![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to twofish.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [truecrypt] / truecrypt / crypto|
Return to twofish.c CVS log | Up to [truecrypt] / truecrypt / crypto |
1.1 root 1: /* This is an independent implementation of the encryption algorithm: */
2: /* */
3: /* Twofish by Bruce Schneier and colleagues */
4: /* */
5: /* which is a candidate algorithm in the Advanced Encryption Standard */
6: /* programme of the US National Institute of Standards and Technology. */
7: /* */
8: /* Copyright in this implementation is held by Dr B R Gladman but I */
9: /* hereby give permission for its free direct or derivative use subject */
10: /* to acknowledgment of its origin and compliance with any conditions */
11: /* that the originators of the algorithm place on its exploitation. */
12: /* */
13: /* My thanks to Doug Whiting and Niels Ferguson for comments that led */
14: /* to improvements in this implementation. */
15: /* */
16: /* Dr Brian Gladman ([email protected]) 14th January 1999 */
17:
1.1.1.2 root 18: /* Adapted for TrueCrypt by the TrueCrypt Foundation */
1.1 root 19:
1.1.1.2 root 20: #include "Twofish.h"
1.1.1.4 ! root 21: #include "Common/Endian.h"
1.1 root 22:
23: #define Q_TABLES
24: #define M_TABLE
25: #define MK_TABLE
26: #define ONE_STEP
27:
28: //u4byte k_len;
29: //u4byte l_key[40];
30: //u4byte s_key[4];
31:
32: /* finite field arithmetic for GF(2**8) with the modular */
33: /* polynomial x^8 + x^6 + x^5 + x^3 + 1 (0x169) */
34:
35: #define G_M 0x0169
36:
37: u1byte tab_5b[4] = { 0, G_M >> 2, G_M >> 1, (G_M >> 1) ^ (G_M >> 2) };
38: u1byte tab_ef[4] = { 0, (G_M >> 1) ^ (G_M >> 2), G_M >> 1, G_M >> 2 };
39:
40: #define ffm_01(x) (x)
41: #define ffm_5b(x) ((x) ^ ((x) >> 2) ^ tab_5b[(x) & 3])
42: #define ffm_ef(x) ((x) ^ ((x) >> 1) ^ ((x) >> 2) ^ tab_ef[(x) & 3])
43:
44: u1byte ror4[16] = { 0, 8, 1, 9, 2, 10, 3, 11, 4, 12, 5, 13, 6, 14, 7, 15 };
45: u1byte ashx[16] = { 0, 9, 2, 11, 4, 13, 6, 15, 8, 1, 10, 3, 12, 5, 14, 7 };
46:
47: u1byte qt0[2][16] =
48: { { 8, 1, 7, 13, 6, 15, 3, 2, 0, 11, 5, 9, 14, 12, 10, 4 },
49: { 2, 8, 11, 13, 15, 7, 6, 14, 3, 1, 9, 4, 0, 10, 12, 5 }
50: };
51:
52: u1byte qt1[2][16] =
53: { { 14, 12, 11, 8, 1, 2, 3, 5, 15, 4, 10, 6, 7, 0, 9, 13 },
54: { 1, 14, 2, 11, 4, 12, 3, 7, 6, 13, 10, 5, 15, 9, 0, 8 }
55: };
56:
57: u1byte qt2[2][16] =
58: { { 11, 10, 5, 14, 6, 13, 9, 0, 12, 8, 15, 3, 2, 4, 7, 1 },
59: { 4, 12, 7, 5, 1, 6, 9, 10, 0, 14, 13, 8, 2, 11, 3, 15 }
60: };
61:
62: u1byte qt3[2][16] =
63: { { 13, 7, 15, 4, 1, 2, 6, 14, 9, 11, 3, 0, 8, 5, 12, 10 },
64: { 11, 9, 5, 1, 12, 3, 13, 14, 6, 4, 7, 15, 2, 0, 8, 10 }
65: };
66:
