diff options
Diffstat (limited to 'src/lib')
-rw-r--r-- | src/lib/libcrypto/Makefile | 3 | ||||
-rw-r--r-- | src/lib/libcrypto/hidden/openssl/mlkem.h | 10 | ||||
-rw-r--r-- | src/lib/libcrypto/mlkem/mlkem.h | 119 | ||||
-rw-r--r-- | src/lib/libcrypto/mlkem/mlkem1024.c | 1121 | ||||
-rw-r--r-- | src/lib/libcrypto/mlkem/mlkem_internal.h | 39 |
5 files changed, 1290 insertions, 2 deletions
diff --git a/src/lib/libcrypto/Makefile b/src/lib/libcrypto/Makefile index f43b09d176..ab2349103d 100644 --- a/src/lib/libcrypto/Makefile +++ b/src/lib/libcrypto/Makefile | |||
@@ -1,4 +1,4 @@ | |||
1 | # $OpenBSD: Makefile,v 1.229 2024/12/13 00:03:57 beck Exp $ | 1 | # $OpenBSD: Makefile,v 1.230 2024/12/13 00:17:17 beck Exp $ |
2 | 2 | ||
3 | LIB= crypto | 3 | LIB= crypto |
4 | LIBREBUILD=y | 4 | LIBREBUILD=y |
@@ -374,6 +374,7 @@ SRCS+= md5.c | |||
374 | 374 | ||
375 | # mlkem/ | 375 | # mlkem/ |
376 | SRCS+= mlkem768.c | 376 | SRCS+= mlkem768.c |
377 | SRCS+= mlkem1024.c | ||
377 | 378 | ||
378 | # modes/ | 379 | # modes/ |
379 | SRCS+= cbc128.c | 380 | SRCS+= cbc128.c |
diff --git a/src/lib/libcrypto/hidden/openssl/mlkem.h b/src/lib/libcrypto/hidden/openssl/mlkem.h index 01ac28cffd..103144d1a1 100644 --- a/src/lib/libcrypto/hidden/openssl/mlkem.h +++ b/src/lib/libcrypto/hidden/openssl/mlkem.h | |||
@@ -1,4 +1,4 @@ | |||
1 | /* $OpenBSD: mlkem.h,v 1.1 2024/12/13 00:03:57 beck Exp $ */ | 1 | /* $OpenBSD: mlkem.h,v 1.2 2024/12/13 00:17:17 beck Exp $ */ |
2 | /* | 2 | /* |
3 | * Copyright (c) 2024 Bob Beck <beck@obtuse.com> | 3 | * Copyright (c) 2024 Bob Beck <beck@obtuse.com> |
4 | * | 4 | * |
@@ -35,6 +35,14 @@ LCRYPTO_USED(MLKEM768_marshal_public_key); | |||
35 | LCRYPTO_USED(MLKEM768_parse_public_key); | 35 | LCRYPTO_USED(MLKEM768_parse_public_key); |
36 | LCRYPTO_USED(MLKEM768_private_key_from_seed); | 36 | LCRYPTO_USED(MLKEM768_private_key_from_seed); |
37 | LCRYPTO_USED(MLKEM768_parse_private_key); | 37 | LCRYPTO_USED(MLKEM768_parse_private_key); |
38 | LCRYPTO_USED(MLKEM1024_generate_key); | ||
39 | LCRYPTO_USED(MLKEM1024_public_from_private); | ||
40 | LCRYPTO_USED(MLKEM1024_encap); | ||
41 | LCRYPTO_USED(MLKEM1024_decap); | ||
42 | LCRYPTO_USED(MLKEM1024_marshal_public_key); | ||
43 | LCRYPTO_USED(MLKEM1024_parse_public_key); | ||
44 | LCRYPTO_USED(MLKEM1024_private_key_from_seed); | ||
45 | LCRYPTO_USED(MLKEM1024_parse_private_key); | ||
38 | #endif | 46 | #endif |
39 | 47 | ||
40 | #endif /* _LIBCRYPTO_MLKEM_H */ | 48 | #endif /* _LIBCRYPTO_MLKEM_H */ |
diff --git a/src/lib/libcrypto/mlkem/mlkem.h b/src/lib/libcrypto/mlkem/mlkem.h index 8040f4844b..1033b89a60 100644 --- a/src/lib/libcrypto/mlkem/mlkem.h +++ b/src/lib/libcrypto/mlkem/mlkem.h | |||
@@ -161,6 +161,125 @@ int MLKEM768_parse_public_key(struct MLKEM768_public_key *out_public_key, | |||
161 | int MLKEM768_parse_private_key(struct MLKEM768_private_key *out_private_key, | 161 | int MLKEM768_parse_private_key(struct MLKEM768_private_key *out_private_key, |
162 | struct cbs_st *in); | 162 | struct cbs_st *in); |
163 | 163 | ||
164 | /* | ||
165 | * ML-KEM-1024 | ||
166 | * | ||
167 | * ML-KEM-1024 also exists. You should prefer ML-KEM-768 where possible. | ||
168 | */ | ||
169 | |||
170 | /* | ||
171 | * MLKEM1024_public_key contains an ML-KEM-1024 public key. The contents of this | ||
172 | * object should never leave the address space since the format is unstable. | ||
173 | */ | ||
174 | struct MLKEM1024_public_key { | ||
175 | union { | ||
176 | uint8_t bytes[512 * (4 + 16) + 32 + 32]; | ||
177 | uint16_t alignment; | ||
178 | } opaque; | ||
179 | }; | ||
180 | |||
181 | /* | ||
182 | * MLKEM1024_private_key contains a ML-KEM-1024 private key. The contents of | ||
183 | * this object should never leave the address space since the format is | ||
184 | * unstable. | ||
185 | */ | ||
186 | struct MLKEM1024_private_key { | ||
187 | union { | ||
188 | uint8_t bytes[512 * (4 + 4 + 16) + 32 + 32 + 32]; | ||
189 | uint16_t alignment; | ||
190 | } opaque; | ||
191 | }; | ||
192 | |||
193 | /* | ||
194 | * MLKEM1024_PUBLIC_KEY_BYTES is the number of bytes in an encoded ML-KEM-1024 | ||
195 | * public key. | ||
196 | */ | ||
197 | #define MLKEM1024_PUBLIC_KEY_BYTES 1568 | ||
198 | |||
199 | /* | ||
200 | * MLKEM1024_generate_key generates a random public/private key pair, writes the | ||
201 | * encoded public key to |out_encoded_public_key| and sets |out_private_key| to | ||
202 | * the private key. If |optional_out_seed| is not NULL then the seed used to | ||
203 | * generate the private key is written to it. | ||
204 | */ | ||
205 | void MLKEM1024_generate_key( | ||
206 | uint8_t out_encoded_public_key[MLKEM1024_PUBLIC_KEY_BYTES], | ||
207 | uint8_t optional_out_seed[MLKEM_SEED_BYTES], | ||
208 | struct MLKEM1024_private_key *out_private_key); | ||
209 | |||
210 | /* | ||
211 | * MLKEM1024_private_key_from_seed derives a private key from a seed that was | ||
212 | * generated by |MLKEM1024_generate_key|. It fails and returns 0 if |seed_len| | ||
213 | * is incorrect, otherwise it writes |*out_private_key| and returns 1. | ||
214 | */ | ||
215 | int MLKEM1024_private_key_from_seed( | ||
216 | struct MLKEM1024_private_key *out_private_key, const uint8_t *seed, | ||
217 | size_t seed_len); | ||
218 | |||
219 | /* | ||
220 | * MLKEM1024_public_from_private sets |*out_public_key| to the public key that | ||
221 | * corresponds to |private_key|. (This is faster than parsing the output of | ||
222 | * |MLKEM1024_generate_key| if, for some reason, you need to encapsulate to a | ||
223 | * key that was just generated.) | ||
224 | */ | ||
225 | void MLKEM1024_public_from_private(struct MLKEM1024_public_key *out_public_key, | ||
226 | const struct MLKEM1024_private_key *private_key); | ||
227 | |||
228 | /* MLKEM1024_CIPHERTEXT_BYTES is number of bytes in the ML-KEM-1024 ciphertext. */ | ||
229 | #define MLKEM1024_CIPHERTEXT_BYTES 1568 | ||
230 | |||
231 | /* | ||
232 | * MLKEM1024_encap encrypts a random shared secret for |public_key|, writes the | ||
233 | * ciphertext to |out_ciphertext|, and writes the random shared secret to | ||
234 | * |out_shared_secret|. | ||
235 | */ | ||
236 | void MLKEM1024_encap(uint8_t out_ciphertext[MLKEM1024_CIPHERTEXT_BYTES], | ||
237 | uint8_t out_shared_secret[MLKEM_SHARED_SECRET_BYTES], | ||
238 | const struct MLKEM1024_public_key *public_key); | ||
239 | |||
240 | /* | ||
241 | * MLKEM1024_decap decrypts a shared secret from |ciphertext| using | ||
242 | * |private_key| and writes it to |out_shared_secret|. If |ciphertext_len| is | ||
243 | * incorrect it returns 0, otherwise it returns 1. If |ciphertext| is invalid | ||
