kris/pqc
1
1
포크 0
코드 이슈 풀 리퀘스트 릴리즈 위키 활동

Fix UB in qTESLA

소스 검색 
Lots of shifts left of signed integers
This commit is contained in:
Thom Wiggers 2020-03-13 15:02:24 -04:00 committed by Kris Kwiatkowski
부모 507617f30f
커밋 d257525360
8개의 변경된 파일과 46개의 추가작업 그리고 46개의 파일을 삭제
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
crypto_sign/qtesla-p-I/clean/gauss.c
파일 보기

@ -23,7 +23,7 @@ void PQCLEAN_QTESLAPI_CLEAN_sample_gauss_poly(poly z, const uint8_t *seed, uint1
cSHAKE(buf, CHUNK_SIZE * CDT_COLS * sizeof(int32_t), (uint8_t *)NULL, 0, dmsp_bytes, 2, seed, CRYPTO_RANDOMBYTES);
++dmsp;
for (size_t i = 0, j = 0; i < CHUNK_SIZE * CDT_COLS; i += 1, j += 4) {
samp[i] = (int32_t)(buf[j] | (buf[j + 1] << 8) | (buf[j + 2] << 16) | (buf[j + 3] << 24));
samp[i] = (int32_t)(buf[j] | (buf[j + 1] << 8) | (buf[j + 2] << 16) | (int32_t)((uint32_t)buf[j + 3] << 24));
}
for (size_t i = 0; i < CHUNK_SIZE; i++) {
z[chunk + i] = 0;

16
crypto_sign/qtesla-p-I/clean/pack.c
파일 보기

@ -75,13 +75,13 @@ void PQCLEAN_QTESLAPI_CLEAN_decode_pk(int32_t *pk, uint8_t *seedA, const uint8_t
const uint8_t *a = pk_in;
for (i = 0, j = 0; i < PARAM_N * PARAM_K; i += 8, j += 29) {
pk[i ] = (int32_t)(( a[j ] | (a[j + 1] << 8) | (a[j + 2] << 16) | (a[j + 3] << 24) ) & mask29);
pk[i + 1] = (int32_t)(((a[j + 3] >> 5) | (a[j + 4] << 3) | (a[j + 5] << 11) | (a[j + 6] << 19) | (a[j + 7] << 27)) & mask29);
pk[i ] = (int32_t)(( a[j ] | (a[j + 1] << 8) | (a[j + 2] << 16) | (int32_t)((uint32_t)a[j + 3] << 24) ) & mask29);
pk[i + 1] = (int32_t)(((a[j + 3] >> 5) | (a[j + 4] << 3) | (a[j + 5] << 11) | (a[j + 6] << 19) | (int32_t)((uint32_t)a[j + 7] << 27)) & mask29);
pk[i + 2] = (int32_t)(((a[j + 7] >> 2) | (a[j + 8] << 6) | (a[j + 9] << 14) | (a[j + 10] << 22) ) & mask29);
pk[i + 3] = (int32_t)(((a[j + 10] >> 7) | (a[j + 11] << 1) | (a[j + 12] << 9) | (a[j + 13] << 17) | (a[j + 14] << 25)) & mask29);
pk[i + 4] = (int32_t)(((a[j + 14] >> 4) | (a[j + 15] << 4) | (a[j + 16] << 12) | (a[j + 17] << 20) | (a[j + 18] << 28)) & mask29);
pk[i + 3] = (int32_t)(((a[j + 10] >> 7) | (a[j + 11] << 1) | (a[j + 12] << 9) | (a[j + 13] << 17) | (int32_t)((uint32_t)a[j + 14] << 25)) & mask29);
pk[i + 4] = (int32_t)(((a[j + 14] >> 4) | (a[j + 15] << 4) | (a[j + 16] << 12) | (a[j + 17] << 20) | (int32_t)((uint32_t)a[j + 18] << 28)) & mask29);
pk[i + 5] = (int32_t)(((a[j + 18] >> 1) | (a[j + 19] << 7) | (a[j + 20] << 15) | (a[j + 21] << 23) ) & mask29);
pk[i + 6] = (int32_t)(((a[j + 21] >> 6) | (a[j + 22] << 2) | (a[j + 23] << 10) | (a[j + 24] << 18) | (a[j + 25] << 26)) & mask29);
pk[i + 6] = (int32_t)(((a[j + 21] >> 6) | (a[j + 22] << 2) | (a[j + 23] << 10) | (a[j + 24] << 18) | (int32_t)((uint32_t)a[j + 25] << 26)) & mask29);
pk[i + 7] = (int32_t)( (a[j + 25] >> 3) | (a[j + 26] << 5) | (a[j + 27] << 13) | (a[j + 28] << 21) );
}
@ -96,7 +96,7 @@ void PQCLEAN_QTESLAPI_CLEAN_encode_sig(uint8_t *sm, uint8_t *c, const poly z) {
for (i = 0, j = 0; i < PARAM_N; i += 2, j += 5) {
sm[j ] = (uint8_t)( z[i ] );
sm[j + 1] = (uint8_t)( z[i ] >> 8);
sm[j + 2] = (uint8_t)(((z[i ] >> 16) & 0x0F) | (z[i + 1] << 4));
sm[j + 2] = (uint8_t)(((z[i ] >> 16) & 0x0F) | (int64_t)((uint64_t)z[i + 1] << 4));
sm[j + 3] = (uint8_t)( z[i + 1] >> 4);
sm[j + 4] = (uint8_t)( z[i + 1] >> 12);
}
@ -109,8 +109,8 @@ void PQCLEAN_QTESLAPI_CLEAN_decode_sig(uint8_t *c, poly z, const uint8_t *sm) {
size_t i, j;
for (i = 0, j = 0; i < PARAM_N; i += 2, j += 5) {
z[i ] = sm[j ] | (sm[j + 1] << 8) | (((int64_t)sm[j + 2] << 60) >> 44);
z[i + 1] = (sm[j + 2] >> 4) | (sm[j + 3] << 4) | (((int64_t)sm[j + 4] << 56) >> 44);
z[i ] = sm[j ] | (sm[j + 1] << 8) | ((int64_t)((uint64_t)sm[j + 2] << 60) >> 44);
z[i + 1] = (sm[j + 2] >> 4) | (sm[j + 3] << 4) | ((int64_t)((uint64_t)sm[j + 4] << 56) >> 44);
}
memcpy(c, &sm[j], CRYPTO_C_BYTES);

