| 1 |
/* |
| 2 |
This file is for Niederreiter decryption |
| 3 |
*/ |
| 4 |
// 20240805 djb: more mask usage |
| 5 |
// 20240503 djb: remove #ifdef KAT ... #endif |
| 6 |
// 20221230 djb: add linker lines |
| 7 |
|
| 8 |
// linker define decrypt |
| 9 |
// linker use benes bm fft fft_tr |
| 10 |
// linker use load_bits |
| 11 |
// linker use vec256_mul_asm vec256_sq vec256_inv |
| 12 |
|
| 13 |
#include "decrypt.h" |
| 14 |
|
| 15 |
#include "params.h" |
| 16 |
#include "fft_tr.h" |
| 17 |
#include "benes.h" |
| 18 |
#include "util.h" |
| 19 |
#include "fft.h" |
| 20 |
#include "bm.h" |
| 21 |
|
| 22 |
#include "crypto_int16.h" |
| 23 |
|
| 24 |
static void scaling(vec256 out[][GFBITS], vec256 inv[][GFBITS], const unsigned char *sk, vec256 *recv) |
| 25 |
{ |
| 26 |
int i, j; |
| 27 |
|
| 28 |
vec128 sk_int[ GFBITS ]; |
| 29 |
vec256 eval[32][ GFBITS ]; |
| 30 |
vec256 tmp[ GFBITS ]; |
| 31 |
|
| 32 |
// computing inverses |
| 33 |
|
| 34 |
irr_load(sk_int, sk); |
| 35 |
|
| 36 |
fft(eval, sk_int); |
| 37 |
|
| 38 |
for (i = 0; i < 32; i++) |
| 39 |
vec256_sq(eval[i], eval[i]); |
| 40 |
|
| 41 |
vec256_copy(inv[0], eval[0]); |
| 42 |
|
| 43 |
for (i = 1; i < 32; i++) |
| 44 |
vec256_mul(inv[i], inv[i-1], eval[i]); |
| 45 |
|
| 46 |
vec256_inv(tmp, inv[31]); |
| 47 |
|
| 48 |
for (i = 30; i >= 0; i--) |
| 49 |
{ |
| 50 |
vec256_mul(inv[i+1], tmp, inv[i]); |
| 51 |
vec256_mul(tmp, tmp, eval[i+1]); |
| 52 |
} |
| 53 |
|
| 54 |
vec256_copy(inv[0], tmp); |
| 55 |
|
| 56 |
// |
| 57 |
|
| 58 |
for (i = 0; i < 32; i++) |
| 59 |
for (j = 0; j < GFBITS; j++) |
| 60 |
out[i][j] = vec256_and(inv[i][j], recv[i]); |
| 61 |
} |
| 62 |
|
| 63 |
static void preprocess(vec128 *recv, const unsigned char *s) |
| 64 |
{ |
| 65 |
int i; |
| 66 |
unsigned char r[ 1024 ]; |
| 67 |
|
| 68 |
for (i = 0; i < SYND_BYTES; i++) |
| 69 |
r[i] = s[i]; |
| 70 |
|
| 71 |
for (i = SYND_BYTES; i < 1024; i++) |
| 72 |
r[i] = 0; |
| 73 |
|
| 74 |
for (i = 0; i < 64; i++) |
| 75 |
recv[i] = load16(r + i*16); |
| 76 |
} |
| 77 |
|
| 78 |
static void postprocess(unsigned char * e, vec128 * err) |
| 79 |
{ |
| 80 |
int i; |
| 81 |
unsigned char error8[ (1 << GFBITS)/8 ]; |
| 82 |
uint64_t v[2]; |
| 83 |
|
| 84 |
for (i = 0; i < 64; i++) |
| 85 |
{ |
| 86 |
v[0] = vec128_extract(err[i], 0); |
| 87 |
v[1] = vec128_extract(err[i], 1); |
| 88 |
|
| 89 |
store8(error8 + i*16 + 0, v[0]); |
| 90 |
store8(error8 + i*16 + 8, v[1]); |
| 91 |
} |
| 92 |
|
| 93 |
for (i = 0; i < SYS_N/8; i++) |
| 94 |
e[i] = error8[i]; |
| 95 |
} |
| 96 |
|
| 97 |
static void scaling_inv(vec256 out[][GFBITS], vec256 inv[][GFBITS], vec256 *recv) |
| 98 |
{ |
| 99 |
int i, j; |
| 100 |
|
| 101 |
for (i = 0; i < 32; i++) |
| 102 |
for (j = 0; j < GFBITS; j++) |
| 103 |
out[i][j] = vec256_and(inv[i][j], recv[i]); |
| 104 |
} |
| 105 |
|
| 106 |
static int weight_check(unsigned char * e, vec128 * error) |
| 107 |
{ |
| 108 |
int i; |
| 109 |
uint16_t w0 = 0; |
| 110 |
uint16_t w1 = 0; |
| 111 |
uint16_t check; |
| 112 |
|
| 113 |
for (i = 0; i < 64; i++) |
| 114 |
{ |
| 115 |
w0 += __builtin_popcountll( vec128_extract(error[i], 0) ); |
| 116 |
w0 += __builtin_popcountll( vec128_extract(error[i], 1) ); |
| 117 |
} |
| 118 |
|
| 119 |
for (i = 0; i < SYS_N/8; i++) |
| 120 |
w1 += __builtin_popcountll( e[i] ); |
| 121 |
|
| 122 |
check = (w0 ^ SYS_T) | (w1 ^ SYS_T); |
| 123 |
return -crypto_int16_zero_mask(check); |
