twofish.c 14.3 KB
Newer Older
1 2
/* twofish.c
 *
3
 * The twofish block cipher.
4
 */
5

6
/* twofish - An implementation of the twofish cipher.
Ruud de Rooij's avatar
Ruud de Rooij committed
7 8 9 10 11
 * Copyright (C) 1999 Ruud de Rooij <ruud@debian.org>
 *
 * Modifications for lsh, integrated testing
 * Copyright (C) 1999 J.H.M. Dassen (Ray) <jdassen@wi.LeidenUniv.nl>
 *
12
 * Integrated with the nettle library,
Niels Möller's avatar
Niels Möller committed
13
 * Copyright (C) 2001 Niels Möller
Ruud de Rooij's avatar
Ruud de Rooij committed
14 15
 */

16 17 18 19 20 21 22
/* nettle, low-level cryptographics library
 *
 * The nettle library 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.
 * 
Niels Möller's avatar
Niels Möller committed
23
 * The nettle Library is distributed in the hope that it will be useful, but
24 25 26 27 28
 * 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
Niels Möller's avatar
Niels Möller committed
29
 * along with the nettle library; see the file COPYING.LIB.  If not, write to
30 31
 * the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
 * MA 02111-1301, USA.
Ruud de Rooij's avatar
Ruud de Rooij committed
32 33
 */

34 35 36
#if HAVE_CONFIG_H
# include "config.h"
#endif
37 38

#include <assert.h>
Niels Möller's avatar
Niels Möller committed
39
#include <string.h>
40

41 42 43 44
#include "twofish.h"

#include "macros.h"

Ruud de Rooij's avatar
Ruud de Rooij committed
45 46 47 48 49 50 51 52 53 54 55 56 57
/* Bitwise rotations on 32-bit words.  These are defined as macros that
 * evaluate their argument twice, so do not apply to any expressions with
 * side effects.
 */

#define rol1(x) (((x) << 1) | (((x) & 0x80000000) >> 31))
#define rol8(x) (((x) << 8) | (((x) & 0xFF000000) >> 24))
#define rol9(x) (((x) << 9) | (((x) & 0xFF800000) >> 23))
#define ror1(x) (((x) >> 1) | (((x) & 0x00000001) << 31))

/* ------------------------------------------------------------------------- */

/* The permutations q0 and q1.  These are fixed permutations on 8-bit values.
58
 * The permutations have been computed using the program twofish-data,
Ruud de Rooij's avatar
Ruud de Rooij committed
59 60 61
 * which is distributed along with this file.
 */

62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
static const uint8_t q0[256] = {
  0xA9,0x67,0xB3,0xE8,0x04,0xFD,0xA3,0x76,
  0x9A,0x92,0x80,0x78,0xE4,0xDD,0xD1,0x38,
  0x0D,0xC6,0x35,0x98,0x18,0xF7,0xEC,0x6C,
  0x43,0x75,0x37,0x26,0xFA,0x13,0x94,0x48,
  0xF2,0xD0,0x8B,0x30,0x84,0x54,0xDF,0x23,
  0x19,0x5B,0x3D,0x59,0xF3,0xAE,0xA2,0x82,
  0x63,0x01,0x83,0x2E,0xD9,0x51,0x9B,0x7C,
  0xA6,0xEB,0xA5,0xBE,0x16,0x0C,0xE3,0x61,
  0xC0,0x8C,0x3A,0xF5,0x73,0x2C,0x25,0x0B,
  0xBB,0x4E,0x89,0x6B,0x53,0x6A,0xB4,0xF1,
  0xE1,0xE6,0xBD,0x45,0xE2,0xF4,0xB6,0x66,
  0xCC,0x95,0x03,0x56,0xD4,0x1C,0x1E,0xD7,
  0xFB,0xC3,0x8E,0xB5,0xE9,0xCF,0xBF,0xBA,
  0xEA,0x77,0x39,0xAF,0x33,0xC9,0x62,0x71,
  0x81,0x79,0x09,0xAD,0x24,0xCD,0xF9,0xD8,
  0xE5,0xC5,0xB9,0x4D,0x44,0x08,0x86,0xE7,
  0xA1,0x1D,0xAA,0xED,0x06,0x70,0xB2,0xD2,
  0x41,0x7B,0xA0,0x11,0x31,0xC2,0x27,0x90,
  0x20,0xF6,0x60,0xFF,0x96,0x5C,0xB1,0xAB,
  0x9E,0x9C,0x52,0x1B,0x5F,0x93,0x0A,0xEF,
  0x91,0x85,0x49,0xEE,0x2D,0x4F,0x8F,0x3B,
  0x47,0x87,0x6D,0x46,0xD6,0x3E,0x69,0x64,
  0x2A,0xCE,0xCB,0x2F,0xFC,0x97,0x05,0x7A,
  0xAC,0x7F,0xD5,0x1A,0x4B,0x0E,0xA7,0x5A,
  0x28,0x14,0x3F,0x29,0x88,0x3C,0x4C,0x02,
  0xB8,0xDA,0xB0,0x17,0x55,0x1F,0x8A,0x7D,
  0x57,0xC7,0x8D,0x74,0xB7,0xC4,0x9F,0x72,
  0x7E,0x15,0x22,0x12,0x58,0x07,0x99,0x34,
  0x6E,0x50,0xDE,0x68,0x65,0xBC,0xDB,0xF8,
  0xC8,0xA8,0x2B,0x40,0xDC,0xFE,0x32,0xA4,
  0xCA,0x10,0x21,0xF0,0xD3,0x5D,0x0F,0x00,
  0x6F,0x9D,0x36,0x42,0x4A,0x5E,0xC1,0xE0,
};
Ruud de Rooij's avatar
Ruud de Rooij committed
96