244 | * (but of the correct length), |out_shared_secret| is filled with a key that | ||
245 | * will always be the same for the same |ciphertext| and |private_key|, but | ||
246 | * which appears to be random unless one has access to |private_key|. These | ||
247 | * alternatives occur in constant time. Any subsequent symmetric encryption | ||
248 | * using |out_shared_secret| must use an authenticated encryption scheme in | ||
249 | * order to discover the decapsulation failure. | ||
250 | */ | ||
251 | int MLKEM1024_decap(uint8_t out_shared_secret[MLKEM_SHARED_SECRET_BYTES], | ||
252 | const uint8_t *ciphertext, size_t ciphertext_len, | ||
253 | const struct MLKEM1024_private_key *private_key); | ||
254 | |||
255 | /* | ||
256 | * Serialisation of ML-KEM-1024 keys. | ||
257 | * MLKEM1024_marshal_public_key serializes |public_key| to |out| in the standard | ||
258 | * format for ML-KEM-1024 public keys. It returns one on success or zero on | ||
259 | * allocation error. | ||
260 | */ | ||
261 | int MLKEM1024_marshal_public_key(struct cbb_st *out, | ||
262 | const struct MLKEM1024_public_key *public_key); | ||
263 | |||
264 | /* | ||
265 | * MLKEM1024_parse_public_key parses a public key, in the format generated by | ||
266 | * |MLKEM1024_marshal_public_key|, from |in| and writes the result to | ||
267 | * |out_public_key|. It returns one on success or zero on parse error or if | ||
268 | * there are trailing bytes in |in|. | ||
269 | */ | ||
270 | int MLKEM1024_parse_public_key(struct MLKEM1024_public_key *out_public_key, | ||
271 | struct cbs_st *in); | ||
272 | |||
273 | /* | ||
274 | * MLKEM1024_parse_private_key parses a private key, in NIST's format for | ||
275 | * private keys, from |in| and writes the result to |out_private_key|. It | ||
276 | * returns one on success or zero on parse error or if there are trailing bytes | ||
277 | * in |in|. This format is verbose and should be avoided. Private keys should be | ||
278 | * stored as seeds and parsed using |MLKEM1024_private_key_from_seed|. | ||
279 | */ | ||
280 | int MLKEM1024_parse_private_key(struct MLKEM1024_private_key *out_private_key, | ||
281 | struct cbs_st *in); | ||
282 | |||
164 | #if defined(__cplusplus) | 283 | #if defined(__cplusplus) |
165 | } | 284 | } |
166 | #endif | 285 | #endif |
diff --git a/src/lib/libcrypto/mlkem/mlkem1024.c b/src/lib/libcrypto/mlkem/mlkem1024.c new file mode 100644 index 0000000000..e0a71f335b --- /dev/null +++ b/src/lib/libcrypto/mlkem/mlkem1024.c | |||
@@ -0,0 +1,1121 @@ | |||
1 | /* $OpenBSD: mlkem1024.c,v 1.1 2024/12/13 00:17:17 beck Exp $ */ | ||
2 | /* | ||
3 | * Copyright (c) 2024, Google Inc. | ||
4 | * Copyright (c) 2024, Bob Beck <beck@obtuse.com> | ||
5 | * | ||
6 | * Permission to use, copy, modify, and/or distribute this software for any | ||
7 | * purpose with or without fee is hereby granted, provided that the above | ||
8 | * copyright notice and this permission notice appear in all copies. | ||
9 | * | ||
10 | * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES | ||
11 | * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF | ||
12 | * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY | ||
13 | * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES | ||
14 | * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION | ||
15 | * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN | ||
16 | * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. | ||
17 | */ | ||
18 | |||
19 | #include <openssl/mlkem.h> | ||
20 | |||
21 | #include <assert.h> | ||
22 | #include <stdlib.h> | ||
23 | #include <string.h> | ||
24 | |||
25 | #include "bytestring.h" | ||
26 | |||
27 | #include "sha3_internal.h" | ||
28 | #include "mlkem_internal.h" | ||
29 | #include "constant_time.h" | ||
30 | #include "crypto_internal.h" | ||
31 | |||
32 | /* Remove later */ | ||
33 | #undef LCRYPTO_ALIAS | ||
34 | #define LCRYPTO_ALIAS(A) | ||
35 | |||
36 | /* | ||
37 | * See | ||
38 | * https://csrc.nist.gov/pubs/fips/203/final | ||
39 | */ | ||
40 | |||
41 | static void | ||
42 | prf(uint8_t *out, size_t out_len, const uint8_t in[33]) | ||
43 | { | ||
44 | sha3_ctx ctx; | ||
45 | shake256_init(&ctx); | ||
46 | shake_update(&ctx, in, 33); | ||
47 | shake_xof(&ctx); | ||
48 | shake_out(&ctx, out, out_len); | ||
49 | } | ||
50 | |||
51 | /* Section 4.1 */ | ||
52 | static void | ||
53 | hash_h(uint8_t out[32], const uint8_t *in, size_t len) | ||
54 | { | ||
55 | sha3_ctx ctx; | ||
56 | sha3_init(&ctx, 32); | ||
57 | sha3_update(&ctx, in, len); | ||
58 | sha3_final(out, &ctx); | ||
59 | } | ||
60 | |||
61 | static void | ||
62 | hash_g(uint8_t out[64], const uint8_t *in, size_t len) | ||
63 | { | ||
64 | sha3_ctx ctx; | ||
65 | sha3_init(&ctx, 64); | ||
66 | sha3_update(&ctx, in, len); | ||
67 | sha3_final(out, &ctx); | ||
68 | } | ||
69 | |||
70 | /* this is called 'J' in the spec */ | ||
71 | static void | ||
72 | kdf(uint8_t out[MLKEM_SHARED_SECRET_BYTES], const uint8_t failure_secret[32], | ||
73 | const uint8_t *in, size_t len) | ||
74 | { | ||
75 | sha3_ctx ctx; | ||
76 | shake256_init(&ctx); | ||
77 | shake_update(&ctx, failure_secret, 32); | ||
78 | shake_update(&ctx, in, len); | ||
79 | shake_xof(&ctx); | ||
80 | shake_out(&ctx, out, MLKEM_SHARED_SECRET_BYTES); | ||
81 | } | ||
82 | |||
83 | #define DEGREE 256 | ||
84 | #define RANK1024 4 | ||
85 | |||
86 | static const size_t kBarrettMultiplier = 5039; | ||
87 | static const unsigned kBarrettShift = 24; | ||
88 | static const uint16_t kPrime = 3329; | ||
89 | static const int kLog2Prime = 12; | ||
90 | static const uint16_t kHalfPrime = (/*kPrime=*/3329 - 1) / 2; | ||
91 | static const int kDU1024 = 11; | ||
92 | static const int kDV1024 = 5; | ||
93 | |||
94 | /* | ||
95 | * kInverseDegree is 128^-1 mod 3329; 128 because kPrime does not have a 512th | ||
96 | * root of unity. | ||
97 | */ | ||
98 | static const uint16_t kInverseDegree = 3303; | ||
99 | static const size_t kEncodedVectorSize = | ||
100 | (/*kLog2Prime=*/12 * DEGREE / 8) * RANK1024; | ||
101 | static const size_t kCompressedVectorSize = /*kDU1024=*/ 11 * RANK1024 * DEGREE / | ||
102 | 8; | ||
103 | |||
104 | typedef struct scalar { | ||
105 | /* On every function entry and exit, 0 <= c < kPrime. */ | ||
106 | uint16_t c[DEGREE]; | ||
107 | } scalar; | ||
108 | |||
109 | typedef struct vector { | ||
110 | scalar v[RANK1024]; | ||
111 | } vector; | ||
112 | |||
113 | typedef struct matrix { | ||
114 | scalar v[RANK1024][RANK1024]; | ||
115 | } matrix; | ||
116 | |||
117 | /* | ||
118 | * This bit of Python will be referenced in some of the following comments: | ||