67: static u1byte qp(const u4byte n, const u1byte x)
68: { u1byte a0, a1, a2, a3, a4, b0, b1, b2, b3, b4;
69:
70: a0 = x >> 4; b0 = x & 15;
71: a1 = a0 ^ b0; b1 = ror4[b0] ^ ashx[a0];
72: a2 = qt0[n][a1]; b2 = qt1[n][b1];
73: a3 = a2 ^ b2; b3 = ror4[b2] ^ ashx[a2];
74: a4 = qt2[n][a3]; b4 = qt3[n][b3];
75: return (b4 << 4) | a4;
76: };
77:
78: #ifdef Q_TABLES
79:
80: u4byte qt_gen = 0;
81: u1byte q_tab[2][256];
82:
83: #define q(n,x) q_tab[n][x]
84:
85: static void gen_qtab(void)
86: { u4byte i;
87:
88: for(i = 0; i < 256; ++i)
89: {
90: q(0,i) = qp(0, (u1byte)i);
91: q(1,i) = qp(1, (u1byte)i);
92: }
93: };
94:
95: #else
96:
97: #define q(n,x) qp(n, x)
98:
99: #endif
100:
101: #ifdef M_TABLE
102:
103: u4byte mt_gen = 0;
104: u4byte m_tab[4][256];
105:
106: static void gen_mtab(void)
107: { u4byte i, f01, f5b, fef;
108:
109: for(i = 0; i < 256; ++i)
110: {
111: f01 = q(1,i); f5b = ffm_5b(f01); fef = ffm_ef(f01);
112: m_tab[0][i] = f01 + (f5b << 8) + (fef << 16) + (fef << 24);
113: m_tab[2][i] = f5b + (fef << 8) + (f01 << 16) + (fef << 24);
114:
115: f01 = q(0,i); f5b = ffm_5b(f01); fef = ffm_ef(f01);
116: m_tab[1][i] = fef + (fef << 8) + (f5b << 16) + (f01 << 24);
117: m_tab[3][i] = f5b + (f01 << 8) + (fef << 16) + (f5b << 24);
118: }
119: };
120:
121: #define mds(n,x) m_tab[n][x]
122:
123: #else
124:
125: #define fm_00 ffm_01
126: #define fm_10 ffm_5b
127: #define fm_20 ffm_ef
128: #define fm_30 ffm_ef
129: #define q_0(x) q(1,x)
130:
131: #define fm_01 ffm_ef
132: #define fm_11 ffm_ef
133: #define fm_21 ffm_5b
134: #define fm_31 ffm_01
135: #define q_1(x) q(0,x)
136:
137: #define fm_02 ffm_5b
138: #define fm_12 ffm_ef
139: #define fm_22 ffm_01
140: #define fm_32 ffm_ef
141: #define q_2(x) q(1,x)
142:
143: #define fm_03 ffm_5b
144: #define fm_13 ffm_01
145: #define fm_23 ffm_ef
146: #define fm_33 ffm_5b
147: #define q_3(x) q(0,x)
148:
149: #define f_0(n,x) ((u4byte)fm_0##n(x))
150: #define f_1(n,x) ((u4byte)fm_1##n(x) << 8)
151: #define f_2(n,x) ((u4byte)fm_2##n(x) << 16)
152: #define f_3(n,x) ((u4byte)fm_3##n(x) << 24)
153:
154: #define mds(n,x) f_0(n,q_##n(x)) ^ f_1(n,q_##n(x)) ^ f_2(n,q_##n(x)) ^ f_3(n,q_##n(x))
155:
156: #endif
157:
158: static u4byte h_fun(TwofishInstance *instance, const u4byte x, const u4byte key[])
159: { u4byte b0, b1, b2, b3;
160:
161: #ifndef M_TABLE
162: u4byte m5b_b0, m5b_b1, m5b_b2, m5b_b3;
163: u4byte mef_b0, mef_b1, mef_b2, mef_b3;
164: #endif
165:
166: b0 = extract_byte(x, 0); b1 = extract_byte(x, 1); b2 = extract_byte(x, 2); b3 = extract_byte(x, 3);
167:
168: switch(instance->k_len)
169: {
170: case 4: b0 = q(1, b0) ^ extract_byte(key[3],0);
171: b1 = q(0, b1) ^ extract_byte(key[3],1);
172: b2 = q(0, b2) ^ extract_byte(key[3],2);
