author | zautrix <zautrix> | 2004-10-19 20:16:14 (UTC) |
---|---|---|
committer | zautrix <zautrix> | 2004-10-19 20:16:14 (UTC) |
commit | eca49bb06a71980ef61d078904573f25890fc7f2 (patch) (side-by-side diff) | |
tree | c5338e3b12430248979a9ac2c1c7e6646ea9ecdf /pwmanager/libcrypt/cipher/elgamal.c | |
parent | 53cc32b6e7b1f672bf91b2baf2df6c1e8baf3e0a (diff) | |
download | kdepimpi-eca49bb06a71980ef61d078904573f25890fc7f2.zip kdepimpi-eca49bb06a71980ef61d078904573f25890fc7f2.tar.gz kdepimpi-eca49bb06a71980ef61d078904573f25890fc7f2.tar.bz2 |
Initial revision
Diffstat (limited to 'pwmanager/libcrypt/cipher/elgamal.c') (more/less context) (ignore whitespace changes)
-rw-r--r-- | pwmanager/libcrypt/cipher/elgamal.c | 682 |
1 files changed, 682 insertions, 0 deletions
diff --git a/pwmanager/libcrypt/cipher/elgamal.c b/pwmanager/libcrypt/cipher/elgamal.c new file mode 100644 index 0000000..e62b1e5 --- a/dev/null +++ b/pwmanager/libcrypt/cipher/elgamal.c @@ -0,0 +1,682 @@ +/* Elgamal.c - ElGamal Public Key encryption + * Copyright (C) 1998, 2000, 2001, 2002, 2003 Free Software Foundation, Inc. + * + * This file is part of Libgcrypt. + * + * Libgcrypt is free software; you can redistribute it and/or modify + * it under the terms of the GNU Lesser general Public License as + * published by the Free Software Foundation; either version 2.1 of + * the License, or (at your option) any later version. + * + * Libgcrypt is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA + * + * For a description of the algorithm, see: + * Bruce Schneier: Applied Cryptography. John Wiley & Sons, 1996. + * ISBN 0-471-11709-9. Pages 476 ff. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include "g10lib.h" +#include "mpi.h" +#include "cipher.h" + +typedef struct +{ + gcry_mpi_t p; /* prime */ + gcry_mpi_t g; /* group generator */ + gcry_mpi_t y; /* g^x mod p */ +} ELG_public_key; + + +typedef struct +{ + gcry_mpi_t p; /* prime */ + gcry_mpi_t g; /* group generator */ + gcry_mpi_t y; /* g^x mod p */ + gcry_mpi_t x; /* secret exponent */ +} ELG_secret_key; + + +static void test_keys (ELG_secret_key *sk, unsigned nbits); +static gcry_mpi_t gen_k (gcry_mpi_t p, int small_k); +static void generate (ELG_secret_key *sk, unsigned nbits, gcry_mpi_t **factors); +static int check_secret_key (ELG_secret_key *sk); +static void do_encrypt (gcry_mpi_t a, gcry_mpi_t b, gcry_mpi_t input, + ELG_public_key *pkey); +static void decrypt (gcry_mpi_t output, gcry_mpi_t a, gcry_mpi_t b, + ELG_secret_key *skey); +static void sign (gcry_mpi_t a, gcry_mpi_t b, gcry_mpi_t input, + ELG_secret_key *skey); +static int verify (gcry_mpi_t a, gcry_mpi_t b, gcry_mpi_t input, + ELG_public_key *pkey); + + +static void (*progress_cb) (void *, const char *, int, int, int); +static void *progress_cb_data; + +void +_gcry_register_pk_elg_progress (void (*cb) (void *, const char *, + int, int, int), + void *cb_data) +{ + progress_cb = cb; + progress_cb_data = cb_data; +} + + +static void +progress (int c) +{ + if (progress_cb) + progress_cb (progress_cb_data, "pk_elg", c, 0, 0); +} + + +/**************** + * Michael Wiener's table on subgroup sizes to match field sizes + * (floating around somewhere - Fixme: need a reference) + */ +static unsigned int +wiener_map( unsigned int n ) +{ + static struct { unsigned int p_n, q_n; } t[] = + { /* p q attack cost */ + { 512, 119 }, /* 9 x 10^17 */ + { 768, 145 }, /* 6 x 10^21 */ + { 1024, 165 }, /* 7 x 10^24 */ + { 1280, 183 }, /* 3 x 10^27 */ + { 1536, 198 }, /* 7 x 10^29 */ + { 1792, 212 }, /* 9 x 10^31 */ + { 2048, 225 }, /* 8 x 10^33 */ + { 2304, 237 }, /* 5 x 10^35 */ + { 2560, 249 }, /* 3 x 10^37 */ + { 2816, 259 }, /* 1 x 10^39 */ + { 3072, 269 }, /* 3 x 10^40 */ + { 3328, 279 }, /* 8 x 10^41 */ + { 3584, 288 }, /* 2 x 10^43 */ + { 3840, 296 }, /* 4 x 10^44 */ + { 4096, 305 }, /* 7 x 10^45 */ + { 4352, 313 }, /* 1 x 10^47 */ + { 4608, 320 }, /* 2 x 10^48 */ + { 4864, 328 }, /* 2 x 10^49 */ + { 5120, 335 }, /* 3 x 10^50 */ + { 0, 0 } + }; + int i; + + for(i=0; t[i].p_n; i++ ) + { + if( n <= t[i].p_n ) + return t[i].q_n; + } + /* Not in table - use an arbitrary high number. */ + return n / 8 + 200; +} + +static void +test_keys( ELG_secret_key *sk, unsigned nbits ) +{ + ELG_public_key pk; + gcry_mpi_t test = gcry_mpi_new ( 0 ); + gcry_mpi_t out1_a = gcry_mpi_new ( nbits ); + gcry_mpi_t out1_b = gcry_mpi_new ( nbits ); + gcry_mpi_t out2 = gcry_mpi_new ( nbits ); + + pk.p = sk->p; + pk.g = sk->g; + pk.y = sk->y; + + gcry_mpi_randomize( test, nbits, GCRY_WEAK_RANDOM ); + + do_encrypt( out1_a, out1_b, test, &pk ); + decrypt( out2, out1_a, out1_b, sk ); + if( mpi_cmp( test, out2 ) ) + log_fatal("ElGamal operation: encrypt, decrypt failed\n"); + + sign( out1_a, out1_b, test, sk ); + if( !verify( out1_a, out1_b, test, &pk ) ) + log_fatal("ElGamal operation: sign, verify failed\n"); + + gcry_mpi_release ( test ); + gcry_mpi_release ( out1_a ); + gcry_mpi_release ( out1_b ); + gcry_mpi_release ( out2 ); +} + + +/**************** + * Generate a random secret exponent k from prime p, so that k is + * relatively prime to p-1. With SMALL_K set, k will be selected for + * better encryption performance - this must never be used signing! + */ +static gcry_mpi_t +gen_k( gcry_mpi_t p, int small_k ) +{ + gcry_mpi_t k = mpi_alloc_secure( 0 ); + gcry_mpi_t temp = mpi_alloc( mpi_get_nlimbs(p) ); + gcry_mpi_t p_1 = mpi_copy(p); + unsigned int orig_nbits = mpi_get_nbits(p); + unsigned int nbits, nbytes; + char *rndbuf = NULL; + + if (small_k) + { + /* Using a k much lesser than p is sufficient for encryption and + * it greatly improves the encryption performance. We use + * Wiener's table and add a large safety margin. */ + nbits = wiener_map( orig_nbits ) * 3 / 2; + if( nbits >= orig_nbits ) + BUG(); + } + else + nbits = orig_nbits; + + + nbytes = (nbits+7)/8; + if( DBG_CIPHER ) + log_debug("choosing a random k "); + mpi_sub_ui( p_1, p, 1); + for(;;) + { + if( !rndbuf || nbits < 32 ) + { + gcry_free(rndbuf); + rndbuf = gcry_random_bytes_secure( nbytes, GCRY_STRONG_RANDOM ); + } + else + { + /* Change only some of the higher bits. We could improve + this by directly requesting more memory at the first call + to get_random_bytes() and use this the here maybe it is + easier to do this directly in random.c Anyway, it is + highly inlikely that we will ever reach this code. */ + char *pp = gcry_random_bytes_secure( 4, GCRY_STRONG_RANDOM ); + memcpy( rndbuf, pp, 4 ); + gcry_free(pp); + } + _gcry_mpi_set_buffer( k, rndbuf, nbytes, 0 ); + + for(;;) + { + if( !(mpi_cmp( k, p_1 ) < 0) ) /* check: k < (p-1) */ + { + if( DBG_CIPHER ) + progress('+'); + break; /* no */ + } + if( !(mpi_cmp_ui( k, 0 ) > 0) ) /* check: k > 0 */ + { + if( DBG_CIPHER ) + progress('-'); + break; /* no */ + } + if (gcry_mpi_gcd( temp, k, p_1 )) + goto found; /* okay, k is relative prime to (p-1) */ + mpi_add_ui( k, k, 1 ); + if( DBG_CIPHER ) + progress('.'); + } + } + found: + gcry_free(rndbuf); + if( DBG_CIPHER ) + progress('\n'); + mpi_free(p_1); + mpi_free(temp); + + return k; +} + +/**************** + * Generate a key pair with a key of size NBITS + * Returns: 2 structures filles with all needed values + * and an array with n-1 factors of (p-1) + */ +static void +generate ( ELG_secret_key *sk, unsigned int nbits, gcry_mpi_t **ret_factors ) +{ + gcry_mpi_t p; /* the prime */ + gcry_mpi_t p_min1; + gcry_mpi_t g; + gcry_mpi_t x; /* the secret exponent */ + gcry_mpi_t y; + gcry_mpi_t temp; + unsigned int qbits; + unsigned int xbits; + byte *rndbuf; + + p_min1 = gcry_mpi_new ( nbits ); + temp = gcry_mpi_new( nbits ); + qbits = wiener_map( nbits ); + if( qbits & 1 ) /* better have a even one */ + qbits++; + g = mpi_alloc(1); + p = _gcry_generate_elg_prime( 0, nbits, qbits, g, ret_factors ); + mpi_sub_ui(p_min1, p, 1); + + + /* Select a random number which has these properties: + * 0 < x < p-1 + * This must be a very good random number because this is the + * secret part. The prime is public and may be shared anyway, + * so a random generator level of 1 is used for the prime. + * + * I don't see a reason to have a x of about the same size + * as the p. It should be sufficient to have one about the size + * of q or the later used k plus a large safety margin. Decryption + * will be much faster with such an x. + */ + xbits = qbits * 3 / 2; + if( xbits >= nbits ) + BUG(); + x = gcry_mpi_snew ( xbits ); + if( DBG_CIPHER ) + log_debug("choosing a random x of size %u", xbits ); + rndbuf = NULL; + do + { + if( DBG_CIPHER ) + progress('.'); + if( rndbuf ) + { /* Change only some of the higher bits */ + if( xbits < 16 ) /* should never happen ... */ + { + gcry_free(rndbuf); + rndbuf = gcry_random_bytes_secure( (xbits+7)/8, + GCRY_VERY_STRONG_RANDOM ); + } + else + { + char *r = gcry_random_bytes_secure( 2, + GCRY_VERY_STRONG_RANDOM ); + memcpy(rndbuf, r, 2 ); + gcry_free(r); + } + } + else + { + rndbuf = gcry_random_bytes_secure( (xbits+7)/8, + GCRY_VERY_STRONG_RANDOM ); + } + _gcry_mpi_set_buffer( x, rndbuf, (xbits+7)/8, 0 ); + mpi_clear_highbit( x, xbits+1 ); + } + while( !