119 | * | ||
120 | * p = 3329 | ||
121 | * | ||
122 | * def bitreverse(i): | ||
123 | * ret = 0 | ||
124 | * for n in range(7): | ||
125 | * bit = i & 1 | ||
126 | * ret <<= 1 | ||
127 | * ret |= bit | ||
128 | * i >>= 1 | ||
129 | * return ret | ||
130 | */ | ||
131 | |||
132 | /* kNTTRoots = [pow(17, bitreverse(i), p) for i in range(128)] */ | ||
133 | static const uint16_t kNTTRoots[128] = { | ||
134 | 1, 1729, 2580, 3289, 2642, 630, 1897, 848, 1062, 1919, 193, 797, | ||
135 | 2786, 3260, 569, 1746, 296, 2447, 1339, 1476, 3046, 56, 2240, 1333, | ||
136 | 1426, 2094, 535, 2882, 2393, 2879, 1974, 821, 289, 331, 3253, 1756, | ||
137 | 1197, 2304, 2277, 2055, 650, 1977, 2513, 632, 2865, 33, 1320, 1915, | ||
138 | 2319, 1435, 807, 452, 1438, 2868, 1534, 2402, 2647, 2617, 1481, 648, | ||
139 | 2474, 3110, 1227, 910, 17, 2761, 583, 2649, 1637, 723, 2288, 1100, | ||
140 | 1409, 2662, 3281, 233, 756, 2156, 3015, 3050, 1703, 1651, 2789, 1789, | ||
141 | 1847, 952, 1461, 2687, 939, 2308, 2437, 2388, 733, 2337, 268, 641, | ||
142 | 1584, 2298, 2037, 3220, 375, 2549, 2090, 1645, 1063, 319, 2773, 757, | ||
143 | 2099, 561, 2466, 2594, 2804, 1092, 403, 1026, 1143, 2150, 2775, 886, | ||
144 | 1722, 1212, 1874, 1029, 2110, 2935, 885, 2154, | ||
145 | }; | ||
146 | |||
147 | /* kInverseNTTRoots = [pow(17, -bitreverse(i), p) for i in range(128)] */ | ||
148 | static const uint16_t kInverseNTTRoots[128] = { | ||
149 | 1, 1600, 40, 749, 2481, 1432, 2699, 687, 1583, 2760, 69, 543, | ||
150 | 2532, 3136, 1410, 2267, 2508, 1355, 450, 936, 447, 2794, 1235, 1903, | ||
151 | 1996, 1089, 3273, 283, 1853, 1990, 882, 3033, 2419, 2102, 219, 855, | ||
152 | 2681, 1848, 712, 682, 927, 1795, 461, 1891, 2877, 2522, 1894, 1010, | ||
153 | 1414, 2009, 3296, 464, 2697, 816, 1352, 2679, 1274, 1052, 1025, 2132, | ||
154 | 1573, 76, 2998, 3040, 1175, 2444, 394, 1219, 2300, 1455, 2117, 1607, | ||
155 | 2443, 554, 1179, 2186, 2303, 2926, 2237, 525, 735, 863, 2768, 1230, | ||
156 | 2572, 556, 3010, 2266, 1684, 1239, 780, 2954, 109, 1292, 1031, 1745, | ||
157 | 2688, 3061, 992, 2596, 941, 892, 1021, 2390, 642, 1868, 2377, 1482, | ||
158 | 1540, 540, 1678, 1626, 279, 314, 1173, 2573, 3096, 48, 667, 1920, | ||
159 | 2229, 1041, 2606, 1692, 680, 2746, 568, 3312, | ||
160 | }; | ||
161 | |||
162 | /* kModRoots = [pow(17, 2*bitreverse(i) + 1, p) for i in range(128)] */ | ||
163 | static const uint16_t kModRoots[128] = { | ||
164 | 17, 3312, 2761, 568, 583, 2746, 2649, 680, 1637, 1692, 723, 2606, | ||
165 | 2288, 1041, 1100, 2229, 1409, 1920, 2662, 667, 3281, 48, 233, 3096, | ||
166 | 756, 2573, 2156, 1173, 3015, 314, 3050, 279, 1703, 1626, 1651, 1678, | ||
167 | 2789, 540, 1789, 1540, 1847, 1482, 952, 2377, 1461, 1868, 2687, 642, | ||
168 | 939, 2390, 2308, 1021, 2437, 892, 2388, 941, 733, 2596, 2337, 992, | ||
169 | 268, 3061, 641, 2688, 1584, 1745, 2298, 1031, 2037, 1292, 3220, 109, | ||
170 | 375, 2954, 2549, 780, 2090, 1239, 1645, 1684, 1063, 2266, 319, 3010, | ||
171 | 2773, 556, 757, 2572, 2099, 1230, 561, 2768, 2466, 863, 2594, 735, | ||
172 | 2804, 525, 1092, 2237, 403, 2926, 1026, 2303, 1143, 2186, 2150, 1179, | ||
173 | 2775, 554, 886, 2443, 1722, 1607, 1212, 2117, 1874, 1455, 1029, 2300, | ||
174 | 2110, 1219, 2935, 394, 885, 2444, 2154, 1175, | ||
175 | }; | ||
176 | |||
177 | /* reduce_once reduces 0 <= x < 2*kPrime, mod kPrime. */ | ||
178 | static uint16_t | ||
179 | reduce_once(uint16_t x) | ||
180 | { | ||
181 | assert(x < 2 * kPrime); | ||
182 | const uint16_t subtracted = x - kPrime; | ||
183 | uint16_t mask = 0u - (subtracted >> 15); | ||
184 | /* | ||
185 | * On Aarch64, omitting a |value_barrier_u16| results in a 2x speedup of | ||
186 | * ML-KEM overall and Clang still produces constant-time code using | ||
187 | * `csel`. On other platforms & compilers on godbolt that we care about, | ||
188 | * this code also produces constant-time output. | ||
189 | */ | ||
190 | return (mask & x) | (~mask & subtracted); | ||
191 | } | ||
192 | |||
193 | /* | ||
194 | * constant time reduce x mod kPrime using Barrett reduction. x must be less | ||
195 | * than kPrime + 2×kPrime². | ||
196 | */ | ||
197 | static uint16_t | ||
198 | reduce(uint32_t x) | ||
199 | { | ||
200 | uint64_t product = (uint64_t)x * kBarrettMultiplier; | ||
201 | uint32_t quotient = (uint32_t)(product >> kBarrettShift); | ||
202 | uint32_t remainder = x - quotient * kPrime; | ||
203 | |||
204 | assert(x < kPrime + 2u * kPrime * kPrime); | ||
205 | return reduce_once(remainder); | ||
206 | } | ||
207 | |||
208 | static void | ||
209 | scalar_zero(scalar *out) | ||
210 | { | ||
211 | memset(out, 0, sizeof(*out)); | ||
212 | } | ||
213 | |||
214 | static void | ||
215 | vector_zero(vector *out) | ||
216 | { | ||
217 | memset(out, 0, sizeof(*out)); | ||
218 | } | ||
219 | |||
220 | /* | ||
221 | * In place number theoretic transform of a given scalar. | ||
222 | * Note that MLKEM's kPrime 3329 does not have a 512th root of unity, so this | ||
223 | * transform leaves off the last iteration of the usual FFT code, with the 128 | ||
224 | * relevant roots of unity being stored in |kNTTRoots|. This means the output | ||
225 | * should be seen as 128 elements in GF(3329^2), with the coefficients of the | ||
226 | * elements being consecutive entries in |s->c|. | ||
227 | */ | ||
228 | static void | ||
229 | scalar_ntt(scalar *s) | ||
230 | { | ||
231 | int offset = DEGREE; | ||
232 | int step; | ||
233 | /* | ||
234 | * `int` is used here because using `size_t` throughout caused a ~5% slowdown | ||
235 | * with Clang 14 on Aarch64. | ||
236 | */ | ||
237 | for (step = 1; step < DEGREE / 2; step <<= 1) { | ||
238 | int i, j, k = 0; | ||
239 | |||
240 | offset >>= 1; | ||
241 | for (i = 0; i < step; i++) { | ||
242 | const uint32_t step_root = kNTTRoots[i + step]; | ||
243 | |||
244 | for (j = k; j < k + offset; j++) { | ||
245 | uint16_t odd, even; | ||
246 | |||
247 | odd = reduce(step_root * s->c[j + offset]); | ||
248 | even = s->c[j]; | ||
249 | s->c[j] = reduce_once(odd + even); | ||
250 | s->c[j + offset] = reduce_once(even - odd + | ||
251 | kPrime); | ||
252 | } | ||
253 | k += 2 * offset; | ||
254 | } | ||
255 | } | ||
256 | } | ||
257 | |||
258 | static void | ||
259 | vector_ntt(vector *a) | ||
260 | { | ||
261 | int i; | ||
262 | |||
263 | for (i = 0; i < RANK1024; i++) { | ||
264 | scalar_ntt(&a->v[i]); | ||
265 | } | ||
266 | } | ||
267 | |||
268 | /* | ||
269 | * In place inverse number theoretic transform of a given scalar, with pairs of | ||
270 | * entries of s->v being interpreted as elements of GF(3329^2). Just as with the | ||