173: b3 = q(1, b3) ^ extract_byte(key[3],3);
174: case 3: b0 = q(1, b0) ^ extract_byte(key[2],0);
175: b1 = q(1, b1) ^ extract_byte(key[2],1);
176: b2 = q(0, b2) ^ extract_byte(key[2],2);
177: b3 = q(0, b3) ^ extract_byte(key[2],3);
178: case 2: b0 = q(0,q(0,b0) ^ extract_byte(key[1],0)) ^ extract_byte(key[0],0);
179: b1 = q(0,q(1,b1) ^ extract_byte(key[1],1)) ^ extract_byte(key[0],1);
180: b2 = q(1,q(0,b2) ^ extract_byte(key[1],2)) ^ extract_byte(key[0],2);
181: b3 = q(1,q(1,b3) ^ extract_byte(key[1],3)) ^ extract_byte(key[0],3);
182: }
183: #ifdef M_TABLE
184:
185: return mds(0, b0) ^ mds(1, b1) ^ mds(2, b2) ^ mds(3, b3);
186:
187: #else
188:
189: b0 = q(1, b0); b1 = q(0, b1); b2 = q(1, b2); b3 = q(0, b3);
190: m5b_b0 = ffm_5b(b0); m5b_b1 = ffm_5b(b1); m5b_b2 = ffm_5b(b2); m5b_b3 = ffm_5b(b3);
191: mef_b0 = ffm_ef(b0); mef_b1 = ffm_ef(b1); mef_b2 = ffm_ef(b2); mef_b3 = ffm_ef(b3);
192: b0 ^= mef_b1 ^ m5b_b2 ^ m5b_b3; b3 ^= m5b_b0 ^ mef_b1 ^ mef_b2;
193: b2 ^= mef_b0 ^ m5b_b1 ^ mef_b3; b1 ^= mef_b0 ^ mef_b2 ^ m5b_b3;
194:
195: return b0 | (b3 << 8) | (b2 << 16) | (b1 << 24);
196:
197: #endif
198: };
199:
200: #ifdef MK_TABLE
201:
202: #ifdef ONE_STEP
203: //u4byte mk_tab[4][256];
204: #else
205: u1byte sb[4][256];
206: #endif
207:
208: #define q20(x) q(0,q(0,x) ^ extract_byte(key[1],0)) ^ extract_byte(key[0],0)
209: #define q21(x) q(0,q(1,x) ^ extract_byte(key[1],1)) ^ extract_byte(key[0],1)
210: #define q22(x) q(1,q(0,x) ^ extract_byte(key[1],2)) ^ extract_byte(key[0],2)
211: #define q23(x) q(1,q(1,x) ^ extract_byte(key[1],3)) ^ extract_byte(key[0],3)
212:
213: #define q30(x) q(0,q(0,q(1, x) ^ extract_byte(key[2],0)) ^ extract_byte(key[1],0)) ^ extract_byte(key[0],0)
214: #define q31(x) q(0,q(1,q(1, x) ^ extract_byte(key[2],1)) ^ extract_byte(key[1],1)) ^ extract_byte(key[0],1)
215: #define q32(x) q(1,q(0,q(0, x) ^ extract_byte(key[2],2)) ^ extract_byte(key[1],2)) ^ extract_byte(key[0],2)
216: #define q33(x) q(1,q(1,q(0, x) ^ extract_byte(key[2],3)) ^ extract_byte(key[1],3)) ^ extract_byte(key[0],3)
217:
218: #define q40(x) q(0,q(0,q(1, q(1, x) ^ extract_byte(key[3],0)) ^ extract_byte(key[2],0)) ^ extract_byte(key[1],0)) ^ extract_byte(key[0],0)
219: #define q41(x) q(0,q(1,q(1, q(0, x) ^ extract_byte(key[3],1)) ^ extract_byte(key[2],1)) ^ extract_byte(key[1],1)) ^ extract_byte(key[0],1)
220: #define q42(x) q(1,q(0,q(0, q(0, x) ^ extract_byte(key[3],2)) ^ extract_byte(key[2],2)) ^ extract_byte(key[1],2)) ^ extract_byte(key[0],2)
221: #define q43(x) q(1,q(1,q(0, q(1, x) ^ extract_byte(key[3],3)) ^ extract_byte(key[2],3)) ^ extract_byte(key[1],3)) ^ extract_byte(key[0],3)
222:
1.1.1.3 root 223: static void gen_mk_tab(TwofishInstance *instance, u4byte key[])