24
crypto_sign/qtesla-p-I/clean/poly.c
파일 보기

@ -215,27 +215,27 @@ void PQCLEAN_QTESLAPI_CLEAN_poly_uniform(poly_k a, const uint8_t *seed) {
pos = 0;
}
val1 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
val2 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
val3 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
val4 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
if (val1 < PARAM_Q && i < PARAM_K * PARAM_N) {

2
crypto_sign/qtesla-p-I/clean/sign.c
파일 보기

@ -77,7 +77,7 @@ static int test_correctness(const poly v) {
left = val;
val = (val + (1 << (PARAM_D - 1)) - 1) >> PARAM_D;
val = left - (val << PARAM_D);
val = left - (int32_t)((uint32_t)val << PARAM_D);
// If (Abs(val) < (1<<(PARAM_D-1))-PARAM_E) then t1 = 0, else t1 = 1
t1 = (uint32_t)(~(Abs(val) - ((1 << (PARAM_D - 1)) - PARAM_E))) >> (RADIX32 - 1);

2
crypto_sign/qtesla-p-III/clean/gauss.c
파일 보기

@ -23,7 +23,7 @@ void PQCLEAN_QTESLAPIII_CLEAN_sample_gauss_poly(poly z, const uint8_t *seed, uin
cSHAKE(buf, CHUNK_SIZE * CDT_COLS * sizeof(int32_t), (uint8_t *)NULL, 0, dmsp_bytes, 2, seed, CRYPTO_RANDOMBYTES);
++dmsp;
for (size_t i = 0, j = 0; i < CHUNK_SIZE * CDT_COLS; i += 1, j += 4) {
samp[i] = (int32_t)(buf[j] | (buf[j + 1] << 8) | (buf[j + 2] << 16) | (buf[j + 3] << 24));
samp[i] = (int32_t)(buf[j] | (buf[j + 1] << 8) | (buf[j + 2] << 16) | (int32_t)((uint32_t)buf[j + 3] << 24));
}
for (size_t i = 0; i < CHUNK_SIZE; i++) {
z[chunk + i] = 0;