( mpi_cmp_ui( x, 0 )>0 && mpi_cmp( x, p_min1 )<0 ) ); + gcry_free(rndbuf); + + y = gcry_mpi_new (nbits); + gcry_mpi_powm( y, g, x, p ); + + if( DBG_CIPHER ) + { + progress('\n'); + log_mpidump("elg p= ", p ); + log_mpidump("elg g= ", g ); + log_mpidump("elg y= ", y ); + log_mpidump("elg x= ", x ); + } + + /* Copy the stuff to the key structures */ + sk->p = p; + sk->g = g; + sk->y = y; + sk->x = x; + + /* Now we can test our keys (this should never fail!) */ + test_keys( sk, nbits - 64 ); + + gcry_mpi_release ( p_min1 ); + gcry_mpi_release ( temp ); +} + + +/**************** + * Test whether the secret key is valid. + * Returns: if this is a valid key. + */ +static int +check_secret_key( ELG_secret_key *sk ) +{ + int rc; + gcry_mpi_t y = mpi_alloc( mpi_get_nlimbs(sk->y) ); + + gcry_mpi_powm( y, sk->g, sk->x, sk->p ); + rc = !mpi_cmp( y, sk->y ); + mpi_free( y ); + return rc; +} + + +static void +do_encrypt(gcry_mpi_t a, gcry_mpi_t b, gcry_mpi_t input, ELG_public_key *pkey ) +{ + gcry_mpi_t k; + + /* Note: maybe we should change the interface, so that it + * is possible to check that input is < p and return an + * error code. + */ + + k = gen_k( pkey->p, 1 ); + gcry_mpi_powm( a, pkey->g, k, pkey->p ); + /* b = (y^k * input) mod p + * = ((y^k mod p) * (input mod p)) mod p + * and because input is < p + * = ((y^k mod p) * input) mod p + */ + gcry_mpi_powm( b, pkey->y, k, pkey->p ); + gcry_mpi_mulm( b, b, input, pkey->p ); +#if 0 + if( DBG_CIPHER ) + { + log_mpidump("elg encrypted y= ", pkey->y); + log_mpidump("elg encrypted p= ", pkey->p); + log_mpidump("elg encrypted k= ", k); + log_mpidump("elg encrypted M= ", input); + log_mpidump("elg encrypted a= ", a); + log_mpidump("elg encrypted b= ", b); + } +#endif + mpi_free(k); +} + + + + +static void +decrypt(gcry_mpi_t output, gcry_mpi_t a, gcry_mpi_t b, ELG_secret_key *skey ) +{ + gcry_mpi_t t1 = mpi_alloc_secure( mpi_get_nlimbs( skey->p ) ); + + /* output = b/(a^x) mod p */ + gcry_mpi_powm( t1, a, skey->x, skey->p ); + mpi_invm( t1, t1, skey->p ); + mpi_mulm( output, b, t1, skey->p ); +#if 0 + if( DBG_CIPHER ) + { + log_mpidump("elg decrypted x= ", skey->x); + log_mpidump("elg decrypted p= ", skey->p); + log_mpidump("elg decrypted a= ", a); + log_mpidump("elg decrypted b= ", b); + log_mpidump("elg decrypted M= ", output); + } +#endif + mpi_free(t1); +} + + +/**************** + * Make an Elgamal signature out of INPUT + */ + +static void +sign(gcry_mpi_t a, gcry_mpi_t