271 | * number theoretic transform, this leaves off the first step of the normal iFFT | ||
272 | * to account for the fact that 3329 does not have a 512th root of unity, using | ||
273 | * the precomputed 128 roots of unity stored in |kInverseNTTRoots|. | ||
274 | */ | ||
275 | static void | ||
276 | scalar_inverse_ntt(scalar *s) | ||
277 | { | ||
278 | int i, j, k, offset, step = DEGREE / 2; | ||
279 | |||
280 | /* | ||
281 | * `int` is used here because using `size_t` throughout caused a ~5% slowdown | ||
282 | * with Clang 14 on Aarch64. | ||
283 | */ | ||
284 | for (offset = 2; offset < DEGREE; offset <<= 1) { | ||
285 | step >>= 1; | ||
286 | k = 0; | ||
287 | for (i = 0; i < step; i++) { | ||
288 | uint32_t step_root = kInverseNTTRoots[i + step]; | ||
289 | for (j = k; j < k + offset; j++) { | ||
290 | uint16_t odd, even; | ||
291 | odd = s->c[j + offset]; | ||
292 | even = s->c[j]; | ||
293 | s->c[j] = reduce_once(odd + even); | ||
294 | s->c[j + offset] = reduce(step_root * | ||
295 | (even - odd + kPrime)); | ||
296 | } | ||
297 | k += 2 * offset; | ||
298 | } | ||
299 | } | ||
300 | for (i = 0; i < DEGREE; i++) { | ||
301 | s->c[i] = reduce(s->c[i] * kInverseDegree); | ||
302 | } | ||
303 | } | ||
304 | |||
305 | static void | ||
306 | vector_inverse_ntt(vector *a) | ||
307 | { | ||
308 | int i; | ||
309 | |||
310 | for (i = 0; i < RANK1024; i++) { | ||
311 | scalar_inverse_ntt(&a->v[i]); | ||
312 | } | ||
313 | } | ||
314 | |||
315 | static void | ||
316 | scalar_add(scalar *lhs, const scalar *rhs) | ||
317 | { | ||
318 | int i; | ||
319 | |||
320 | for (i = 0; i < DEGREE; i++) { | ||
321 | lhs->c[i] = reduce_once(lhs->c[i] + rhs->c[i]); | ||
322 | } | ||
323 | } | ||
324 | |||
325 | static void | ||
326 | scalar_sub(scalar *lhs, const scalar *rhs) | ||
327 | { | ||
328 | int i; | ||
329 | |||
330 | for (i = 0; i < DEGREE; i++) { | ||
331 | lhs->c[i] = reduce_once(lhs->c[i] - rhs->c[i] + kPrime); | ||
332 | } | ||
333 | } | ||
334 | |||
335 | /* | ||
336 | * Multiplying two scalars in the number theoretically transformed state. Since | ||
337 | * 3329 does not have a 512th root of unity, this means we have to interpret | ||
338 | * the 2*ith and (2*i+1)th entries of the scalar as elements of GF(3329)[X]/(X^2 | ||
339 | * - 17^(2*bitreverse(i)+1)) The value of 17^(2*bitreverse(i)+1) mod 3329 is | ||
340 | * stored in the precomputed |kModRoots| table. Note that our Barrett transform | ||
341 | * only allows us to multipy two reduced numbers together, so we need some | ||
342 | * intermediate reduction steps, even if an uint64_t could hold 3 multiplied | ||
343 | * numbers. | ||
344 | */ | ||
345 | static void | ||
346 | scalar_mult(scalar *out, const scalar *lhs, const scalar *rhs) | ||
347 | { | ||
348 | int i; | ||
349 | |||
350 | for (i = 0; i < DEGREE / 2; i++) { | ||
351 | uint32_t real_real = (uint32_t)lhs->c[2 * i] * rhs->c[2 * i]; | ||
352 | uint32_t img_img = (uint32_t)lhs->c[2 * i + 1] * | ||
353 | rhs->c[2 * i + 1]; | ||
354 | uint32_t real_img = (uint32_t)lhs->c[2 * i] * rhs->c[2 * i + 1]; | ||
355 | uint32_t img_real = (uint32_t)lhs->c[2 * i + 1] * rhs->c[2 * i]; | ||
356 | |||
357 | out->c[2 * i] = | ||
358 | reduce(real_real + | ||
359 | (uint32_t)reduce(img_img) * kModRoots[i]); | ||
360 | out->c[2 * i + 1] = reduce(img_real + real_img); | ||
361 | } | ||
362 | } | ||
363 | |||
364 | static void | ||
365 | vector_add(vector *lhs, const vector *rhs) | ||
366 | { | ||
367 | int i; | ||
368 | |||
369 | for (i = 0; i < RANK1024; i++) { | ||
370 | scalar_add(&lhs->v[i], &rhs->v[i]); | ||
371 | } | ||
372 | } | ||
373 | |||
374 | static void | ||
375 | matrix_mult(vector *out, const matrix *m, const vector *a) | ||
376 | { | ||
377 | int i, j; | ||
378 | |||
379 | vector_zero(out); | ||
380 | for (i = 0; i < RANK1024; i++) { | ||
381 | for (j = 0; j < RANK1024; j++) { | ||
382 | scalar product; | ||
383 | |||
384 | scalar_mult(&product, &m->v[i][j], &a->v[j]); | ||
385 | scalar_add(&out->v[i], &product); | ||
386 | } | ||
387 | } | ||
388 | } | ||
389 | |||
390 | static void | ||
391 | matrix_mult_transpose(vector *out, const matrix *m, | ||
392 | const vector *a) | ||
393 | { | ||
394 | int i, j; | ||
395 | |||
396 | vector_zero(out); | ||
397 | for (i = 0; i < RANK1024; i++) { | ||
398 | for (j = 0; j < RANK1024; j++) { | ||
399 | scalar product; | ||
400 | |||
401 | scalar_mult(&product, &m->v[j][i], &a->v[j]); | ||
402 | scalar_add(&out->v[i], &product); | ||
403 | } | ||
404 | } | ||
405 | } | ||
406 | |||
407 | static void | ||
408 | scalar_inner_product(scalar *out, const vector *lhs, | ||
409 | const vector *rhs) | ||
410 | { | ||
411 | int i; | ||
412 | scalar_zero(out); | ||
413 | for (i = 0; i < RANK1024; i++) { | ||
414 | scalar product; | ||
415 | |||
416 | scalar_mult(&product, &lhs->v[i], &rhs->v[i]); | ||
417 | scalar_add(out, &product); | ||
418 | } | ||
419 | } | ||
420 | |||
421 | /* | ||
422 | * Algorithm 6 of spec. Rejection samples a Keccak stream to get uniformly | ||
423 | * distributed elements. This is used for matrix expansion and only operates on | ||
424 | * public inputs. | ||
425 | */ | ||
426 | static void | ||
427 | scalar_from_keccak_vartime(scalar *out, sha3_ctx *keccak_ctx) | ||
428 | { | ||
429 | int i, done = 0; | ||
430 | |||
431 | while (done < DEGREE) { | ||
432 | uint8_t block[168]; | ||
433 | |||
434 | shake_out(keccak_ctx, block, sizeof(block)); | ||
435 | for (i = 0; i < sizeof(block) && done < DEGREE; i += 3) { | ||
436 | uint16_t d1 = block[i] + 256 * (block[i + 1] % 16); | ||
437 | uint16_t d2 = block[i + 1] / 16 + 16 * block[i + 2]; | ||
438 | |||
439 | if (d1 < kPrime) { | ||
440 | out->c[done++] = d1; | ||
441 | } | ||
442 | if (d2 < kPrime && done < DEGREE) { | ||
443 | out->c[done++] = d2; | ||
444 | } | ||
445 | } | ||
446 | } | ||
447 | } | ||
448 | |||
449 | /* | ||
450 | * Algorithm 7 of the spec, with eta fixed to two and the PRF call | ||
451 | * included. Creates binominally distributed elements by sampling 2*|eta| bits, | ||
452 | * and setting the coefficient to the count of the first bits minus the count of | ||
453 | * the second bits, resulting in a centered binomial distribution. Since eta is | ||
454 | * two this gives -2/2 with a probability of 1/16, -1/1 with probability 1/4, | ||
455 | * and 0 with probability 3/8. | ||
456 | */ | ||
457 | static void | ||
458 | scalar_centered_binomial_distribution_eta_2_with_prf(scalar *out, | ||
459 | const uint8_t input[33]) | ||
460 | { | ||
461 | uint8_t entropy[128]; | ||
462 | int i; | ||
463 | |||
