add Saber

crypto_kem/saber/META.yml

@@ -0,0 +1,13 @@
+name: Saber 
+type: kem
+claimed-nist-level: 3
+claimed-security: IND-CCA2
+length-public-key: 992
+length-ciphertext: 1088
+length-secret-key: 2304
+length-shared-secret: 32
+nistkat-sha256: c9e2c16f41f162c607a1d5704107159e5e12713b9bb8c356b1d68b216e79096e
+principal-submitter: Jan-Pieter D'Anvers, Angshuman Karmakar, Sujoy Sinha Roy, Frederik Vercauteren
+implementations:
+    - name: clean
+      version: https://github.com/KULeuven-COSIC/SABER/commit/14ede83f1ff3bcc41f0464543542366c68b55871

crypto_kem/saber/clean/LICENSE

@@ -0,0 +1 @@
+TODO

crypto_kem/saber/clean/Makefile

@@ -0,0 +1,19 @@
+# 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 
+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)
+
+all: $(LIB)
+
+%.o: %.c $(HEADERS)
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(LIB): $(OBJECTS)
+	$(AR) -r $@ $(OBJECTS)
+
+clean:
+	$(RM) $(OBJECTS)
+	$(RM) $(LIB)

crypto_kem/saber/clean/Makefile.Microsoft_nmake

@@ -0,0 +1,19 @@
+# This Makefile can be used with Microsoft Visual Studio's nmake using the command:
+#    nmake /f Makefile.Microsoft_nmake
+
+LIBRARY=libsaber_clean.lib
+OBJECTS=cbd.obj kem.obj pack_unpack.obj poly.obj poly_mul.obj SABER_indcpa.obj verify.obj 
+
+CFLAGS=/nologo /I ..\..\..\common /W4 /WX
+
+all: $(LIBRARY)
+
+# Make sure objects are recompiled if headers change.
+$(OBJECTS): *.h
+
+$(LIBRARY): $(OBJECTS)
+    LIB.EXE /NOLOGO /WX /OUT:$@ $**
+
+clean:
+    -DEL $(OBJECTS)
+    -DEL $(LIBRARY)