1.1 root 224: { u4byte i;
225: u1byte by;
226:
1.1.1.2 root 227: // u4byte *l_key = instance->l_key;
228: // u4byte *s_key = instance->s_key;
1.1 root 229: u4byte *mk_tab = instance->mk_tab;
230:
231: switch(instance->k_len)
232: {
233: case 2: for(i = 0; i < 256; ++i)
234: {
235: by = (u1byte)i;
236: #ifdef ONE_STEP
237: mk_tab[0 + 4*i] = mds(0, q20(by)); mk_tab[1 + 4*i] = mds(1, q21(by));
238: mk_tab[2 + 4*i] = mds(2, q22(by)); mk_tab[3 + 4*i] = mds(3, q23(by));
239: #else
240: sb[0][i] = q20(by); sb[1][i] = q21(by);
241: sb[2][i] = q22(by); sb[3][i] = q23(by);
242: #endif
243: }
244: break;
245:
246: case 3: for(i = 0; i < 256; ++i)
247: {
248: by = (u1byte)i;
249: #ifdef ONE_STEP
250: mk_tab[0 + 4*i] = mds(0, q30(by)); mk_tab[1 + 4*i] = mds(1, q31(by));
251: mk_tab[2 + 4*i] = mds(2, q32(by)); mk_tab[3 + 4*i] = mds(3, q33(by));
252: #else
253: sb[0][i] = q30(by); sb[1][i] = q31(by);
254: sb[2][i] = q32(by); sb[3][i] = q33(by);
255: #endif
256: }
257: break;
258:
259: case 4: for(i = 0; i < 256; ++i)
260: {
261: by = (u1byte)i;
262: #ifdef ONE_STEP
263: mk_tab[0 + 4*i] = mds(0, q40(by)); mk_tab[1 + 4*i] = mds(1, q41(by));
264: mk_tab[2 + 4*i] = mds(2, q42(by)); mk_tab[3 + 4*i] = mds(3, q43(by));
265: #else
266: sb[0][i] = q40(by); sb[1][i] = q41(by);
267: sb[2][i] = q42(by); sb[3][i] = q43(by);
268: #endif
269: }
270: }
271: };
272:
273: # ifdef ONE_STEP
274: # define g0_fun(x) ( mk_tab[0 + 4*extract_byte(x,0)] ^ mk_tab[1 + 4*extract_byte(x,1)] \
275: ^ mk_tab[2 + 4*extract_byte(x,2)] ^ mk_tab[3 + 4*extract_byte(x,3)] )
276: # define g1_fun(x) ( mk_tab[0 + 4*extract_byte(x,3)] ^ mk_tab[1 + 4*extract_byte(x,0)] \
277: ^ mk_tab[2 + 4*extract_byte(x,1)] ^ mk_tab[3 + 4*extract_byte(x,2)] )
278:
279:
280: # else
281: # define g0_fun(x) ( mds(0, sb[0][extract_byte(x,0)]) ^ mds(1, sb[1][extract_byte(x,1)]) \
282: ^ mds(2, sb[2][extract_byte(x,2)]) ^ mds(3, sb[3][extract_byte(x,3)]) )
283: # define g1_fun(x) ( mds(0, sb[0][extract_byte(x,3)]) ^ mds(1, sb[1][extract_byte(x,0)]) \
284: ^ mds(2, sb[2][extract_byte(x,1)]) ^ mds(3, sb[3][extract_byte(x,2)]) )
285: # endif
286:
287: #else
288:
289: #define g0_fun(x) h_fun(instance, x,s_key)
290: #define g1_fun(x) h_fun(instance, rotl(x,8),s_key)
291:
292: #endif
293:
294: /* The (12,8) Reed Soloman code has the generator polynomial
295:
296: g(x) = x^4 + (a + 1/a) * x^3 + a * x^2 + (a + 1/a) * x + 1
297:
298: where the coefficients are in the finite field GF(2^8) with a
299: modular polynomial a^8 + a^6 + a^3 + a^2 + 1. To generate the
300: remainder we have to start with a 12th order polynomial with our
301: eight input bytes as the coefficients of the 4th to 11th terms.