464 | CTASSERT(sizeof(entropy) == 2 * /*kEta=*/ 2 * DEGREE / 8); | ||
465 | prf(entropy, sizeof(entropy), input); | ||
466 | |||
467 | for (i = 0; i < DEGREE; i += 2) { | ||
468 | uint8_t byte = entropy[i / 2]; | ||
469 | uint16_t value = kPrime; | ||
470 | |||
471 | value += (byte & 1) + ((byte >> 1) & 1); | ||
472 | value -= ((byte >> 2) & 1) + ((byte >> 3) & 1); | ||
473 | out->c[i] = reduce_once(value); | ||
474 | |||
475 | byte >>= 4; | ||
476 | value = kPrime; | ||
477 | value += (byte & 1) + ((byte >> 1) & 1); | ||
478 | value -= ((byte >> 2) & 1) + ((byte >> 3) & 1); | ||
479 | out->c[i + 1] = reduce_once(value); | ||
480 | } | ||
481 | } | ||
482 | |||
483 | /* | ||
484 | * Generates a secret vector by using | ||
485 | * |scalar_centered_binomial_distribution_eta_2_with_prf|, using the given seed | ||
486 | * appending and incrementing |counter| for entry of the vector. | ||
487 | */ | ||
488 | static void | ||
489 | vector_generate_secret_eta_2(vector *out, uint8_t *counter, | ||
490 | const uint8_t seed[32]) | ||
491 | { | ||
492 | uint8_t input[33]; | ||
493 | int i; | ||
494 | |||
495 | memcpy(input, seed, 32); | ||
496 | for (i = 0; i < RANK1024; i++) { | ||
497 | input[32] = (*counter)++; | ||
498 | scalar_centered_binomial_distribution_eta_2_with_prf(&out->v[i], | ||
499 | input); | ||
500 | } | ||
501 | } | ||
502 | |||
503 | /* Expands the matrix of a seed for key generation and for encaps-CPA. */ | ||
504 | static void | ||
505 | matrix_expand(matrix *out, const uint8_t rho[32]) | ||
506 | { | ||
507 | uint8_t input[34]; | ||
508 | int i, j; | ||
509 | |||
510 | memcpy(input, rho, 32); | ||
511 | for (i = 0; i < RANK1024; i++) { | ||
512 | for (j = 0; j < RANK1024; j++) { | ||
513 | sha3_ctx keccak_ctx; | ||
514 | |||
515 | input[32] = i; | ||
516 | input[33] = j; | ||
517 | shake128_init(&keccak_ctx); | ||
518 | shake_update(&keccak_ctx, input, sizeof(input)); | ||
519 | shake_xof(&keccak_ctx); | ||
520 | scalar_from_keccak_vartime(&out->v[i][j], &keccak_ctx); | ||
521 | } | ||
522 | } | ||
523 | } | ||
524 | |||
525 | static const uint8_t kMasks[8] = {0x01, 0x03, 0x07, 0x0f, | ||
526 | 0x1f, 0x3f, 0x7f, 0xff}; | ||
527 | |||
528 | static void | ||
529 | scalar_encode(uint8_t *out, const scalar *s, int bits) | ||
530 | { | ||
531 | uint8_t out_byte = 0; | ||
532 | int i, out_byte_bits = 0; | ||
533 | |||
534 | assert(bits <= (int)sizeof(*s->c) * 8 && bits != 1); | ||
535 | for (i = 0; i < DEGREE; i++) { | ||
536 | uint16_t element = s->c[i]; | ||
537 | int element_bits_done = 0; | ||
538 | |||
539 | while (element_bits_done < bits) { | ||
540 | int chunk_bits = bits - element_bits_done; | ||
541 | int out_bits_remaining = 8 - out_byte_bits; | ||
542 | |||
543 | if (chunk_bits >= out_bits_remaining) { | ||
544 | chunk_bits = out_bits_remaining; | ||
545 | out_byte |= (element & | ||
546 | kMasks[chunk_bits - 1]) << out_byte_bits; | ||
547 | *out = out_byte; | ||
548 | out++; | ||
549 | out_byte_bits = 0; | ||
550 | out_byte = 0; | ||
551 | } else { | ||
552 | out_byte |= (element & | ||
553 | kMasks[chunk_bits - 1]) << out_byte_bits; | ||
554 | out_byte_bits += chunk_bits; | ||
555 | } | ||
556 | |||
557 | element_bits_done += chunk_bits; | ||
558 | element >>= chunk_bits; | ||
559 | } | ||
560 | } | ||
561 | |||
562 | if (out_byte_bits > 0) { | ||
563 | *out = out_byte; | ||
564 | } | ||
565 | } | ||
566 | |||
567 | /* scalar_encode_1 is |scalar_encode| specialised for |bits| == 1. */ | ||
568 | static void | ||
569 | scalar_encode_1(uint8_t out[32], const scalar *s) | ||
570 | { | ||
571 | int i, j; | ||
572 | |||
573 | for (i = 0; i < DEGREE; i += 8) { | ||
574 | uint8_t out_byte = 0; | ||
575 | |||
576 | for (j = 0; j < 8; j++) { | ||
577 | out_byte |= (s->c[i + j] & 1) << j; | ||
578 | } | ||
579 | *out = out_byte; | ||
580 | out++; | ||
581 | } | ||
582 | } | ||
583 | |||
584 | /* | ||
585 | * Encodes an entire vector into 32*|RANK1024|*|bits| bytes. Note that since 256 | ||
586 | * (DEGREE) is divisible by 8, the individual vector entries will always fill a | ||
587 | * whole number of bytes, so we do not need to worry about bit packing here. | ||
588 | */ | ||
589 | static void | ||
590 | vector_encode(uint8_t *out, const vector *a, int bits) | ||
591 | { | ||
592 | int i; | ||
593 | |||
594 | for (i = 0; i < RANK1024; i++) { | ||
595 | scalar_encode(out + i * bits * DEGREE / 8, &a->v[i], bits); | ||
596 | } | ||
597 | } | ||
598 | |||
599 | /* | ||
600 | * scalar_decode parses |DEGREE * bits| bits from |in| into |DEGREE| values in | ||
601 | * |out|. It returns one on success and zero if any parsed value is >= | ||
602 | * |kPrime|. | ||
603 | */ | ||
604 | static int | ||
605 | scalar_decode(scalar *out, const uint8_t *in, int bits) | ||
606 | { | ||
607 | uint8_t in_byte = 0; | ||
608 | int i, in_byte_bits_left = 0; | ||
609 | |||
610 | assert(bits <= (int)sizeof(*out->c) * 8 && bits != 1); | ||
611 | |||
612 | for (i = 0; i < DEGREE; i++) { | ||
613 | uint16_t element = 0; | ||
614 | int element_bits_done = 0; | ||
615 | |||
616 | while (element_bits_done < bits) { | ||
617 | int chunk_bits = bits - element_bits_done; | ||
618 | |||
619 | if (in_byte_bits_left == 0) { | ||
620 | in_byte = *in; | ||
621 | in++; | ||
622 | in_byte_bits_left = 8; | ||
623 | } | ||
624 | |||
625 | if (chunk_bits > in_byte_bits_left) { | ||
626 | chunk_bits = in_byte_bits_left; | ||
627 | } | ||
628 | |||
629 | element |= (in_byte & kMasks[chunk_bits - 1]) << | ||
630 | element_bits_done; | ||
631 | in_byte_bits_left -= chunk_bits; | ||
632 | in_byte >>= chunk_bits; | ||
633 | |||
634 | element_bits_done += chunk_bits; | ||
635 | } | ||
636 | |||
637 | if (element >= kPrime) { | ||
638 | return 0; | ||
639 | } | ||
640 | out->c[i] = element; | ||
641 | } | ||
642 | |||
643 | return 1; | ||
644 | } | ||
645 | |||
646 | /* scalar_decode_1 is |scalar_decode| specialised for |bits| == 1. */ | ||
647 | static void | ||
648 | scalar_decode_1(scalar *out, const uint8_t in[32]) | ||
649 | { | ||
650 | int i, j; | ||
651 | |||
652 | for (i = 0; i < DEGREE; i += 8) { | ||
653 | uint8_t in_byte = *in; | ||
654 | |||
655 | in++; | ||
656 | for (j = 0; j < 8; j++) { | ||
657 | out->c[i + j] = in_byte & 1; | ||
658 | in_byte >>= 1; | ||
659 | } | ||
660 | } | ||
661 | } | ||
662 | |||
663 | /* | ||
664 | * Decodes 32*|RANK1024|*|bits| bytes from |in| into |out|. It returns one on | ||
665 | * success or zero if any parsed value is >= |kPrime|. | ||
666 | */ | ||
667 | static int | ||
668 | vector_decode(vector *out, const uint8_t *in, int bits) | ||
669 | { | ||
670 | int i; | ||
671 | |||