crypto_kem/saber/clean/SABER_indcpa.c

@@ -0,0 +1,351 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "fips202.h"
+#include "pack_unpack.h"
+#include "poly.h"
+#include "poly_mul.h"
+#include "randombytes.h"
+#include <stdint.h>
+#include <string.h>
+
+
+
+/*-----------------------------------------------------------------------------------
+    This routine generates a=[Matrix K x K] of 256-coefficient polynomials
+-------------------------------------------------------------------------------------*/
+
+#define h1 4 //2^(EQ-EP-1)
+
+#define h2 ( (1<<(SABER_EP-2)) - (1<<(SABER_EP-SABER_ET-1)) + (1<<(SABER_EQ-SABER_EP-1)) )
+
+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);
+
+static void POL2MSG(const uint16_t *message_dec_unpacked, unsigned char *message_dec);
+
+static void GenMatrix(polyvec *a, const unsigned char *seed) {
+    unsigned int one_vector = 13 * SABER_N / 8;
+    unsigned int byte_bank_length = SABER_K * SABER_K * one_vector;
+    unsigned char buf[byte_bank_length];
+
+    uint16_t temp_ar[SABER_N];
+
+    int i, j, k;
+    uint16_t mod = (SABER_Q - 1);
+
+    shake128(buf, byte_bank_length, 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)*one_vector, 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];// skpv;
+
+    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);
+    shake128(seed, SABER_SEEDBYTES, seed, SABER_SEEDBYTES); // for not revealing system RNG state
+    randombytes(noiseseed, SABER_COINBYTES);
+
+    GenMatrix(a, seed);   //sample matrix A
+
+    PQCLEAN_SABER_CLEAN_GenSecret(skpv, noiseseed); //generate secret from constant-time binomial distribution
+
+    //------------------------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++) { //shift right 3 bits
+        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));
+        }
+    }
+
+    //------------------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)-------
+
+
+    PQCLEAN_SABER_CLEAN_POLVEC2BS(pk, res, SABER_P); // load the public-key coefficients
+
+
+
+    for (i = 0; i < SABER_SEEDBYTES; i++) { // now load the seedbytes in PK. Easy since seed bytes are kept in byte format.
+        pk[SABER_POLYVECCOMPRESSEDBYTES + i] = seed[i];
+    }
+
+}
+
+
+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];     // skpv;
+    unsigned char seed[SABER_SEEDBYTES];
+    uint16_t pkcl[SABER_K][SABER_N];    //public key of received by the client
+
+
+
+    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];
+
+    for (i = 0; i < SABER_SEEDBYTES; i++) { // extract the seedbytes from Public Key.
+        seed[i] = pk[ SABER_POLYVECCOMPRESSEDBYTES + i];
+    }
+
+    GenMatrix(a, seed);
+
+    PQCLEAN_SABER_CLEAN_GenSecret(skpv1, noiseseed); //generate secret from constant-time binomial distribution
+
+    //-----------------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
+
+    for (i = 0; i < SABER_K; i++) { //shift right 3 bits
+        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++) {
+        for (j = 0; j < SABER_N; j++) {
+            skpv1[i][j] = skpv1[i][j] & (mod_p);
+        }
+    }
+
+    // vector-vector scalar multiplication with mod p
+    InnerProd(pkcl, skpv1, mod_p, vprime);
+
+    //addition of h1 to vprime
+    for (i = 0; i < SABER_N; i++) {
+        vprime[i] = vprime[i] + h1;
+    }
+
+
+    // 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);
+    }
+
+
+    #if Saber_type == 1
+    PQCLEAN_SABER_CLEAN_pack_3bit(msk_c, vprime);
+    #elif Saber_type == 2
+    PQCLEAN_SABER_CLEAN_pack_4bit(msk_c, vprime);
+    #elif Saber_type == 3
+    PQCLEAN_SABER_CLEAN_pack_6bit(msk_c, vprime);
+    #endif
+
+
+    for (j = 0; j < SABER_SCALEBYTES_KEM; j++) {
+        ciphertext[SABER_POLYVECCOMPRESSEDBYTES + j] = msk_c[j];
+    }
+}
+
+
+void PQCLEAN_SABER_CLEAN_indcpa_kem_dec(const unsigned char *sk, const unsigned char *ciphertext, unsigned char message_dec[]) {
+
+    uint32_t i, j;
+
+
+    uint16_t sksv[SABER_K][SABER_N]; //secret key of the server
+
+
+    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];
+
+
+    PQCLEAN_SABER_CLEAN_BS2POLVEC(sk, sksv, SABER_Q); //sksv is the secret-key
+    PQCLEAN_SABER_CLEAN_BS2POLVEC(ciphertext, pksv, SABER_P); //pksv is the ciphertext
+
+    // 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);
+
+
+    //Extraction
+    for (i = 0; i < SABER_SCALEBYTES_KEM; i++) {
+        scale_ar[i] = ciphertext[SABER_POLYVECCOMPRESSEDBYTES + i];
+    }
+
+    #if Saber_type == 1
+    PQCLEAN_SABER_CLEAN_un_pack3bit(scale_ar, op);
+    #elif Saber_type == 2
+    PQCLEAN_SABER_CLEAN_un_pack4bit(scale_ar, op);
+    #elif Saber_type == 3
+    PQCLEAN_SABER_CLEAN_un_pack6bit(scale_ar, op);
+    #endif
+
+
+    //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);
+    }
+
+    // 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];
+                    res[i][k] = (res[i][k] & mod); //reduction mod p
+                    acc[k] = 0; //clear the accumulator
+                }
+
+            }
+        }
+    } 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];
+                    res[i][k] = res[i][k] & mod; //reduction
+                    acc[k] = 0; //clear the accumulator
+                }
+
+            }
+        }
+    }
+
+
+}
+
+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] | (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];
+            res[k] = res[k] & mod; //reduction
+            acc[k] = 0; //clear the accumulator
+        }
+    }
+}

