data + 2] = (bytes[offset_byte + 3] >> 2 & (0x3f)) |
-                                ((bytes[offset_byte + 4] & 0x7f) << 6);
-        data[offset_data + 3] = (bytes[offset_byte + 4] >> 7 & (0x01)) |
-                                ((bytes[offset_byte + 5] & 0xff) << 1) |
-                                ((bytes[offset_byte + 6] & 0x0f) << 9);
-        data[offset_data + 4] = (bytes[offset_byte + 6] >> 4 & (0x0f)) |
-                                ((bytes[offset_byte + 7] & 0xff) << 4) |
-                                ((bytes[offset_byte + 8] & 0x01) << 12);
-        data[offset_data + 5] = (bytes[offset_byte + 8] >> 1 & (0x7f)) |
-                                ((bytes[offset_byte + 9] & 0x3f) << 7);
-        data[offset_data + 6] = (bytes[offset_byte + 9] >> 6 & (0x03)) |
-                                ((bytes[offset_byte + 10] & 0xff) << 2) |
-                                ((bytes[offset_byte + 11] & 0x07) << 10);
-        data[offset_data + 7] = (bytes[offset_byte + 11] >> 3 & (0x1f)) |
-                                ((bytes[offset_byte + 12] & 0xff) << 5);
+void PQCLEAN_SABER_CLEAN_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]) {
+    size_t i, j;
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            data->coeffs[j * 8 + i] = ((bytes[j] >> i) & 0x01);
+        }
     }
 }
 
-void PQCLEAN_SABER_CLEAN_POLVEC2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus) {
-    if (modulus == 1024) {
-        POLVECp2BS(bytes, data);
-    } else if (modulus == 8192) {
-        POLVECq2BS(bytes, data);
-    }
-}
+void PQCLEAN_SABER_CLEAN_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data) {
+    size_t i, j;
+    memset(bytes, 0, SABER_KEYBYTES);
 
-void PQCLEAN_SABER_CLEAN_BS2POLVEC(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus) {
-    if (modulus == 1024) {
-        BS2POLVECp(bytes, data);
-    } else if (modulus == 8192) {
-        BS2POLVECq(bytes, data);
+    for (j = 0; j < SABER_KEYBYTES; j++) {
+        for (i = 0; i < 8; i++) {
+            bytes[j] = bytes[j] | ((data->coeffs[j * 8 + i] & 0x01) << i);
+        }
     }
 }

crypto_kem/saber/clean/pack_unpack.h

@@ -1,28 +1,28 @@
 #ifndef PACK_UNPACK_H
 #define PACK_UNPACK_H
-
 #include "SABER_params.h"
+#include "poly.h"
 #include <stdint.h>
 #include <stdio.h>
 
+void PQCLEAN_SABER_CLEAN_POLT2BS(uint8_t bytes[SABER_SCALEBYTES_KEM], const poly *data);
+
+void PQCLEAN_SABER_CLEAN_BS2POLT(poly *data, const uint8_t bytes[SABER_SCALEBYTES_KEM]);
 
-void PQCLEAN_SABER_CLEAN_pack_3bit(uint8_t *bytes, const uint16_t *data);
 
-void PQCLEAN_SABER_CLEAN_un_pack3bit(const uint8_t *bytes, uint16_t *data);
+void PQCLEAN_SABER_CLEAN_POLVECq2BS(uint8_t bytes[SABER_POLYVECBYTES], const poly data[SABER_L]);
 
-void PQCLEAN_SABER_CLEAN_pack_4bit(uint8_t *bytes, const uint16_t *data);
+void PQCLEAN_SABER_CLEAN_POLVECp2BS(uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES], const poly data[SABER_L]);
 
-void PQCLEAN_SABER_CLEAN_un_pack4bit(const unsigned char *bytes, uint16_t *ar);
 
-void PQCLEAN_SABER_CLEAN_pack_6bit(uint8_t *bytes, const uint16_t *data);
+void PQCLEAN_SABER_CLEAN_BS2POLVECq(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECBYTES]);
 
-void PQCLEAN_SABER_CLEAN_un_pack6bit(const unsigned char *bytes, uint16_t *data);
+void PQCLEAN_SABER_CLEAN_BS2POLVECp(poly data[SABER_L], const uint8_t bytes[SABER_POLYVECCOMPRESSEDBYTES]);
 
 
-void PQCLEAN_SABER_CLEAN_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]);
+void PQCLEAN_SABER_CLEAN_BS2POLmsg(poly *data, const uint8_t bytes[SABER_KEYBYTES]);
 