20
crypto_sign/qtesla-p-III/clean/pack.c
파일 보기

@ -61,9 +61,9 @@ void PQCLEAN_QTESLAPIII_CLEAN_decode_pk(int32_t *pk, uint8_t *seedA, const uint8
const uint8_t *a = pk_in;
for (i = 0, j = 0; i < PARAM_N * PARAM_K; i += 4, j += 15) {
pk[i ] = (int32_t)(( a[j ] | (a[j + 1] << 8) | (a[j + 2] << 16) | (a[j + 3] << 24) ) & mask30);
pk[i + 1] = (int32_t)(((a[j + 3] >> 6) | (a[j + 4] << 2) | (a[j + 5] << 10) | (a[j + 6] << 18) | (a[j + 7] << 26)) & mask30);
pk[i + 2] = (int32_t)(((a[j + 7] >> 4) | (a[j + 8] << 4) | (a[j + 9] << 12) | (a[j + 10] << 20) | (a[j + 11] << 28)) & mask30);
pk[i ] = (int32_t)(( a[j ] | (a[j + 1] << 8) | (a[j + 2] << 16) | (int32_t)((uint32_t)a[j + 3] << 24) ) & mask30);
pk[i + 1] = (int32_t)(((a[j + 3] >> 6) | (a[j + 4] << 2) | (a[j + 5] << 10) | (a[j + 6] << 18) | (int32_t)((uint32_t)a[j + 7] << 26)) & mask30);
pk[i + 2] = (int32_t)(((a[j + 7] >> 4) | (a[j + 8] << 4) | (a[j + 9] << 12) | (a[j + 10] << 20) | (int32_t)((uint32_t)a[j + 11] << 28)) & mask30);
pk[i + 3] = (int32_t)( (a[j + 11] >> 2) | (a[j + 12] << 6) | (a[j + 13] << 14) | (a[j + 14] << 22) );
}
@ -78,13 +78,13 @@ void PQCLEAN_QTESLAPIII_CLEAN_encode_sig(uint8_t *sm, uint8_t *c, const poly z)
for (i = 0, j = 0; i < PARAM_N; i += 4, j += 11) {
sm[j ] = (uint8_t)( z[i ] );
sm[j + 1] = (uint8_t)( z[i ] >> 8);
sm[j + 2] = (uint8_t)(((z[i ] >> 16) & 0x3F) | (z[i + 1] << 6));
sm[j + 2] = (uint8_t)(((z[i ] >> 16) & 0x3F) | ((uint64_t)z[i + 1] << 6));
sm[j + 3] = (uint8_t)( z[i + 1] >> 2);
sm[j + 4] = (uint8_t)( z[i + 1] >> 10);
sm[j + 5] = (uint8_t)(((z[i + 1] >> 18) & 0x0F) | (z[i + 2] << 4));
sm[j + 5] = (uint8_t)(((z[i + 1] >> 18) & 0x0F) | ((uint64_t)z[i + 2] << 4));
sm[j + 6] = (uint8_t)( z[i + 2] >> 4);
sm[j + 7] = (uint8_t)( z[i + 2] >> 12);
sm[j + 8] = (uint8_t)(((z[i + 2] >> 20) & 0x03) | (z[i + 3] << 2));
sm[j + 8] = (uint8_t)(((z[i + 2] >> 20) & 0x03) | ((uint64_t)z[i + 3] << 2));
sm[j + 9] = (uint8_t)( z[i + 3] >> 6);
sm[j + 10] = (uint8_t)( z[i + 3] >> 14);
}
@ -98,10 +98,10 @@ void PQCLEAN_QTESLAPIII_CLEAN_decode_sig(uint8_t *c, poly z, const uint8_t *sm)
size_t i, j;
for (i = 0, j = 0; i < PARAM_N; i += 4, j += 11) {
z[i ] = sm[j ] | (sm[j + 1] << 8) | (((int64_t)sm[j + 2] << 58) >> 42);
z[i + 1] = (sm[j + 2] >> 6) | (sm[j + 3] << 2) | (sm[j + 4] << 10) | (((int64_t)sm[j + 5] << 60) >> 42);
z[i + 2] = (sm[j + 5] >> 4) | (sm[j + 6] << 4) | (sm[j + 7] << 12) | (((int64_t)sm[j + 8] << 62) >> 42);
z[i + 3] = (sm[j + 8] >> 2) | (sm[j + 9] << 6) | (((int64_t)sm[j + 10] << 56) >> 42);
z[i ] = sm[j ] | (sm[j + 1] << 8) | ((int64_t)((uint64_t)sm[j + 2] << 58) >> 42);
z[i + 1] = (sm[j + 2] >> 6) | (sm[j + 3] << 2) | (sm[j + 4] << 10) | ((int64_t)((uint64_t)sm[j + 5] << 60) >> 42);
z[i + 2] = (sm[j + 5] >> 4) | (sm[j + 6] << 4) | (sm[j + 7] << 12) | ((int64_t)((uint64_t)sm[j + 8] << 62) >> 42);
z[i + 3] = (sm[j + 8] >> 2) | (sm[j + 9] << 6) | ((int64_t)((uint64_t)sm[j + 10] << 56) >> 42);
}
memcpy(c, &sm[j], CRYPTO_C_BYTES);

24
crypto_sign/qtesla-p-III/clean/poly.c
파일 보기

@ -206,27 +206,27 @@ void PQCLEAN_QTESLAPIII_CLEAN_poly_uniform(poly_k a, const uint8_t *seed) {
pos = 0;
}
val1 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
val2 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
val3 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
val4 = ((uint32_t)(buf[pos])
| (uint32_t)(buf[pos + 1] << 8)
| (uint32_t)(buf[pos + 2] << 16)
| (uint32_t)(buf[pos + 3] << 24))
| ((uint32_t)buf[pos + 1] << 8)
| ((uint32_t)buf[pos + 2] << 16)
| ((uint32_t)buf[pos + 3] << 24))
& mask;
pos += nbytes;
if (val1 < PARAM_Q && i < PARAM_K * PARAM_N) {

2
crypto_sign/qtesla-p-III/clean/sign.c
파일 보기

@ -77,7 +77,7 @@ static int test_correctness(const poly v) {
left = val;
val = (val + (1 << (PARAM_D - 1)) - 1) >> PARAM_D;
val = left - (val << PARAM_D);
val = left - (int32_t)((uint32_t)val << PARAM_D);
// If (Abs(val) < (1<<(PARAM_D-1))-PARAM_E) then t1 = 0, else t1 = 1
t1 = (uint32_t)(~(Abs(val) - ((1 << (PARAM_D - 1)) - PARAM_E))) >> (RADIX32 - 1);