crypto_kem/saber/clean/SABER_indcpa.h

@@ -0,0 +1,9 @@
+#ifndef INDCPA_H
+#define INDCPA_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);
+
+#endif
+

crypto_kem/saber/clean/SABER_params.h

@@ -0,0 +1,63 @@
+#include "api.h"
+
+#ifndef PARAMS_H
+#define PARAMS_H
+
+
+
+#if Saber_type == 1
+#define SABER_K 2
+#define SABER_MU 10
+#define SABER_ET 3
+
+#elif Saber_type == 2
+#define SABER_K 3
+#define SABER_MU 8
+#define SABER_ET 4
+
+#elif Saber_type == 3
+#define SABER_K 4
+#define SABER_MU 6
+#define SABER_ET 6
+#endif
+
+#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_POLYBYTES       416 //13*256/8 
+
+#define SABER_POLYVECBYTES    (SABER_K * SABER_POLYBYTES)
+
+#define SABER_POLYVECCOMPRESSEDBYTES (SABER_K * 320) //10*256/8 NOTE : changed till here due to parameter adaptation
+
+#define SABER_CIPHERTEXTBYTES (SABER_POLYVECCOMPRESSEDBYTES)
+
+#define SABER_SCALEBYTES (SABER_DELTA*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_BYTES_CCA_DEC   (SABER_POLYVECCOMPRESSEDBYTES + SABER_SCALEBYTES_KEM) /* Second part is for Targhi-Unruh */
+
+
+
+
+#endif
+

crypto_kem/saber/clean/api.h

@@ -0,0 +1,16 @@
+#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 Saber_type 2
+
+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 */

crypto_kem/saber/clean/cbd.c

@@ -0,0 +1,106 @@
+/*---------------------------------------------------------------------
+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 <stdint.h>
+
+static uint64_t load_littleendian(const unsigned char *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_CLEAN_cbd(uint16_t *r, const unsigned char *buf) {
+    uint16_t Qmod_minus1 = SABER_Q - 1;
+
+    #if Saber_type == 3
+    uint32_t t, d, a[4], b[4];
+    int i, j;
+
+    for (i = 0; i < SABER_N / 4; i++) {
+        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);
+
+        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;
+
+    }
+    #elif Saber_type == 2
+    uint32_t t, d, a[4], b[4];
+    int i, j;
+
+    for (i = 0; i < SABER_N / 4; i++) {
+        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);
+
+        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;
+    }
+    #elif Saber_type == 1
+    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);
+
+        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;
+    }
+    #else
+#error "Unsupported SABER parameter."
+    #endif
+}

crypto_kem/saber/clean/cbd.h

@@ -0,0 +1,16 @@
+/*---------------------------------------------------------------------
+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
+----------------------------------------------------------------------*/
+#ifndef CBD_H
+#define CBD_H
+
+#include "poly.h"
+#include <stdint.h>
+
+void PQCLEAN_SABER_CLEAN_cbd(uint16_t *r, const unsigned char *buf);
+
+#endif