b, gcry_mpi_t input, ELG_secret_key *skey ) +{ + gcry_mpi_t k; + gcry_mpi_t t = mpi_alloc( mpi_get_nlimbs(a) ); + gcry_mpi_t inv = mpi_alloc( mpi_get_nlimbs(a) ); + gcry_mpi_t p_1 = mpi_copy(skey->p); + + /* + * b = (t * inv) mod (p-1) + * b = (t * inv(k,(p-1),(p-1)) mod (p-1) + * b = (((M-x*a) mod (p-1)) * inv(k,(p-1),(p-1))) mod (p-1) + * + */ + mpi_sub_ui(p_1, p_1, 1); + k = gen_k( skey->p, 0 /* no small K ! */ ); + gcry_mpi_powm( a, skey->g, k, skey->p ); + mpi_mul(t, skey->x, a ); + mpi_subm(t, input, t, p_1 ); + mpi_invm(inv, k, p_1 ); + mpi_mulm(b, t, inv, p_1 ); + +#if 0 + if( DBG_CIPHER ) + { + log_mpidump("elg sign p= ", skey->p); + log_mpidump("elg sign g= ", skey->g); + log_mpidump("elg sign y= ", skey->y); + log_mpidump("elg sign x= ", skey->x); + log_mpidump("elg sign k= ", k); + log_mpidump("elg sign M= ", input); + log_mpidump("elg sign a= ", a); + log_mpidump("elg sign b= ", b); + } +#endif + mpi_free(k); + mpi_free(t); + mpi_free(inv); + mpi_free(p_1); +} + + +/**************** + * Returns true if the signature composed of A and B is valid. + */ +static int +verify(gcry_mpi_t a, gcry_mpi_t b, gcry_mpi_t input, ELG_public_key *pkey ) +{ + int rc; + gcry_mpi_t t1; + gcry_mpi_t t2; + gcry_mpi_t base[4]; + gcry_mpi_t ex[4]; + + if( !(mpi_cmp_ui( a, 0 ) > 0 && mpi_cmp( a, pkey->p ) < 0) ) + return 0; /* assertion 0 < a < p failed */ + + t1 = mpi_alloc( mpi_get_nlimbs(a) ); + t2 = mpi_alloc( mpi_get_nlimbs(a) ); + +#if 0 + /* t1 = (y^a mod p) * (a^b mod p) mod p */ + gcry_mpi_powm( t1, pkey->y, a, pkey->p ); + gcry_mpi_powm( t2, a, b, pkey->p ); + mpi_mulm( t1, t1, t2, pkey->p ); + + /* t2 = g ^ input mod p */ + gcry_mpi_powm( t2, pkey->g, input, pkey->p ); + + rc = !mpi_cmp( t1, t2 ); +#elif 0 + /* t1 = (y^a mod p) * (a^b mod p) mod p */ + base[0] = pkey->y; ex[0] = a; + base[1] = a; ex[1] = b; + base[2] = NULL; ex[2] = NULL; + mpi_mulpowm( t1, base, ex, pkey->p ); + + /* t2 = g ^ input mod p */ + gcry_mpi_powm( t2, pkey->g, input, pkey->p ); + + rc = !mpi_cmp( t1, t2 ); +#else + /* t1 = g ^ - input * y ^ a * a ^ b mod p */ + mpi_invm(t2, pkey->g, pkey->p ); + base[0] = t2 ; ex[0] = input; + base[1] = pkey->y; ex[1] = a; + base[2] = a; ex[2] = b; + base[3] = NULL; ex[3] = NULL; + mpi_mulpowm( t1, base, ex, pkey->p ); + rc = !mpi_cmp_ui( t1, 1 ); + +#endif + + mpi_free(t1); + mpi_free(t2); + return rc; +} + +/********************************************* + ************** interface ****************** + *********************************************/ + +gcry_err_code_t +_gcry_elg_generate (int algo, unsigned nbits, unsigned long dummy, + gcry_mpi_t *skey, gcry_mpi_t **retfactors) +{ + ELG_secret_key sk; + + generate (&sk, nbits, retfactors); + skey[0] = sk.p; + skey[1] = sk.g; + skey[2] = sk.y; + skey[3] = sk.x; + + return GPG_ERR_NO_ERROR; +} + + +gcry_err_code_t +_gcry_elg_check_secret_key (int algo, gcry_mpi_t *skey) +{ + gcry_err_code_t err = GPG_ERR_NO_ERROR; + ELG_secret_key sk; + + if ((! skey[0]) || (! skey[1]) || (! skey[2]) || (! skey[3])) + err = GPG_ERR_BAD_MPI; + else + { + sk.p = skey[0]; + sk.g = skey[1]; + sk.y = skey[2]; + sk.x = skey[3]; + + if (! check_secret_key (&sk)) + err = GPG_ERR_BAD_SECKEY; + } + + return err; +} + + +gcry_err_code_t +_gcry_elg_encrypt (int algo, gcry_mpi_t *resarr, + gcry_mpi_t data, gcry_mpi_t *pkey, int flags) +{ + gcry_err_code_t err = GPG_ERR_NO_ERROR; + ELG_public_key pk; + + if ((! data) || (! pkey[0]) || (! pkey[1]) || (! pkey[2])) + err = GPG_ERR_BAD_MPI; + else + { + pk.p = pkey[0]; + pk.g = pkey[1]; + pk.y = pkey[2]; + resarr[0] = mpi_alloc (mpi_get_nlimbs (pk.p)); + resarr[1] = mpi_alloc (mpi_get_nlimbs (pk.p)); + do_encrypt (resarr[0], resarr[1], data, &pk); + } + return err; +} + + +gcry_err_code_t +_gcry_elg_decrypt (int algo, gcry_mpi_t *result, + gcry_mpi_t *data, gcry_mpi_t *skey, int flags) +{ + gcry_err_code_t err = GPG_ERR_NO_ERROR; + ELG_secret_key sk; + + if ((! data[0]) || (! data[1]) + || (! skey[0]) || (! skey[1]) || (! skey[2]) || (! skey[3])) + err = GPG_ERR_BAD_MPI; + else + { + sk.p = skey[0]; + sk.g = skey[1]; + sk.y = skey[2]; + sk.x = skey[3]; + *result = mpi_alloc_secure (mpi_get_nlimbs (sk.p)); + decrypt (*result, data[0], data[1], &sk); + } + return err; +} + + +gcry_err_code_t +_gcry_elg_sign (int algo, gcry_mpi_t *resarr, gcry_mpi_t data, gcry_mpi_t *skey) +{ + gcry_err_code_t err = GPG_ERR_NO_ERROR; + ELG_secret_key sk; + + if ((! data) + || (! skey[0]) || (! skey[1]) || (! skey[2]) || (! skey[3])) + err = GPG_ERR_BAD_MPI; + else + { + sk.p = skey[0]; + sk.g = skey[1]; + sk.y = skey[2]; + sk.x = skey[3]; + resarr[0] = mpi_alloc (mpi_get_nlimbs (sk.p)); + resarr[1] = mpi_alloc (mpi_get_nlimbs (sk.p)); + sign (resarr[0], resarr[1], data, &sk); + } + + return err; +} + +gcry_err_code_t +_gcry_elg_verify (int algo, gcry_mpi_t hash, gcry_mpi_t *data, gcry_mpi_t *pkey, + int (*cmp) (void *, gcry_mpi_t), void *opaquev) +{ + gcry_err_code_t err = GPG_ERR_NO_ERROR; + ELG_public_key pk; + + if ((! data[0]) || (! data[1]) || (! hash) + || (! pkey[0]) || (! pkey[1]) || (! pkey[2])) + err = GPG_ERR_BAD_MPI; + else + { + pk.p = pkey[0]; + pk.g = pkey[1]; + pk.y = pkey[2]; + if (! verify (data[0], data[1], hash, &pk)) + err = GPG_ERR_BAD_SIGNATURE; + } + + return err; +} + + +unsigned int +_gcry_elg_get_nbits (int algo, gcry_mpi_t *pkey) +{ + return mpi_get_nbits (pkey[0]); +} + +static char *elg_names[] = + { + "elg", + "openpgp-elg", + "openpgp-elg-sig", + NULL, + }; + + +gcry_pk_spec_t _gcry_pubkey_spec_elg = + { + "ELG", elg_names, + "pgy", "pgyx", "ab", "rs", "pgy", + GCRY_PK_USAGE_SIGN | GCRY_PK_USAGE_ENCR, + _gcry_elg_generate, + _gcry_elg_check_secret_key, + _gcry_elg_encrypt, + _gcry_elg_decrypt, + _gcry_elg_sign, + _gcry_elg_verify, + _gcry_elg_get_nbits, + }; |