-void PQCLEAN_SABER_CLEAN_POLVEC2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus);
+void PQCLEAN_SABER_CLEAN_POLmsg2BS(uint8_t bytes[SABER_KEYBYTES], const poly *data);
 
-void PQCLEAN_SABER_CLEAN_BS2POLVEC(const unsigned char *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus);
 
 #endif

crypto_kem/saber/clean/poly.c

@@ -1,21 +1,57 @@
-/*---------------------------------------------------------------------
-This file has been adapted from the implementation
-(available at, Public Domain https://github.com/pq-crystals/kyber)
-of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
-by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
-Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
-#include "SABER_params.h"
+#include "api.h"
 #include "cbd.h"
 #include "fips202.h"
+#include "pack_unpack.h"
 #include "poly.h"
+#include <stddef.h>
 
-void PQCLEAN_SABER_CLEAN_GenSecret(uint16_t r[SABER_K][SABER_N], const unsigned char *seed) {
-    uint8_t buf[SABER_MU * SABER_N * SABER_K / 8];
+void PQCLEAN_SABER_CLEAN_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const poly s[SABER_L], int16_t transpose) {
+    size_t i, j;
+
+    if (transpose) {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_SABER_CLEAN_poly_mul(&c[i], &A[0][i], &s[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_SABER_CLEAN_poly_mul(&c[i], &A[j][i], &s[j], 1);
+            }
+        }
+    } else {
+        for (i = 0; i < SABER_L; i++) {
+            PQCLEAN_SABER_CLEAN_poly_mul(&c[i], &A[i][0], &s[0], 0);
+            for (j = 1; j < SABER_L; j++) {
+                PQCLEAN_SABER_CLEAN_poly_mul(&c[i], &A[i][j], &s[j], 1);
+            }
+        }
+    }
+}
+
+void PQCLEAN_SABER_CLEAN_InnerProd(poly *c, const poly b[SABER_L], const poly s[SABER_L]) {
+    size_t i;
+
+    PQCLEAN_SABER_CLEAN_poly_mul(c, &b[0], &s[0], 0);
+    for (i = 1; i < SABER_L; i++) {
+        PQCLEAN_SABER_CLEAN_poly_mul(c, &b[i], &s[i], 1);
+    }
+}
+
+void PQCLEAN_SABER_CLEAN_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYVECBYTES];
+
+    shake128(buf, sizeof(buf), seed, SABER_SEEDBYTES);
+
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_SABER_CLEAN_BS2POLVECq(A[i], buf + i * SABER_POLYVECBYTES);
+    }
+}
+
+void PQCLEAN_SABER_CLEAN_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]) {
+    size_t i;
+    uint8_t buf[SABER_L * SABER_POLYCOINBYTES];
 
     shake128(buf, sizeof(buf), seed, SABER_NOISESEEDBYTES);
 
-    for (size_t i = 0; i < SABER_K; i++) {
-        PQCLEAN_SABER_CLEAN_cbd(r[i], buf + i * SABER_MU * SABER_N / 8);
+    for (i = 0; i < SABER_L; i++) {
+        PQCLEAN_SABER_CLEAN_cbd(s[i].coeffs, buf + i * SABER_POLYCOINBYTES);
     }
 }

crypto_kem/saber/clean/poly.h

@@ -1,26 +1,23 @@
 #ifndef POLY_H
 #define POLY_H
-
-/*---------------------------------------------------------------------
-This file has been adapted from the implementation
-(available at, Public Domain https://github.com/pq-crystals/kyber)
-of "CRYSTALS – Kyber: a CCA-secure module-lattice-based KEM"
-by : Joppe Bos, Leo Ducas, Eike Kiltz, Tancrede Lepoint,
-Vadim Lyubashevsky, John M. Schanck, Peter Schwabe & Damien stehle
-
-
 #include "SABER_params.h"
 #include <stdint.h>
 
-typedef struct {
+typedef union {
     uint16_t coeffs[SABER_N];
 } poly;
 
-typedef struct {
-    poly vec[SABER_K];
-} polyvec;
 
-void PQCLEAN_SABER_CLEAN_GenSecret(uint16_t r[SABER_K][SABER_N], const unsigned char *seed);
+void PQCLEAN_SABER_CLEAN_MatrixVectorMul(poly c[SABER_L], const poly A[SABER_L][SABER_L], const poly s[SABER_L], int16_t transpose);
+
+void PQCLEAN_SABER_CLEAN_InnerProd(poly *c, const poly b[SABER_L], const poly s[SABER_L]);
+
+void PQCLEAN_SABER_CLEAN_GenMatrix(poly A[SABER_L][SABER_L], const uint8_t seed[SABER_SEEDBYTES]);
+
+void PQCLEAN_SABER_CLEAN_GenSecret(poly s[SABER_L], const uint8_t seed[SABER_NOISESEEDBYTES]);
+
+
+void PQCLEAN_SABER_CLEAN_poly_mul(poly *c, const poly *a, const poly *b, int accumulate);
+
 