crypto_kem/saber/clean/kem.c

@@ -0,0 +1,78 @@
+#include "SABER_indcpa.h"
+#include "SABER_params.h"
+#include "fips202.h"
+#include "randombytes.h"
+#include "verify.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;
+
+    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[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 crypto_kem_dec() fails.
+    return (0);
+}
+
+int PQCLEAN_SABER_CLEAN_crypto_kem_enc(unsigned char *ct, unsigned char *ss, const unsigned char *pk) {
+
+    unsigned char kr[64];                                   // Will contain key, coins
+    unsigned char 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_CLEAN_indcpa_kem_enc(buf, kr + 32, pk,  ct); // buf[0:31] contains message; kr[32:63] contains randomness r;
+
+    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);
+
+    sha3_256(ss, kr, 64);                                              // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}
+
+
+int PQCLEAN_SABER_CLEAN_crypto_kem_dec(unsigned char *ss, const unsigned char *ct, const unsigned char *sk) {
+    int i, fail;
+    unsigned char cmp[SABER_BYTES_CCA_DEC];
+    unsigned char buf[64];
+    unsigned char kr[64];                             // Will contain key, coins
+    const unsigned char *pk = sk + SABER_INDCPA_SECRETKEYBYTES;
+
+    PQCLEAN_SABER_CLEAN_indcpa_kem_dec(sk, ct, buf);               // 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_CLEAN_indcpa_kem_enc(buf, kr + 32, pk, cmp);
+
+
+    fail = PQCLEAN_SABER_CLEAN_verify(ct, cmp, SABER_BYTES_CCA_DEC);
+
+    sha3_256(kr + 32, ct, SABER_BYTES_CCA_DEC);                 // overwrite coins in kr with h(c)
+
+    PQCLEAN_SABER_CLEAN_cmov(kr, sk + SABER_SECRETKEYBYTES - SABER_KEYBYTES, SABER_KEYBYTES, fail);
+
+    sha3_256(ss, kr, 64);                                       // hash concatenation of pre-k and h(c) to k
+
+    return (0);
+}

crypto_kem/saber/clean/pack_unpack.c

@@ -0,0 +1,242 @@
+#include "pack_unpack.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_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_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 );
+    }
+}
+
+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;
+    }
+}
+
+void PQCLEAN_SABER_CLEAN_pack_6bit(uint8_t *bytes, const uint16_t *data) {
+    uint32_t j;
+    uint32_t offset_data, offset_byte;
+
+    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);
+    }
+}
+
+
+void PQCLEAN_SABER_CLEAN_un_pack6bit(const unsigned char *bytes, uint16_t *data) {
+    uint32_t j;
+    uint32_t offset_data, offset_byte;
+
+    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);
+    }
+
+}
+
+
+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 );
+        }
+    }
+
+
+}
+
+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);
+
+        }
+    }
+
+
+}
+
+
+
+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 );
+
+        }
+    }
+
+
+}
+
+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_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]) { //only BS2POLq no BS2POLp
+
+    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_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_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);
+    }
+
+}

crypto_kem/saber/clean/pack_unpack.h

@@ -0,0 +1,28 @@
+#ifndef PACK_UNPACK_H
+#define PACK_UNPACK_H
+
+#include "SABER_params.h"
+#include <stdint.h>
+#include <stdio.h>
+
+
+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_pack_4bit(uint8_t *bytes, const uint16_t *data);
+
+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_un_pack6bit(const unsigned char *bytes, uint16_t *data);
+
+
+void PQCLEAN_SABER_CLEAN_BS2POL(const unsigned char *bytes, uint16_t data[SABER_N]);
+
+void PQCLEAN_SABER_CLEAN_POLVEC2BS(uint8_t *bytes, uint16_t data[SABER_K][SABER_N], uint16_t modulus);
+
+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

@@ -0,0 +1,27 @@
+/*---------------------------------------------------------------------
+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 "api.h"
+#include "cbd.h"
+#include "fips202.h"
+#include "poly.h"
+
+void PQCLEAN_SABER_CLEAN_GenSecret(uint16_t r[SABER_K][SABER_N], const unsigned char *seed) {
+
+
+    uint32_t i;
+
+    int32_t buf_size = SABER_MU * SABER_N * SABER_K / 8;
+
+    uint8_t buf[buf_size];
+
+    shake128(buf, buf_size, seed, SABER_NOISESEEDBYTES);
+
+    for (i = 0; i < SABER_K; i++) {
+        PQCLEAN_SABER_CLEAN_cbd(r[i], buf + i * SABER_MU * SABER_N / 8);
+    }
+}