672 | for (i = 0; i < RANK1024; i++) { | ||
673 | if (!scalar_decode(&out->v[i], in + i * bits * DEGREE / 8, | ||
674 | bits)) { | ||
675 | return 0; | ||
676 | } | ||
677 | } | ||
678 | return 1; | ||
679 | } | ||
680 | |||
681 | /* | ||
682 | * Compresses (lossily) an input |x| mod 3329 into |bits| many bits by grouping | ||
683 | * numbers close to each other together. The formula used is | ||
684 | * round(2^|bits|/kPrime*x) mod 2^|bits|. | ||
685 | * Uses Barrett reduction to achieve constant time. Since we need both the | ||
686 | * remainder (for rounding) and the quotient (as the result), we cannot use | ||
687 | * |reduce| here, but need to do the Barrett reduction directly. | ||
688 | */ | ||
689 | static uint16_t | ||
690 | compress(uint16_t x, int bits) | ||
691 | { | ||
692 | uint32_t shifted = (uint32_t)x << bits; | ||
693 | uint64_t product = (uint64_t)shifted * kBarrettMultiplier; | ||
694 | uint32_t quotient = (uint32_t)(product >> kBarrettShift); | ||
695 | uint32_t remainder = shifted - quotient * kPrime; | ||
696 | |||
697 | /* | ||
698 | * Adjust the quotient to round correctly: | ||
699 | * 0 <= remainder <= kHalfPrime round to 0 | ||
700 | * kHalfPrime < remainder <= kPrime + kHalfPrime round to 1 | ||
701 | * kPrime + kHalfPrime < remainder < 2 * kPrime round to 2 | ||
702 | */ | ||
703 | assert(remainder < 2u * kPrime); | ||
704 | quotient += 1 & constant_time_lt(kHalfPrime, remainder); | ||
705 | quotient += 1 & constant_time_lt(kPrime + kHalfPrime, remainder); | ||
706 | return quotient & ((1 << bits) - 1); | ||
707 | } | ||
708 | |||
709 | /* | ||
710 | * Decompresses |x| by using an equi-distant representative. The formula is | ||
711 | * round(kPrime/2^|bits|*x). Note that 2^|bits| being the divisor allows us to | ||
712 | * implement this logic using only bit operations. | ||
713 | */ | ||
714 | static uint16_t | ||
715 | decompress(uint16_t x, int bits) | ||
716 | { | ||
717 | uint32_t product = (uint32_t)x * kPrime; | ||
718 | uint32_t power = 1 << bits; | ||
719 | /* This is |product| % power, since |power| is a power of 2. */ | ||
720 | uint32_t remainder = product & (power - 1); | ||
721 | /* This is |product| / power, since |power| is a power of 2. */ | ||
722 | uint32_t lower = product >> bits; | ||
723 | |||
724 | /* | ||
725 | * The rounding logic works since the first half of numbers mod |power| have a | ||
726 | * 0 as first bit, and the second half has a 1 as first bit, since |power| is | ||
727 | * a power of 2. As a 12 bit number, |remainder| is always positive, so we | ||
728 | * will shift in 0s for a right shift. | ||
729 | */ | ||
730 | return lower + (remainder >> (bits - 1)); | ||
731 | } | ||
732 | |||
733 | static void | ||
734 | scalar_compress(scalar *s, int bits) | ||
735 | { | ||
736 | int i; | ||
737 | |||
738 | for (i = 0; i < DEGREE; i++) { | ||
739 | s->c[i] = compress(s->c[i], bits); | ||
740 | } | ||
741 | } | ||
742 | |||
743 | static void | ||
744 | scalar_decompress(scalar *s, int bits) | ||
745 | { | ||
746 | int i; | ||
747 | |||
748 | for (i = 0; i < DEGREE; i++) { | ||
749 | s->c[i] = decompress(s->c[i], bits); | ||
750 | } | ||
751 | } | ||
752 | |||
753 | static void | ||
754 | vector_compress(vector *a, int bits) | ||
755 | { | ||
756 | int i; | ||
757 | |||
758 | for (i = 0; i < RANK1024; i++) { | ||
759 | scalar_compress(&a->v[i], bits); | ||
760 | } | ||
761 | } | ||
762 | |||
763 | static void | ||
764 | vector_decompress(vector *a, int bits) | ||
765 | { | ||
766 | int i; | ||
767 | |||
768 | for (i = 0; i < RANK1024; i++) { | ||
769 | scalar_decompress(&a->v[i], bits); | ||
770 | } | ||
771 | } | ||
772 | |||
773 | struct public_key { | ||
774 | vector t; | ||
775 | uint8_t rho[32]; | ||
776 | uint8_t public_key_hash[32]; | ||
777 | matrix m; | ||
778 | }; | ||
779 | |||
780 | static struct public_key * | ||
781 | public_key_1024_from_external(const struct MLKEM1024_public_key *external) | ||
782 | { | ||
783 | return (struct public_key *)external; | ||
784 | } | ||
785 | |||
786 | struct private_key { | ||
787 | struct public_key pub; | ||
788 | vector s; | ||
789 | uint8_t fo_failure_secret[32]; | ||
790 | }; | ||
791 | |||
792 | static struct private_key * | ||
793 | private_key_1024_from_external(const struct MLKEM1024_private_key *external) | ||
794 | { | ||
795 | return (struct private_key *)external; | ||
796 | } | ||
797 | |||
798 | /* | ||
799 | * Calls |MLKEM1024_generate_key_external_entropy| with random bytes from | ||
800 | * |RAND_bytes|. | ||
801 | */ | ||
802 | void | ||
803 | MLKEM1024_generate_key(uint8_t out_encoded_public_key[MLKEM1024_PUBLIC_KEY_BYTES], | ||
804 | uint8_t optional_out_seed[MLKEM_SEED_BYTES], | ||
805 | struct MLKEM1024_private_key *out_private_key) | ||
806 | { | ||
807 | uint8_t entropy_buf[MLKEM_SEED_BYTES]; | ||
808 | uint8_t *entropy = optional_out_seed != NULL ? optional_out_seed : | ||
809 | entropy_buf; | ||
810 | |||
811 | arc4random_buf(entropy, MLKEM_SEED_BYTES); | ||
812 | MLKEM1024_generate_key_external_entropy(out_encoded_public_key, | ||
813 | out_private_key, entropy); | ||
814 | } | ||
815 | LCRYPTO_ALIAS(MLKEM1024_generate_key); | ||
816 | |||
817 | int | ||
818 | MLKEM1024_private_key_from_seed(struct MLKEM1024_private_key *out_private_key, | ||
819 | const uint8_t *seed, size_t seed_len) | ||
820 | { | ||
821 | uint8_t public_key_bytes[MLKEM1024_PUBLIC_KEY_BYTES]; | ||
822 | |||
823 | if (seed_len != MLKEM_SEED_BYTES) { | ||
824 | return 0; | ||
825 | } | ||
826 | MLKEM1024_generate_key_external_entropy(public_key_bytes, | ||
827 | out_private_key, seed); | ||
828 | |||
829 | return 1; | ||
830 | } | ||
831 | LCRYPTO_ALIAS(MLKEM1024_private_key_from_seed); | ||
832 | |||
833 | static int | ||
834 | mlkem_marshal_public_key(CBB *out, const struct public_key *pub) | ||
835 | { | ||
836 | uint8_t *vector_output; | ||
837 | |||
838 | if (!CBB_add_space(out, &vector_output, kEncodedVectorSize)) { | ||
839 | return 0; | ||
840 | } | ||
841 | vector_encode(vector_output, &pub->t, kLog2Prime); | ||
842 | if (!CBB_add_bytes(out, pub->rho, sizeof(pub->rho))) { | ||
843 | return 0; | ||
844 | } | ||
845 | return 1; | ||
846 | } | ||
847 | |||
848 | void | ||
849 | MLKEM1024_generate_key_external_entropy( | ||
850 | uint8_t out_encoded_public_key[MLKEM1024_PUBLIC_KEY_BYTES], | ||
851 | struct MLKEM1024_private_key *out_private_key, | ||
852 | const uint8_t entropy[MLKEM_SEED_BYTES]) | ||
853 | { | ||
854 | struct private_key *priv = private_key_1024_from_external( | ||
855 | out_private_key); | ||
856 | uint8_t augmented_seed[33]; | ||