| 124 |
} |
| 125 |
|
| 126 |
static uint64_t synd_cmp(vec256 *s0 , vec256 *s1) |
| 127 |
{ |
| 128 |
int i; |
| 129 |
vec256 diff; |
| 130 |
|
| 131 |
diff = vec256_xor(s0[0], s1[0]); |
| 132 |
|
| 133 |
for (i = 1; i < GFBITS; i++) |
| 134 |
diff = vec256_or(diff, vec256_xor(s0[i], s1[i])); |
| 135 |
|
| 136 |
return vec256_testz(diff); |
| 137 |
} |
| 138 |
|
| 139 |
static void reformat_128to256(vec256 * out, vec128 * in) |
| 140 |
{ |
| 141 |
int i; |
| 142 |
uint64_t v[4]; |
| 143 |
|
| 144 |
for (i = 0; i < 32; i++) |
| 145 |
{ |
| 146 |
v[0] = vec128_extract(in[2*i+0], 0); |
| 147 |
v[1] = vec128_extract(in[2*i+0], 1); |
| 148 |
v[2] = vec128_extract(in[2*i+1], 0); |
| 149 |
v[3] = vec128_extract(in[2*i+1], 1); |
| 150 |
|
| 151 |
out[i] = vec256_set4x(v[0], v[1], v[2], v[3]); |
| 152 |
} |
| 153 |
} |
| 154 |
|
| 155 |
static void reformat_256to128(vec128 * out, vec256 * in) |
| 156 |
{ |
| 157 |
int i; |
| 158 |
uint64_t v[4]; |
| 159 |
|
| 160 |
for (i = 0; i < 32; i++) |
| 161 |
{ |
| 162 |
v[0] = vec256_extract(in[i], 0); |
| 163 |
v[1] = vec256_extract(in[i], 1); |
| 164 |
v[2] = vec256_extract(in[i], 2); |
| 165 |
v[3] = vec256_extract(in[i], 3); |
| 166 |
|
| 167 |
out[2*i+0] = vec128_set2x(v[0], v[1]); |
| 168 |
out[2*i+1] = vec128_set2x(v[2], v[3]); |
| 169 |
} |
| 170 |
} |
| 171 |
|
| 172 |
/* Niederreiter decryption with the Berlekamp decoder */ |
| 173 |
/* intput: sk, secret key */ |
| 174 |
/* s, ciphertext (syndrome) */ |
| 175 |
/* output: e, error vector */ |
| 176 |
/* return: 0 for success; 1 for failure */ |
| 177 |
int decrypt(unsigned char *e, const unsigned char *sk, const unsigned char *s) |
| 178 |
{ |
| 179 |
int i; |
| 180 |
|
| 181 |
uint16_t check_synd; |
| 182 |
uint16_t check_weight; |
| 183 |
|
| 184 |
vec256 inv[ 32 ][ GFBITS ]; |
| 185 |
vec256 scaled[ 32 ][ GFBITS ]; |
| 186 |
vec256 eval[32][ GFBITS ]; |
| 187 |
|
| 188 |
vec128 error128[ 64 ]; |
| 189 |
vec256 error256[ 32 ]; |
| 190 |
|
| 191 |
vec256 s_priv[ GFBITS ]; |
| 192 |
vec256 s_priv_cmp[ GFBITS ]; |
| 193 |
vec128 locator[ GFBITS ]; |
| 194 |
|
| 195 |
vec128 recv128[ 64 ]; |
| 196 |
vec256 recv256[ 32 ]; |
| 197 |
vec256 allone; |
| 198 |
|
| 199 |
vec128 bits_int[25][32]; |
| 200 |
|
| 201 |
// Berlekamp decoder |
| 202 |
|
| 203 |
preprocess(recv128, s); |
| 204 |
|
| 205 |
load_bits(bits_int, sk + IRR_BYTES); |
| 206 |
benes(recv128, bits_int, 1); |
| 207 |
|
| 208 |
reformat_128to256(recv256, recv128); |
| 209 |
|
| 210 |
scaling(scaled, inv, sk, recv256); |
| 211 |
fft_tr(s_priv, scaled); |
| 212 |
bm(locator, s_priv); |
| 213 |
|
| 214 |
fft(eval, locator); |
| 215 |
|
| 216 |
// reencryption and weight check |
| 217 |
|
| 218 |
allone = vec256_set1_16b(0xFFFF); |
| 219 |
|
| 220 |
for (i = 0; i < 32; i++) |
| 221 |
{ |
| 222 |
error256[i] = vec256_or_reduce(eval[i]); |
| 223 |
error256[i] = vec256_xor(error256[i], allone); |
| 224 |
} |
| 225 |
|
| 226 |
scaling_inv(scaled, inv, error256); |
| 227 |
fft_tr(s_priv_cmp, scaled); |
| 228 |
|
| 229 |
check_synd = synd_cmp(s_priv, s_priv_cmp); |
| 230 |
|
| 231 |
// |
| 232 |
|
| 233 |
reformat_256to128(error128, error256); |
| 234 |
benes(error128, bits_int, 0); |
| 235 |
|
| 236 |
postprocess(e, error128); |
| 237 |
|
| 238 |
check_weight = weight_check(e, error128); |
| 239 |
|
| 240 |
return 1 - (check_synd & check_weight); |
| 241 |
} |
| 242 |
|