crypto_kem/saber/clean/poly.h

@@ -0,0 +1,25 @@
+/*---------------------------------------------------------------------
+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
+----------------------------------------------------------------------*/
+
+#ifndef POLY_H
+#define POLY_H
+
+#include "SABER_params.h"
+#include <stdint.h>
+
+typedef struct {
+    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);
+
+#endif

crypto_kem/saber/clean/poly_mul.c

@@ -0,0 +1,240 @@
+#include "poly_mul.h"
+#include <stdint.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#define SCHB_N 16
+
+#define N_RES (SABER_N << 1)
+#define N_SB (SABER_N >> 2)
+#define N_SB_RES (2*N_SB-1)
+
+static void karatsuba_simple(const uint16_t *a_1, const uint16_t *b_1, uint16_t *result_final) { //uses 10 registers
+
+    uint16_t N = 64;
+    uint16_t d01[N / 2 - 1];
+    uint16_t d0123[N / 2 - 1];
+    uint16_t d23[N / 2 - 1];
+    uint16_t result_d01[N - 1];
+
+    int32_t i, j;
+
+    memset(result_d01, 0, (N - 1)*sizeof(uint16_t));
+    memset(d01, 0, (N / 2 - 1)*sizeof(uint16_t));
+    memset(d0123, 0, (N / 2 - 1)*sizeof(uint16_t));
+    memset(d23, 0, (N / 2 - 1)*sizeof(uint16_t));
+    memset(result_final, 0, (2 * N - 1)*sizeof(uint16_t));
+
+    uint16_t acc1, acc2, acc3, acc4, acc5, acc6, acc7, acc8, acc9, acc10;
+
+
+    for (i = 0; i < N / 4; i++) {
+        acc1 = a_1[i]; //a0
+        acc2 = a_1[i + N / 4]; //a1
+        acc3 = a_1[i + 2 * N / 4]; //a2
+        acc4 = a_1[i + 3 * N / 4]; //a3
+        for (j = 0; j < N / 4; j++) {
+
+            acc5 = b_1[j]; //b0
+            acc6 = b_1[j + N / 4]; //b1
+
+            result_final[i + j + 0 * N / 4] = result_final[i + j + 0 * N / 4] + acc1 * acc5;
+            result_final[i + j + 2 * N / 4] = result_final[i + j + 2 * N / 4] + acc2 * acc6;
+
+            acc7 = acc5 + acc6; //b01
+            acc8 = acc1 + acc2; //a01
+            d01[i + j] = d01[i + j] + acc7 * acc8;
+            //--------------------------------------------------------
+
+            acc7 = b_1[j + 2 * N / 4]; //b2
+            acc8 = b_1[j + 3 * N / 4]; //b3
+            result_final[i + j + 4 * N / 4] = result_final[i + j + 4 * N / 4] + acc7 * acc3;
+
+            result_final[i + j + 6 * N / 4] = result_final[i + j + 6 * N / 4] + acc8 * acc4;
+
+            acc9 = acc3 + acc4;
+            acc10 = acc7 + acc8;
+            d23[i + j] = d23[i + j] + acc9 * acc10;
+            //--------------------------------------------------------
+
+            acc5 = acc5 + acc7; //b02
+            acc7 = acc1 + acc3; //a02
+            result_d01[i + j + 0 * N / 4] = result_d01[i + j + 0 * N / 4] + acc5 * acc7;
+
+            acc6 = acc6 + acc8; //b13
+            acc8 = acc2 + acc4;
+            result_d01[i + j + 2 * N / 4] = result_d01[i + j + 2 * N / 4] + acc6 * acc8;
+
+            acc5 = acc5 + acc6;
+            acc7 = acc7 + acc8;
+            d0123[i + j] = d0123[i + j] + acc5 * acc7;
+        }
+    }
+
+//------------------2nd last stage-------------------------
+
+    for (i = 0; i < N / 2 - 1; i++) {
+        d0123[i] = d0123[i] - result_d01[i + 0 * N / 4] - result_d01[i + 2 * N / 4];
+        d01[i] = d01[i] - result_final[i + 0 * N / 4] - result_final[i + 2 * N / 4];
+        d23[i] = d23[i] - result_final[i + 4 * N / 4] - result_final[i + 6 * N / 4];
+    }
+
+    for (i = 0; i < N / 2 - 1; i++) {
+        result_d01[i + 1 * N / 4] = result_d01[i + 1 * N / 4] + d0123[i];
+        result_final[i + 1 * N / 4] = result_final[i + 1 * N / 4] + d01[i];
+        result_final[i + 5 * N / 4] = result_final[i + 5 * N / 4] + d23[i];
+    }
+
+//------------Last stage---------------------------
+    for (i = 0; i < N - 1; i++) {