302: That is:
303:
304: m[7] * x^11 + m[6] * x^10 ... + m[0] * x^4 + 0 * x^3 +... + 0
305:
306: We then multiply the generator polynomial by m[7] * x^7 and subtract
307: it - xor in GF(2^8) - from the above to eliminate the x^7 term (the
308: artihmetic on the coefficients is done in GF(2^8). We then multiply
309: the generator polynomial by x^6 * coeff(x^10) and use this to remove
310: the x^10 term. We carry on in this way until the x^4 term is removed
311: so that we are left with:
312:
313: r[3] * x^3 + r[2] * x^2 + r[1] 8 x^1 + r[0]
314:
315: which give the resulting 4 bytes of the remainder. This is equivalent
316: to the matrix multiplication in the Twofish description but much faster
317: to implement.
318:
319: */
320:
321: #define G_MOD 0x0000014d
322:
323: static u4byte mds_rem(u4byte p0, u4byte p1)
324: { u4byte i, t, u;
325:
326: for(i = 0; i < 8; ++i)
327: {
328: t = p1 >> 24; // get most significant coefficient
329:
330: p1 = (p1 << 8) | (p0 >> 24); p0 <<= 8; // shift others up
331:
332: // multiply t by a (the primitive element - i.e. left shift)
333:
334: u = (t << 1);
335:
336: if(t & 0x80) // subtract modular polynomial on overflow
337:
338: u ^= G_MOD;
339:
340: p1 ^= t ^ (u << 16); // remove t * (a * x^2 + 1)
341:
342: u ^= (t >> 1); // form u = a * t + t / a = t * (a + 1 / a);
343:
344: if(t & 0x01) // add the modular polynomial on underflow
345:
346: u ^= G_MOD >> 1;
347:
348: p1 ^= (u << 24) | (u << 8); // remove t * (a + 1/a) * (x^3 + x)
349: }
350:
351: return p1;
352: };
353:
354: /* initialise the key schedule from the user supplied key */
355:
356: u4byte *twofish_set_key(TwofishInstance *instance, const u4byte in_key[], const u4byte key_len)
357: { u4byte i, a, b, me_key[4], mo_key[4];
358: u4byte *l_key, *s_key;
359:
360: instance->l_key = (u4byte *) ((__int8 *)instance + sizeof (TwofishInstance));
361: instance->s_key = (u4byte *) ((__int8 *)instance + sizeof (TwofishInstance) + TF_L_KEY_SIZE);
362: instance->mk_tab = (u4byte *) ((__int8 *)instance + sizeof (TwofishInstance) + TF_L_KEY_SIZE + TF_S_KEY_SIZE);
363:
364: l_key = instance->l_key;
365: s_key = instance->s_key;
366:
367: #ifdef Q_TABLES
368: if(!qt_gen)
369: {
370: gen_qtab(); qt_gen = 1;
371: }
372: #endif
373:
374: #ifdef M_TABLE
375: if(!mt_gen)
376: {
377: gen_mtab(); mt_gen = 1;
378: }
379: #endif
380:
381: instance->k_len = key_len / 64; /* 2, 3 or 4 */
382:
383: for(i = 0; i < instance->k_len; ++i)
384: {
1.1.1.2 root 385: a = LE32(in_key[i + i]); me_key[i] = a;
386: b = LE32(in_key[i + i + 1]); mo_key[i] = b;
1.1 root 387: s_key[instance->k_len - i - 1] = mds_rem(a, b);
388: }
389:
390: for(i = 0; i < 40; i += 2)
391: {
392: a = 0x01010101 * i; b = a + 0x01010101;