97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
static const uint8_t q1[256] = {
  0x75,0xF3,0xC6,0xF4,0xDB,0x7B,0xFB,0xC8,
  0x4A,0xD3,0xE6,0x6B,0x45,0x7D,0xE8,0x4B,
  0xD6,0x32,0xD8,0xFD,0x37,0x71,0xF1,0xE1,
  0x30,0x0F,0xF8,0x1B,0x87,0xFA,0x06,0x3F,
  0x5E,0xBA,0xAE,0x5B,0x8A,0x00,0xBC,0x9D,
  0x6D,0xC1,0xB1,0x0E,0x80,0x5D,0xD2,0xD5,
  0xA0,0x84,0x07,0x14,0xB5,0x90,0x2C,0xA3,
  0xB2,0x73,0x4C,0x54,0x92,0x74,0x36,0x51,
  0x38,0xB0,0xBD,0x5A,0xFC,0x60,0x62,0x96,
  0x6C,0x42,0xF7,0x10,0x7C,0x28,0x27,0x8C,
  0x13,0x95,0x9C,0xC7,0x24,0x46,0x3B,0x70,
  0xCA,0xE3,0x85,0xCB,0x11,0xD0,0x93,0xB8,
  0xA6,0x83,0x20,0xFF,0x9F,0x77,0xC3,0xCC,
  0x03,0x6F,0x08,0xBF,0x40,0xE7,0x2B,0xE2,
  0x79,0x0C,0xAA,0x82,0x41,0x3A,0xEA,0xB9,
  0xE4,0x9A,0xA4,0x97,0x7E,0xDA,0x7A,0x17,
  0x66,0x94,0xA1,0x1D,0x3D,0xF0,0xDE,0xB3,
  0x0B,0x72,0xA7,0x1C,0xEF,0xD1,0x53,0x3E,
  0x8F,0x33,0x26,0x5F,0xEC,0x76,0x2A,0x49,
  0x81,0x88,0xEE,0x21,0xC4,0x1A,0xEB,0xD9,
  0xC5,0x39,0x99,0xCD,0xAD,0x31,0x8B,0x01,
  0x18,0x23,0xDD,0x1F,0x4E,0x2D,0xF9,0x48,
  0x4F,0xF2,0x65,0x8E,0x78,0x5C,0x58,0x19,
  0x8D,0xE5,0x98,0x57,0x67,0x7F,0x05,0x64,
  0xAF,0x63,0xB6,0xFE,0xF5,0xB7,0x3C,0xA5,
  0xCE,0xE9,0x68,0x44,0xE0,0x4D,0x43,0x69,
  0x29,0x2E,0xAC,0x15,0x59,0xA8,0x0A,0x9E,
  0x6E,0x47,0xDF,0x34,0x35,0x6A,0xCF,0xDC,
  0x22,0xC9,0xC0,0x9B,0x89,0xD4,0xED,0xAB,
  0x12,0xA2,0x0D,0x52,0xBB,0x02,0x2F,0xA9,
  0xD7,0x61,0x1E,0xB4,0x50,0x04,0xF6,0xC2,
  0x16,0x25,0x86,0x56,0x55,0x09,0xBE,0x91,
};
Ruud de Rooij's avatar
Ruud de Rooij committed
131 132 133