 #endif

crypto_kem/saber/clean/poly_mul.c

@@ -1,4 +1,4 @@
-#include "poly_mul.h"
+#include "poly.h"
 #include <stdint.h>
 #include <string.h>
 
@@ -11,13 +11,13 @@
 #define OVERFLOWING_MUL(X, Y) ((uint16_t)((uint32_t)(X) * (uint32_t)(Y)))
 
 #define KARATSUBA_N 64
-static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t *result_final) {
+static void karatsuba_simple(uint16_t *result_final, const uint16_t *a_1, const uint16_t *b_1) {
     uint16_t d01[KARATSUBA_N / 2 - 1];
     uint16_t d0123[KARATSUBA_N / 2 - 1];
     uint16_t d23[KARATSUBA_N / 2 - 1];
     uint16_t result_d01[KARATSUBA_N - 1];
 
-    int32_t i, j;
+    size_t i, j;
 
     memset(result_d01, 0, (KARATSUBA_N - 1)*sizeof(uint16_t));
     memset(d01, 0, (KARATSUBA_N / 2 - 1)*sizeof(uint16_t));
@@ -110,7 +110,7 @@ static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t
 
 
 
-static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *result) {
+static void toom_cook_4way (uint16_t *result, const uint16_t *a1, const uint16_t *b1) {
     uint16_t inv3 = 43691, inv9 = 36409, inv15 = 61167;
 
     uint16_t aw1[N_SB], aw2[N_SB], aw3[N_SB], aw4[N_SB], aw5[N_SB], aw6[N_SB], aw7[N_SB];
@@ -181,13 +181,13 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
 
     // MULTIPLICATION
 
-    karatsuba_simple(aw1, bw1, w1);
-    karatsuba_simple(aw2, bw2, w2);
-    karatsuba_simple(aw3, bw3, w3);
-    karatsuba_simple(aw4, bw4, w4);
-    karatsuba_simple(aw5, bw5, w5);
-    karatsuba_simple(aw6, bw6, w6);
-    karatsuba_simple(aw7, bw7, w7);
+    karatsuba_simple(w1, aw1, bw1);
+    karatsuba_simple(w2, aw2, bw2);
+    karatsuba_simple(w3, aw3, bw3);
+    karatsuba_simple(w4, aw4, bw4);
+    karatsuba_simple(w5, aw5, bw5);
+    karatsuba_simple(w6, aw6, bw6);
+    karatsuba_simple(w7, aw7, bw7);
 
     // INTERPOLATION
     for (i = 0; i < N_SB_RES; ++i) {
@@ -228,19 +228,21 @@ static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *re
     }
 }
 
-void PQCLEAN_SABER_CLEAN_pol_mul(uint16_t *a, uint16_t *b, uint16_t *res, uint16_t p, uint32_t n) {
-    uint32_t i;
-    // normal multiplication
-    uint16_t c[512];
+/* res += a*b */
+void PQCLEAN_SABER_CLEAN_poly_mul(poly *c, const poly *a, const poly *b, const int accumulate) {
+    uint16_t C[2 * SABER_N] = {0};
+    size_t i;
 
-    for (i = 0; i < 512; i++) {
-        c[i] = 0;
-    }
-
-    toom_cook_4way(a, b, c);
+    toom_cook_4way(C, a->coeffs, b->coeffs);
 
-    // reduction
-    for (i = n; i < 2 * n; i++) {
-        res[i - n] = (c[i - n] - c[i]) & (p - 1);
+    /* reduction */
+    if (accumulate == 0) {
+        for (i = SABER_N; i < 2 * SABER_N; i++) {
+            c->coeffs[i - SABER_N] = (C[i - SABER_N] - C[i]);
+        }
+    } else {
+        for (i = SABER_N; i < 2 * SABER_N; i++) {
+            c->coeffs[i - SABER_N] += (C[i - SABER_N] - C[i]);
+        }
     }
 }

crypto_kem/saber/clean/poly_mul.h

@@ -1,9 +1,3 @@
-#ifndef POLYMUL_H
-#define POLYMUL_H
 
-#include "SABER_params.h"
-#include <stdint.h>
 
-void PQCLEAN_SABER_CLEAN_pol_mul(uint16_t *a, uint16_t *b, uint16_t *res, uint16_t p, uint32_t n);
 
-#endif