857 | uint8_t *rho, *sigma; | ||
858 | uint8_t counter = 0; | ||
859 | uint8_t hashed[64]; | ||
860 | vector error; | ||
861 | CBB cbb; | ||
862 | |||
863 | memcpy(augmented_seed, entropy, 32); | ||
864 | augmented_seed[32] = RANK1024; | ||
865 | hash_g(hashed, augmented_seed, 33); | ||
866 | rho = hashed; | ||
867 | sigma = hashed + 32; | ||
868 | memcpy(priv->pub.rho, hashed, sizeof(priv->pub.rho)); | ||
869 | matrix_expand(&priv->pub.m, rho); | ||
870 | vector_generate_secret_eta_2(&priv->s, &counter, sigma); | ||
871 | vector_ntt(&priv->s); | ||
872 | vector_generate_secret_eta_2(&error, &counter, sigma); | ||
873 | vector_ntt(&error); | ||
874 | matrix_mult_transpose(&priv->pub.t, &priv->pub.m, &priv->s); | ||
875 | vector_add(&priv->pub.t, &error); | ||
876 | |||
877 | CBB_init_fixed(&cbb, out_encoded_public_key, MLKEM1024_PUBLIC_KEY_BYTES); | ||
878 | if (!mlkem_marshal_public_key(&cbb, &priv->pub)) { | ||
879 | abort(); | ||
880 | } | ||
881 | |||
882 | hash_h(priv->pub.public_key_hash, out_encoded_public_key, | ||
883 | MLKEM1024_PUBLIC_KEY_BYTES); | ||
884 | memcpy(priv->fo_failure_secret, entropy + 32, 32); | ||
885 | } | ||
886 | |||
887 | void | ||
888 | MLKEM1024_public_from_private(struct MLKEM1024_public_key *out_public_key, | ||
889 | const struct MLKEM1024_private_key *private_key) | ||
890 | { | ||
891 | struct public_key *const pub = public_key_1024_from_external( | ||
892 | out_public_key); | ||
893 | const struct private_key *const priv = private_key_1024_from_external( | ||
894 | private_key); | ||
895 | |||
896 | *pub = priv->pub; | ||
897 | } | ||
898 | LCRYPTO_ALIAS(MLKEM1024_public_from_private); | ||
899 | |||
900 | /* | ||
901 | * Encrypts a message with given randomness to the ciphertext in |out|. Without | ||
902 | * applying the Fujisaki-Okamoto transform this would not result in a CCA secure | ||
903 | * scheme, since lattice schemes are vulnerable to decryption failure oracles. | ||
904 | */ | ||
905 | static void | ||
906 | encrypt_cpa(uint8_t out[MLKEM1024_CIPHERTEXT_BYTES], | ||
907 | const struct public_key *pub, const uint8_t message[32], | ||
908 | const uint8_t randomness[32]) | ||
909 | { | ||
910 | scalar expanded_message, scalar_error; | ||
911 | vector secret, error, u; | ||
912 | uint8_t counter = 0; | ||
913 | uint8_t input[33]; | ||
914 | scalar v; | ||
915 | |||
916 | vector_generate_secret_eta_2(&secret, &counter, randomness); | ||
917 | vector_ntt(&secret); | ||
918 | vector_generate_secret_eta_2(&error, &counter, randomness); | ||
919 | memcpy(input, randomness, 32); | ||
920 | input[32] = counter; | ||
921 | scalar_centered_binomial_distribution_eta_2_with_prf(&scalar_error, | ||
922 | input); | ||
923 | matrix_mult(&u, &pub->m, &secret); | ||
924 | vector_inverse_ntt(&u); | ||
925 | vector_add(&u, &error); | ||
926 | scalar_inner_product(&v, &pub->t, &secret); | ||
927 | scalar_inverse_ntt(&v); | ||
928 | scalar_add(&v, &scalar_error); | ||
929 | scalar_decode_1(&expanded_message, message); | ||
930 | scalar_decompress(&expanded_message, 1); | ||
931 | scalar_add(&v, &expanded_message); | ||
932 | vector_compress(&u, kDU1024); | ||
933 | vector_encode(out, &u, kDU1024); | ||
934 | scalar_compress(&v, kDV1024); | ||
935 | scalar_encode(out + kCompressedVectorSize, &v, kDV1024); | ||
936 | } | ||
937 | |||
938 | /* Calls MLKEM1024_encap_external_entropy| with random bytes */ | ||
939 | void | ||
940 | MLKEM1024_encap(uint8_t out_ciphertext[MLKEM1024_CIPHERTEXT_BYTES], | ||
941 | uint8_t out_shared_secret[MLKEM_SHARED_SECRET_BYTES], | ||
942 | const struct MLKEM1024_public_key *public_key) | ||
943 | { | ||
944 | uint8_t entropy[MLKEM_ENCAP_ENTROPY]; | ||
945 | |||
946 | arc4random_buf(entropy, MLKEM_ENCAP_ENTROPY); | ||
947 | MLKEM1024_encap_external_entropy(out_ciphertext, out_shared_secret, | ||
948 | public_key, entropy); | ||
949 | } | ||
950 | LCRYPTO_ALIAS(MLKEM1024_encap); | ||
951 | |||
952 | /* See section 6.2 of the spec. */ | ||
953 | void | ||
954 | MLKEM1024_encap_external_entropy( | ||
955 | uint8_t out_ciphertext[MLKEM1024_CIPHERTEXT_BYTES], | ||
956 | uint8_t out_shared_secret[MLKEM_SHARED_SECRET_BYTES], | ||
957 | const struct MLKEM1024_public_key *public_key, | ||
958 | const uint8_t entropy[MLKEM_ENCAP_ENTROPY]) | ||
959 | { | ||
960 | const struct public_key *pub = public_key_1024_from_external(public_key); | ||
961 | uint8_t key_and_randomness[64]; | ||
962 | uint8_t input[64]; | ||
963 | |||
964 | memcpy(input, entropy, MLKEM_ENCAP_ENTROPY); | ||
965 | memcpy(input + MLKEM_ENCAP_ENTROPY, pub->public_key_hash, | ||
966 | sizeof(input) - MLKEM_ENCAP_ENTROPY); | ||
967 | hash_g(key_and_randomness, input, sizeof(input)); | ||
968 | encrypt_cpa(out_ciphertext, pub, entropy, key_and_randomness + 32); | ||
969 | memcpy(out_shared_secret, key_and_randomness, 32); | ||
970 | } | ||
971 | |||
972 | static void | ||
973 | decrypt_cpa(uint8_t out[32], const struct private_key *priv, | ||
974 | const uint8_t ciphertext[MLKEM1024_CIPHERTEXT_BYTES]) | ||
975 | { | ||
976 | scalar mask, v; | ||
977 | vector u; | ||
978 | |||
979 | vector_decode(&u, ciphertext, kDU1024); | ||
980 | vector_decompress(&u, kDU1024); | ||
981 | vector_ntt(&u); | ||
982 | scalar_decode(&v, ciphertext + kCompressedVectorSize, kDV1024); | ||
983 | scalar_decompress(&v, kDV1024); | ||
984 | scalar_inner_product(&mask, &priv->s, &u); | ||
985 | scalar_inverse_ntt(&mask); | ||
986 | scalar_sub(&v, &mask); | ||
987 | scalar_compress(&v, 1); | ||
988 | scalar_encode_1(out, &v); | ||
989 | } | ||
990 | |||
991 | /* See section 6.3 */ | ||
992 | int | ||
993 | MLKEM1024_decap(uint8_t out_shared_secret[MLKEM_SHARED_SECRET_BYTES], | ||
994 | const uint8_t *ciphertext, size_t ciphertext_len, | ||
995 | const struct MLKEM1024_private_key *private_key) | ||
996 | { | ||
997 | const struct private_key *priv = private_key_1024_from_external( | ||
998 | private_key); | ||
999 | uint8_t expected_ciphertext[MLKEM1024_CIPHERTEXT_BYTES]; | ||
1000 | uint8_t key_and_randomness[64]; | ||
1001 | uint8_t failure_key[32]; | ||
1002 | uint8_t decrypted[64]; | ||
1003 | uint8_t mask; | ||
1004 | int i; | ||
1005 | |||
1006 | if (ciphertext_len != MLKEM1024_CIPHERTEXT_BYTES) { | ||
1007 | arc4random_buf(out_shared_secret, MLKEM_SHARED_SECRET_BYTES); | ||
1008 | return 0; | ||
1009 | } | ||
1010 | |||
1011 | decrypt_cpa(decrypted, priv, ciphertext); | ||
1012 | memcpy(decrypted + 32, priv->pub.public_key_hash, | ||
1013 | sizeof(decrypted) - 32); | ||
1014 | hash_g(key_and_randomness, decrypted, sizeof(decrypted)); | ||
1015 | encrypt_cpa(expected_ciphertext, &priv->pub, decrypted, | ||