/* ------------------------------------------------------------------------- */

134
/* uint8_t gf_multiply(uint8_t p, uint8_t a, uint8_t b)
Ruud de Rooij's avatar
Ruud de Rooij committed
135 136 137 138 139 140 141 142 143 144 145 146 147
 *
 * Multiplication in GF(2^8).
 *
 * This function multiplies a times b in the Galois Field GF(2^8) with
 * primitive polynomial p.
 * The representation of the polynomials a, b, and p uses bits with
 * values 2^i to represent the terms x^i.  The polynomial p contains an
 * implicit term x^8.
 *
 * Note that addition and subtraction in GF(2^8) is simply the XOR
 * operation.
 */

148 149
static uint8_t
gf_multiply(uint8_t p, uint8_t a, uint8_t b)
Ruud de Rooij's avatar
Ruud de Rooij committed
150
{
151 152
  uint32_t shift  = b;
  uint8_t result = 0;
153 154 155 156 157 158
  while (a)
    {
      if (a & 1) result ^= shift;
      a = a >> 1;
      shift = shift << 1;
      if (shift & 0x100) shift ^= p;
Ruud de Rooij's avatar
Ruud de Rooij committed
159
    }
160
  return result;
Ruud de Rooij's avatar
Ruud de Rooij committed
161 162 163 164 165 166
}

/* ------------------------------------------------------------------------- */

/* The matrix RS as specified in section 4.3 the twofish paper. */

167
static const uint8_t rs_matrix[4][8] = {
Ruud de Rooij's avatar
Ruud de Rooij committed
168 169 170 171 172
    { 0x01, 0xA4, 0x55, 0x87, 0x5A, 0x58, 0xDB, 0x9E },
    { 0xA4, 0x56, 0x82, 0xF3, 0x1E, 0xC6, 0x68, 0xE5 },
    { 0x02, 0xA1, 0xFC, 0xC1, 0x47, 0xAE, 0x3D, 0x19 },
    { 0xA4, 0x55, 0x87, 0x5A, 0x58, 0xDB, 0x9E, 0x03 } };

173
/* uint32_t compute_s(uint32_t m1, uint32_t m2);
Ruud de Rooij's avatar
Ruud de Rooij committed
174 175 176 177 178 179 180 181 182
 *
 * Computes the value RS * M, where M is a byte vector composed of the
 * bytes of m1 and m2.  Arithmetic is done in GF(2^8) with primitive
 * polynomial x^8 + x^6 + x^3 + x^2 + 1.
 *
 * This function is used to compute the sub-keys S which are in turn used
 * to generate the S-boxes.
 */

183 184
static uint32_t
compute_s(uint32_t m1, uint32_t m2)
Ruud de Rooij's avatar
Ruud de Rooij committed
185
{
186
  uint32_t s = 0;
187 188 189 190 191 192 193 194 195 196 197
  int i;
  for (i = 0; i < 4; i++)
    s |=  ((  gf_multiply(0x4D, m1,       rs_matrix[i][0])
	    ^ gf_multiply(0x4D, m1 >> 8,  rs_matrix[i][1])
	    ^ gf_multiply(0x4D, m1 >> 16, rs_matrix[i][2])
	    ^ gf_multiply(0x4D, m1 >> 24, rs_matrix[i][3])
	    ^ gf_multiply(0x4D, m2,       rs_matrix[i][4])
	    ^ gf_multiply(0x4D, m2 >> 8,  rs_matrix[i][5])
	    ^ gf_multiply(0x4D, m2 >> 16, rs_matrix[i][6])
	    ^ gf_multiply(0x4D, m2 >> 24, rs_matrix[i][7])) << (i*8));
  return s;
Ruud de Rooij's avatar
Ruud de Rooij committed
198 199 200 201 202 203 204 205
}

/* ------------------------------------------------------------------------- */

/* This table describes which q S-boxes are used for each byte in each stage
 * of the function h, cf. figure 2 of the twofish paper.
 */

206 207 208 209 210
static const uint8_t * const q_table[4][5] =
  { { q1, q1, q0, q0, q1 },
    { q0, q1, q1, q0, q0 },
    { q0, q0, q0, q1, q1 },
    { q1, q0, q1, q1, q0 } };
Ruud de Rooij's avatar
Ruud de Rooij committed
211 212 213