+        result_d01[i] = result_d01[i] - result_final[i] - result_final[i + N];
+    }
+
+    for (i = 0; i < N - 1; i++) {
+        result_final[i + 1 * N / 2] = result_final[i + 1 * N / 2] + result_d01[i]; //-result_d0[i]-result_d1[i];
+    }
+
+}
+
+static void toom_cook_4way (const uint16_t *a1, const uint16_t *b1, uint16_t *result) {
+    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];
+    uint16_t bw1[N_SB], bw2[N_SB], bw3[N_SB], bw4[N_SB], bw5[N_SB], bw6[N_SB], bw7[N_SB];
+    uint16_t w1[N_SB_RES] = {0}, w2[N_SB_RES] = {0}, w3[N_SB_RES] = {0}, w4[N_SB_RES] = {0},
+                            w5[N_SB_RES] = {0}, w6[N_SB_RES] = {0}, w7[N_SB_RES] = {0};
+    uint16_t r0, r1, r2, r3, r4, r5, r6, r7;
+    uint16_t *A0, *A1, *A2, *A3, *B0, *B1, *B2, *B3;
+    A0 = (uint16_t *)a1;
+    A1 = (uint16_t *)&a1[N_SB];
+    A2 = (uint16_t *)&a1[2 * N_SB];
+    A3 = (uint16_t *)&a1[3 * N_SB];
+    B0 = (uint16_t *)b1;
+    B1 = (uint16_t *)&b1[N_SB];
+    B2 = (uint16_t *)&b1[2 * N_SB];
+    B3 = (uint16_t *)&b1[3 * N_SB];
+
+    uint16_t *C;
+    C = result;
+
+    int i, j;
+
+// EVALUATION
+    for (j = 0; j < N_SB; ++j) {
+        r0 = A0[j];
+        r1 = A1[j];
+        r2 = A2[j];
+        r3 = A3[j];
+        r4 = r0 + r2;
+        r5 = r1 + r3;
+        r6 = r4 + r5;
+        r7 = r4 - r5;
+        aw3[j] = r6;
+        aw4[j] = r7;
+        r4 = ((r0 << 2) + r2) << 1;
+        r5 = (r1 << 2) + r3;
+        r6 = r4 + r5;
+        r7 = r4 - r5;
+        aw5[j] = r6;
+        aw6[j] = r7;
+        r4 = (r3 << 3) + (r2 << 2) + (r1 << 1) + r0;
+        aw2[j] = r4;
+        aw7[j] = r0;
+        aw1[j] = r3;
+    }
+    for (j = 0; j < N_SB; ++j) {
+        r0 = B0[j];
+        r1 = B1[j];
+        r2 = B2[j];
+        r3 = B3[j];
+        r4 = r0 + r2;
+        r5 = r1 + r3;
+        r6 = r4 + r5;
+        r7 = r4 - r5;
+        bw3[j] = r6;
+        bw4[j] = r7;
+        r4 = ((r0 << 2) + r2) << 1;
+        r5 = (r1 << 2) + r3;
+        r6 = r4 + r5;
+        r7 = r4 - r5;
+        bw5[j] = r6;
+        bw6[j] = r7;
+        r4 = (r3 << 3) + (r2 << 2) + (r1 << 1) + r0;
+        bw2[j] = r4;
+        bw7[j] = r0;
+        bw1[j] = r3;
+    }
+
+// 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);
+
+// INTERPOLATION
+    for (i = 0; i < N_SB_RES; ++i) {
+        r0 = w1[i];
+        r1 = w2[i];
+        r2 = w3[i];
+        r3 = w4[i];
+        r4 = w5[i];
+        r5 = w6[i];
+        r6 = w7[i];
+
+        r1 = r1 + r4;
+        r5 = r5 - r4;
+        r3 = ((r3 - r2) >> 1);
+        r4 = r4 - r0;
+        r4 = r4 - (r6 << 6);
+        r4 = (r4 << 1) + r5;
+        r2 = r2 + r3;
+        r1 = r1 - (r2 << 6) - r2;
+        r2 = r2 - r6;
+        r2 = r2 - r0;
+        r1 = r1 + 45 * r2;
+        r4 = (((r4 - (r2 << 3)) * inv3) >> 3);
+        r5 = r5 + r1;
+        r1 = (((r1 + (r3 << 4)) * inv9) >> 1);
+        r3 = -(r3 + r1);
+        r5 = (((30 * r1 - r5) * inv15) >> 2);
+        r2 = r2 - r4;
+        r1 = r1 - r5;
+
+        C[i]     += r6;
+        C[i + 64]  += r5;
+        C[i + 128] += r4;
+        C[i + 192] += r3;
+        C[i + 256] += r2;
+        C[i + 320] += r1;
+        C[i + 384] += r0;
+    }
+}
+
+void PQCLEAN_SABER_CLEAN_pol_mul(uint16_t *a, uint16_t *b, uint16_t *res, uint16_t p, uint32_t n) {
+    // Polynomial multiplication using the schoolbook method, c[x] = a[x]*b[x]
+    // SECURITY NOTE: TO BE USED FOR TESTING ONLY.
+
+    uint32_t i;
+
+    //-------------------normal multiplication-----------------
+
+    uint16_t c[512];
+
+    for (i = 0; i < 512; i++) {
+        c[i] = 0;
+    }
+
+    toom_cook_4way(a, b, c);
+
+    //---------------reduction-------
+    for (i = n; i < 2 * n; i++) {
+        res[i - n] = (c[i - n] - c[i]) & (p - 1);
+    }
+
+
+}