1016 | key_and_randomness + 32); | ||
1017 | kdf(failure_key, priv->fo_failure_secret, ciphertext, ciphertext_len); | ||
1018 | mask = constant_time_eq_int_8(memcmp(ciphertext, expected_ciphertext, | ||
1019 | sizeof(expected_ciphertext)), 0); | ||
1020 | for (i = 0; i < MLKEM_SHARED_SECRET_BYTES; i++) { | ||
1021 | out_shared_secret[i] = constant_time_select_8(mask, | ||
1022 | key_and_randomness[i], failure_key[i]); | ||
1023 | } | ||
1024 | |||
1025 | return 1; | ||
1026 | } | ||
1027 | LCRYPTO_ALIAS(MLKEM1024_decap); | ||
1028 | |||
1029 | int | ||
1030 | MLKEM1024_marshal_public_key(CBB *out, | ||
1031 | const struct MLKEM1024_public_key *public_key) | ||
1032 | { | ||
1033 | return mlkem_marshal_public_key(out, | ||
1034 | public_key_1024_from_external(public_key)); | ||
1035 | } | ||
1036 | LCRYPTO_ALIAS(MLKEM1024_marshal_public_key); | ||
1037 | |||
1038 | /* | ||
1039 | * mlkem_parse_public_key_no_hash parses |in| into |pub| but doesn't calculate | ||
1040 | * the value of |pub->public_key_hash|. | ||
1041 | */ | ||
1042 | static int | ||
1043 | mlkem_parse_public_key_no_hash(struct public_key *pub, CBS *in) | ||
1044 | { | ||
1045 | CBS t_bytes; | ||
1046 | |||
1047 | if (!CBS_get_bytes(in, &t_bytes, kEncodedVectorSize) || | ||
1048 | !vector_decode(&pub->t, CBS_data(&t_bytes), kLog2Prime)) { | ||
1049 | return 0; | ||
1050 | } | ||
1051 | memcpy(pub->rho, CBS_data(in), sizeof(pub->rho)); | ||
1052 | if (!CBS_skip(in, sizeof(pub->rho))) | ||
1053 | return 0; | ||
1054 | matrix_expand(&pub->m, pub->rho); | ||
1055 | return 1; | ||
1056 | } | ||
1057 | |||
1058 | int | ||
1059 | MLKEM1024_parse_public_key(struct MLKEM1024_public_key *public_key, CBS *in) | ||
1060 | { | ||
1061 | struct public_key *pub = public_key_1024_from_external(public_key); | ||
1062 | CBS orig_in = *in; | ||
1063 | |||
1064 | if (!mlkem_parse_public_key_no_hash(pub, in) || | ||
1065 | CBS_len(in) != 0) { | ||
1066 | return 0; | ||
1067 | } | ||
1068 | hash_h(pub->public_key_hash, CBS_data(&orig_in), CBS_len(&orig_in)); | ||
1069 | return 1; | ||
1070 | } | ||
1071 | LCRYPTO_ALIAS(MLKEM1024_parse_public_key); | ||
1072 | |||
1073 | int | ||
1074 | MLKEM1024_marshal_private_key(CBB *out, | ||
1075 | const struct MLKEM1024_private_key *private_key) | ||
1076 | { | ||
1077 | const struct private_key *const priv = private_key_1024_from_external( | ||
1078 | private_key); | ||
1079 | uint8_t *s_output; | ||
1080 | |||
1081 | if (!CBB_add_space(out, &s_output, kEncodedVectorSize)) { | ||
1082 | return 0; | ||
1083 | } | ||
1084 | vector_encode(s_output, &priv->s, kLog2Prime); | ||
1085 | if (!mlkem_marshal_public_key(out, &priv->pub) || | ||
1086 | !CBB_add_bytes(out, priv->pub.public_key_hash, | ||
1087 | sizeof(priv->pub.public_key_hash)) || | ||
1088 | !CBB_add_bytes(out, priv->fo_failure_secret, | ||
1089 | sizeof(priv->fo_failure_secret))) { | ||
1090 | return 0; | ||
1091 | } | ||
1092 | return 1; | ||
1093 | } | ||
1094 | |||
1095 | int | ||
1096 | MLKEM1024_parse_private_key(struct MLKEM1024_private_key *out_private_key, | ||
1097 | CBS *in) | ||
1098 | { | ||
1099 | struct private_key *const priv = private_key_1024_from_external( | ||
1100 | out_private_key); | ||
1101 | CBS s_bytes; | ||
1102 | |||
1103 | if (!CBS_get_bytes(in, &s_bytes, kEncodedVectorSize) || | ||
1104 | !vector_decode(&priv->s, CBS_data(&s_bytes), kLog2Prime) || | ||
1105 | !mlkem_parse_public_key_no_hash(&priv->pub, in)) { | ||
1106 | return 0; | ||
1107 | } | ||
1108 | memcpy(priv->pub.public_key_hash, CBS_data(in), | ||
1109 | sizeof(priv->pub.public_key_hash)); | ||
1110 | if (!CBS_skip(in, sizeof(priv->pub.public_key_hash))) | ||
1111 | return 0; | ||
1112 | memcpy(priv->fo_failure_secret, CBS_data(in), | ||
1113 | sizeof(priv->fo_failure_secret)); | ||
1114 | if (!CBS_skip(in, sizeof(priv->fo_failure_secret))) | ||
1115 | return 0; | ||
1116 | if (CBS_len(in) != 0) | ||
1117 | return 0; | ||
1118 | |||
1119 | return 1; | ||
1120 | } | ||
1121 | LCRYPTO_ALIAS(MLKEM1024_parse_private_key); | ||
diff --git a/src/lib/libcrypto/mlkem/mlkem_internal.h b/src/lib/libcrypto/mlkem/mlkem_internal.h index 3ef877f6d1..3141160ac2 100644 --- a/src/lib/libcrypto/mlkem/mlkem_internal.h +++ b/src/lib/libcrypto/mlkem/mlkem_internal.h | |||
@@ -69,6 +69,45 @@ void MLKEM768_encap_external_entropy( | |||
69 | const struct MLKEM768_public_key *public_key, | 69 | const struct MLKEM768_public_key *public_key, |
70 | const uint8_t entropy[MLKEM_ENCAP_ENTROPY]); | 70 | const uint8_t entropy[MLKEM_ENCAP_ENTROPY]); |
71 | 71 | ||
72 | /* | ||
73 | * MLKEM1024_generate_key_external_entropy is a deterministic function to create a | ||
74 | * pair of ML-KEM 1024 keys, using the supplied entropy. The entropy needs to be | ||
75 | * uniformly random generated. This function is should only be used for tests, | ||
76 | * regular callers should use the non-deterministic |MLKEM_generate_key| | ||
77 | * directly. | ||
78 | */ | ||
79 | void MLKEM1024_generate_key_external_entropy( | ||
80 | uint8_t out_encoded_public_key[MLKEM1024_PUBLIC_KEY_BYTES], | ||
81 | struct MLKEM1024_private_key *out_private_key, | ||
82 | const uint8_t entropy[MLKEM_SEED_BYTES]); | ||
83 | |||
84 | /* | ||
85 | * MLKEM1024_PRIVATE_KEY_BYTES is the length of the data produced by | ||
86 | * |MLKEM1024_marshal_private_key|. | ||
87 | */ | ||
88 | #define MLKEM1024_PRIVATE_KEY_BYTES 3168 | ||
89 | |||
90 | /* | ||
91 | * MLKEM1024_marshal_private_key serializes |private_key| to |out| in the | ||
92 | * standard format for ML-KEM private keys. It returns one on success or zero on | ||
93 | * allocation error. | ||
94 | */ | ||
95 | int MLKEM1024_marshal_private_key(CBB *out, | ||
96 | const struct MLKEM1024_private_key *private_key); | ||
97 | |||
98 | /* | ||
99 | * MLKEM_encap_external_entropy behaves like |MLKEM_encap|, but uses | ||
100 | * |MLKEM_ENCAP_ENTROPY| bytes of |entropy| for randomization. The decapsulating | ||
101 | * side will be able to recover |entropy| in full. This function should only be | ||
102 | * used for tests, regular callers should use the non-deterministic | ||
103 | * |MLKEM_encap| directly. | ||
104 | */ | ||
105 | void MLKEM1024_encap_external_entropy( | ||
106 | uint8_t out_ciphertext[MLKEM1024_CIPHERTEXT_BYTES], | ||
107 | uint8_t out_shared_secret[MLKEM_SHARED_SECRET_BYTES], | ||
108 | const struct MLKEM1024_public_key *public_key, | ||
109 | const uint8_t entropy[MLKEM_ENCAP_ENTROPY]); | ||
110 | |||
72 | __END_HIDDEN_DECLS | 111 | __END_HIDDEN_DECLS |
73 | 112 | ||
74 | #if defined(__cplusplus) | 113 | #if defined(__cplusplus) |