/* The matrix MDS as specified in section 4.3.2 of the twofish paper. */

214
static const uint8_t mds_matrix[4][4] = { { 0x01, 0xEF, 0x5B, 0x5B },
215 216 217
				 { 0x5B, 0xEF, 0xEF, 0x01 },
				 { 0xEF, 0x5B, 0x01, 0xEF },
				 { 0xEF, 0x01, 0xEF, 0x5B } };
Ruud de Rooij's avatar
Ruud de Rooij committed
218

219
/* uint32_t h_uint8_t(int k, int i, uint8_t x, uint8_t l0, uint8_t l1, uint8_t l2, uint8_t l3);
Ruud de Rooij's avatar
Ruud de Rooij committed
220 221 222 223 224 225 226 227 228
 *
 * Perform the h function (section 4.3.2) on one byte.  It consists of
 * repeated applications of the q permutation, followed by a XOR with
 * part of a sub-key.  Finally, the value is multiplied by one column of
 * the MDS matrix.  To obtain the result for a full word, the results of
 * h for the individual bytes are XORed.
 *
 * k is the key size (/ 64 bits), i is the byte number (0 = LSB), x is the
 * actual byte to apply the function to; l0, l1, l2, and l3 are the
229
 * appropriate bytes from the subkey.  Note that only l0..l(k-1) are used.
Ruud de Rooij's avatar
Ruud de Rooij committed
230 231
 */

232 233
static uint32_t
h_byte(int k, int i, uint8_t x, uint8_t l0, uint8_t l1, uint8_t l2, uint8_t l3)
Ruud de Rooij's avatar
Ruud de Rooij committed
234
{
235
  uint8_t y = q_table[i][4][l0 ^
236 237 238 239
            q_table[i][3][l1 ^
              q_table[i][2][k == 2 ? x : l2 ^
                q_table[i][1][k == 3 ? x : l3 ^ q_table[i][0][x]]]]];

240 241 242 243
  return ( ((uint32_t)gf_multiply(0x69, mds_matrix[0][i], y))
	   | ((uint32_t)gf_multiply(0x69, mds_matrix[1][i], y) << 8)
	   | ((uint32_t)gf_multiply(0x69, mds_matrix[2][i], y) << 16)
	   | ((uint32_t)gf_multiply(0x69, mds_matrix[3][i], y) << 24) );
Ruud de Rooij's avatar
Ruud de Rooij committed
244 245
}

246
/* uint32_t h(int k, uint8_t x, uint32_t l0, uint32_t l1, uint32_t l2, uint32_t l3);
Ruud de Rooij's avatar
Ruud de Rooij committed
247 248 249 250
 *
 * Perform the function h on a word.  See the description of h_byte() above.
 */

251 252
static uint32_t
h(int k, uint8_t x, uint32_t l0, uint32_t l1, uint32_t l2, uint32_t l3)
Ruud de Rooij's avatar
Ruud de Rooij committed
253
{
254 255 256 257
  return (  h_byte(k, 0, x, l0,       l1,       l2,       l3)
	  ^ h_byte(k, 1, x, l0 >> 8,  l1 >> 8,  l2 >> 8,  l3 >> 8)
	  ^ h_byte(k, 2, x, l0 >> 16, l1 >> 16, l2 >> 16, l3 >> 16)
	  ^ h_byte(k, 3, x, l0 >> 24, l1 >> 24, l2 >> 24, l3 >> 24) );
Ruud de Rooij's avatar
Ruud de Rooij committed
258 259 260 261 262 263 264 265 266 267 268 269
}


/* ------------------------------------------------------------------------- */

/* API */

/* Structure which contains the tables containing the subkeys and the
 * key-dependent s-boxes.
 */


270
/* Set up internal tables required for twofish encryption and decryption.
Ruud de Rooij's avatar
Ruud de Rooij committed
271 272
 *
 * The key size is specified in bytes.  Key sizes up to 32 bytes are
273
 * supported.  Larger key sizes are silently truncated.  
Ruud de Rooij's avatar
Ruud de Rooij committed
274 275
 */

276
void
277
twofish_set_key(struct twofish_ctx *context,
278
		size_t keysize, const uint8_t *key)
Ruud de Rooij's avatar
Ruud de Rooij committed
279
{
280 281
  uint8_t key_copy[32];
  uint32_t m[8], s[4], t;
282
  int i, j, k;
Ruud de Rooij's avatar
Ruud de Rooij committed
283