crypto_kem/saber/clean/poly_mul.h

@@ -0,0 +1,9 @@
+#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

crypto_kem/saber/clean/verify.c

@@ -0,0 +1,33 @@
+/*-------------------------------------------------
+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 "verify.h"
+#include <stdint.h>
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+int PQCLEAN_SABER_CLEAN_verify(const unsigned char *a, const unsigned char *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) >> 63;
+    return r;
+}
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_SABER_CLEAN_cmov(unsigned char *r, const unsigned char *x, size_t len, unsigned char b) {
+    size_t i;
+
+    b = -b;
+    for (i = 0; i < len; i++) {
+        r[i] ^= b & (x[i] ^ r[i]);
+    }
+}

crypto_kem/saber/clean/verify.h

@@ -0,0 +1,19 @@
+/*-------------------------------------------------
+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
+----------------------------------------------------*/
+#ifndef VERIFY_H
+#define VERIFY_H
+
+#include <stddef.h>
+
+/* returns 0 for equal strings, 1 for non-equal strings */
+int PQCLEAN_SABER_CLEAN_verify(const unsigned char *a, const unsigned char *b, size_t len);
+
+/* b = 1 means mov, b = 0 means don't mov*/
+void PQCLEAN_SABER_CLEAN_cmov(unsigned char *r, const unsigned char *x, size_t len, unsigned char b);
+
+#endif