393: a = h_fun(instance, a, me_key);
394: b = rotl(h_fun(instance, b, mo_key), 8);
395: l_key[i] = a + b;
396: l_key[i + 1] = rotl(a + 2 * b, 9);
397: }
398:
399: #ifdef MK_TABLE
400: gen_mk_tab(instance, s_key);
401: #endif
402:
403: return l_key;
404: };
405:
406: /* encrypt a block of text */
407:
408: #define f_rnd(i) \
409: t1 = g1_fun(blk[1]); t0 = g0_fun(blk[0]); \
410: blk[2] = rotr(blk[2] ^ (t0 + t1 + l_key[4 * (i) + 8]), 1); \
411: blk[3] = rotl(blk[3], 1) ^ (t0 + 2 * t1 + l_key[4 * (i) + 9]); \
412: t1 = g1_fun(blk[3]); t0 = g0_fun(blk[2]); \
413: blk[0] = rotr(blk[0] ^ (t0 + t1 + l_key[4 * (i) + 10]), 1); \
414: blk[1] = rotl(blk[1], 1) ^ (t0 + 2 * t1 + l_key[4 * (i) + 11])
415:
416: void twofish_encrypt(TwofishInstance *instance, const u4byte in_blk[4], u4byte out_blk[])
417: { u4byte t0, t1, blk[4];
418:
419: u4byte *l_key = instance->l_key;
1.1.1.2 root 420: // u4byte *s_key = instance->s_key;
1.1 root 421: u4byte *mk_tab = instance->mk_tab;
422:
1.1.1.2 root 423: blk[0] = LE32(in_blk[0]) ^ l_key[0];
424: blk[1] = LE32(in_blk[1]) ^ l_key[1];
425: blk[2] = LE32(in_blk[2]) ^ l_key[2];
426: blk[3] = LE32(in_blk[3]) ^ l_key[3];
1.1 root 427:
428: f_rnd(0); f_rnd(1); f_rnd(2); f_rnd(3);
429: f_rnd(4); f_rnd(5); f_rnd(6); f_rnd(7);
430:
1.1.1.2 root 431: out_blk[0] = LE32(blk[2] ^ l_key[4]);
432: out_blk[1] = LE32(blk[3] ^ l_key[5]);
433: out_blk[2] = LE32(blk[0] ^ l_key[6]);
434: out_blk[3] = LE32(blk[1] ^ l_key[7]);
1.1 root 435: };
436:
437: /* decrypt a block of text */
438:
439: #define i_rnd(i) \
440: t1 = g1_fun(blk[1]); t0 = g0_fun(blk[0]); \
441: blk[2] = rotl(blk[2], 1) ^ (t0 + t1 + l_key[4 * (i) + 10]); \
442: blk[3] = rotr(blk[3] ^ (t0 + 2 * t1 + l_key[4 * (i) + 11]), 1); \
443: t1 = g1_fun(blk[3]); t0 = g0_fun(blk[2]); \
444: blk[0] = rotl(blk[0], 1) ^ (t0 + t1 + l_key[4 * (i) + 8]); \
445: blk[1] = rotr(blk[1] ^ (t0 + 2 * t1 + l_key[4 * (i) + 9]), 1)
446:
447: void twofish_decrypt(TwofishInstance *instance, const u4byte in_blk[4], u4byte out_blk[4])
448: { u4byte t0, t1, blk[4];
449:
450: u4byte *l_key = instance->l_key;
1.1.1.2 root 451: // u4byte *s_key = instance->s_key;
1.1 root 452: u4byte *mk_tab = instance->mk_tab;
453:
1.1.1.2 root 454: blk[0] = LE32(in_blk[0]) ^ l_key[4];
455: blk[1] = LE32(in_blk[1]) ^ l_key[5];
456: blk[2] = LE32(in_blk[2]) ^ l_key[6];
457: blk[3] = LE32(in_blk[3]) ^ l_key[7];
1.1 root 458:
459: i_rnd(7); i_rnd(6); i_rnd(5); i_rnd(4);
460: i_rnd(3); i_rnd(2); i_rnd(1); i_rnd(0);
461:
1.1.1.2 root 462: out_blk[0] = LE32(blk[2] ^ l_key[0]);
463: out_blk[1] = LE32(blk[3] ^ l_key[1]);
464: out_blk[2] = LE32(blk[0] ^ l_key[2]);
465: out_blk[3] = LE32(blk[1] ^ l_key[3]);
1.1 root 466: };
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.