284
  /* Extend key as necessary */
Ruud de Rooij's avatar
Ruud de Rooij committed
285

286
  assert(keysize <= 32);
287

288 289
  /* We do a little more copying than necessary, but that doesn't
   * really matter. */
290 291 292
  memset(key_copy, 0, 32);
  memcpy(key_copy, key, keysize);

293 294 295
  for (i = 0; i<8; i++)
    m[i] = LE_READ_UINT32(key_copy + i*4);
  
296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312
  if (keysize <= 16)
    k = 2;
  else if (keysize <= 24)
    k = 3;
  else
    k = 4;

  /* Compute sub-keys */

  for (i = 0; i < 20; i++)
    {
      t = h(k, 2*i+1, m[1], m[3], m[5], m[7]);
      t = rol8(t);
      t += (context->keys[2*i] =
	    t + h(k, 2*i, m[0], m[2], m[4], m[6]));
      t = rol9(t);
      context->keys[2*i+1] = t;
Ruud de Rooij's avatar
Ruud de Rooij committed
313 314
    }

315
  /* Compute key-dependent S-boxes */
Ruud de Rooij's avatar
Ruud de Rooij committed
316

317 318
  for (i = 0; i < k; i++)
    s[k-1-i] = compute_s(m[2*i], m[2*i+1]);
Ruud de Rooij's avatar
Ruud de Rooij committed
319

320 321 322 323 324 325 326
  for (i = 0; i < 4; i++)
    for (j = 0; j < 256; j++)
      context->s_box[i][j] = h_byte(k, i, j,
				    s[0] >> (i*8),
				    s[1] >> (i*8),
				    s[2] >> (i*8),
				    s[3] >> (i*8));
Ruud de Rooij's avatar
Ruud de Rooij committed
327 328
}

329
/* Encrypt blocks of 16 bytes of data with the twofish algorithm.
Ruud de Rooij's avatar
Ruud de Rooij committed
330
 *
331
 * Before this function can be used, twofish_set_key() must be used in order to
Ruud de Rooij's avatar
Ruud de Rooij committed
332
 * set up various tables required for the encryption algorithm.
333
 * 
Ruud de Rooij's avatar
Ruud de Rooij committed
334 335 336 337 338 339
 * This function always encrypts 16 bytes of plaintext to 16 bytes of
 * ciphertext.  The memory areas of the plaintext and the ciphertext can
 * overlap.
 */

void
340
twofish_encrypt(const struct twofish_ctx *context,
341
		size_t length,
342 343
		uint8_t *ciphertext,
		const uint8_t *plaintext)
Ruud de Rooij's avatar
Ruud de Rooij committed
344
{
345 346
  const uint32_t * keys        = context->keys;
  const uint32_t (*s_box)[256] = context->s_box;
347 348

  assert( !(length % TWOFISH_BLOCK_SIZE) );
349
  for ( ; length; length -= TWOFISH_BLOCK_SIZE)
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
    {  
      uint32_t words[4];
      uint32_t r0, r1, r2, r3, t0, t1;
      int i;

      for (i = 0; i<4; i++, plaintext += 4)
	words[i] = LE_READ_UINT32(plaintext);

      r0 = words[0] ^ keys[0];
      r1 = words[1] ^ keys[1];
      r2 = words[2] ^ keys[2];
      r3 = words[3] ^ keys[3];
  
      for (i = 0; i < 8; i++) {
	t1 = (  s_box[1][r1 & 0xFF]
		^ s_box[2][(r1 >> 8) & 0xFF]
		^ s_box[3][(r1 >> 16) & 0xFF]
		^ s_box[0][(r1 >> 24) & 0xFF]);
	t0 = (  s_box[0][r0 & 0xFF]
		^ s_box[1][(r0 >> 8) & 0xFF]
		^ s_box[2][(r0 >> 16) & 0xFF]
		^ s_box[3][(r0 >> 24) & 0xFF]) + t1;
	r3 = (t1 + t0 + keys[4*i+9]) ^ rol1(r3);
	r2 = (t0 + keys[4*i+8]) ^ r2;
	r2 = ror1(r2);

	t1 = (  s_box[1][r3 & 0xFF]
		^ s_box[2][(r3 >> 8) & 0xFF]
		^ s_box[3][(r3 >> 16) & 0xFF]
		^ s_box[0][(r3 >> 24) & 0xFF]);
	t0 = (  s_box[0][r2 & 0xFF]
		^ s_box[1][(r2 >> 8) & 0xFF]
		^ s_box[2][(r2 >> 16) & 0xFF]
		^ s_box[3][(r2 >> 24) & 0xFF]) + t1;
	r1 = (t1 + t0 + keys[4*i+11]) ^ rol1(r1);
	r0 = (t0 + keys[4*i+10]) ^ r0;
	r0 = ror1(r0);
      }

      words[0] = r2 ^ keys[4];
      words[1] = r3 ^ keys[5];
      words[2] = r0 ^ keys[6];
      words[3] = r1 ^ keys[7];

      for (i = 0; i<4; i++, ciphertext += 4)
	LE_WRITE_UINT32(ciphertext, words[i]);
    }
Ruud de Rooij's avatar
Ruud de Rooij committed
397 398
}

399
/* Decrypt blocks of 16 bytes of data with the twofish algorithm.
Ruud de Rooij's avatar
Ruud de Rooij committed
400
 *
401
 * Before this function can be used, twofish_set_key() must be used in order to
Ruud de Rooij's avatar
Ruud de Rooij committed
402
 * set up various tables required for the decryption algorithm.
403
 * 
Ruud de Rooij's avatar
Ruud de Rooij committed
404 405 406 407 408 409
 * This function always decrypts 16 bytes of ciphertext to 16 bytes of
 * plaintext.  The memory areas of the plaintext and the ciphertext can
 * overlap.
 */

void
410
twofish_decrypt(const struct twofish_ctx *context,
411
		size_t length,
412 413 414
		uint8_t *plaintext,
		const uint8_t *ciphertext)

Ruud de Rooij's avatar
Ruud de Rooij committed
415
{
416 417
  const uint32_t *keys  = context->keys;
  const uint32_t (*s_box)[256] = context->s_box;
418 419

  assert( !(length % TWOFISH_BLOCK_SIZE) );
420
  for ( ; length; length -= TWOFISH_BLOCK_SIZE)
421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
    {  
      uint32_t words[4];
      uint32_t r0, r1, r2, r3, t0, t1;
      int i;

      for (i = 0; i<4; i++, ciphertext += 4)
	words[i] = LE_READ_UINT32(ciphertext);

      r0 = words[2] ^ keys[6];
      r1 = words[3] ^ keys[7];
      r2 = words[0] ^ keys[4];
      r3 = words[1] ^ keys[5];

      for (i = 0; i < 8; i++) {
	t1 = (  s_box[1][r3 & 0xFF]
		^ s_box[2][(r3 >> 8) & 0xFF]
		^ s_box[3][(r3 >> 16) & 0xFF]
		^ s_box[0][(r3 >> 24) & 0xFF]);
	t0 = (  s_box[0][r2 & 0xFF]
		^ s_box[1][(r2 >> 8) & 0xFF]
		^ s_box[2][(r2 >> 16) & 0xFF]
		^ s_box[3][(r2 >> 24) & 0xFF]) + t1;
	r1 = (t1 + t0 + keys[39-4*i]) ^ r1;
	r1 = ror1(r1);
	r0 = (t0 + keys[38-4*i]) ^ rol1(r0);

	t1 = (  s_box[1][r1 & 0xFF]
		^ s_box[2][(r1 >> 8) & 0xFF]
		^ s_box[3][(r1 >> 16) & 0xFF]
		^ s_box[0][(r1 >> 24) & 0xFF]);
	t0 = (  s_box[0][r0 & 0xFF]
		^ s_box[1][(r0 >> 8) & 0xFF]
		^ s_box[2][(r0 >> 16) & 0xFF]
		^ s_box[3][(r0 >> 24) & 0xFF]) + t1;
	r3 = (t1 + t0 + keys[37-4*i]) ^ r3;
	r3 = ror1(r3);
	r2 = (t0 + keys[36-4*i]) ^ rol1(r2);
      }

      words[0] = r0 ^ keys[0];
      words[1] = r1 ^ keys[1];
      words[2] = r2 ^ keys[2];
      words[3] = r3 ^ keys[3];

      for (i = 0; i<4; i++, plaintext += 4)
	LE_WRITE_UINT32(plaintext, words[i]);
    }
Ruud de Rooij's avatar
Ruud de Rooij committed
468
}