/* This is an independent implementation of the encryption algorithm: */ /* */ /* Twofish by Bruce Schneier and colleagues */ /* */ /* which is a candidate algorithm in the Advanced Encryption Standard */ /* programme of the US National Institute of Standards and Technology. */ /* */ /* Copyright in this implementation is held by Dr B R Gladman but I */ /* hereby give permission for its free direct or derivative use subject */ /* to acknowledgment of its origin and compliance with any conditions */ /* that the originators of the algorithm place on its exploitation. */ /* */ /* My thanks to Doug Whiting and Niels Ferguson for comments that led */ /* to improvements in this implementation. */ /* */ /* Dr Brian Gladman 14th January 1999 */ /* Timing data for Twofish (twofish.c) 128 bit key: Key Setup: 8414 cycles Encrypt: 376 cycles = 68.1 mbits/sec Decrypt: 374 cycles = 68.4 mbits/sec Mean: 375 cycles = 68.3 mbits/sec 192 bit key: Key Setup: 11628 cycles Encrypt: 376 cycles = 68.1 mbits/sec Decrypt: 374 cycles = 68.4 mbits/sec Mean: 375 cycles = 68.3 mbits/sec 256 bit key: Key Setup: 15457 cycles Encrypt: 381 cycles = 67.2 mbits/sec Decrypt: 374 cycles = 68.4 mbits/sec Mean: 378 cycles = 67.8 mbits/sec */ #include "../std_defs.h" #define Q_TABLES #define M_TABLE #define MK_TABLE #define ONE_STEP static char *alg_name[] = { "twofish", "twofish.c", "twofish" }; char **cipher_name() { return alg_name; } u4byte k_len; u4byte l_key[40]; u4byte s_key[4]; /* finite field arithmetic for GF(2**8) with the modular */ /* polynomial x^8 + x^6 + x^5 + x^3 + 1 (0x169) */ #define G_M 0x0169 u1byte tab_5b[4] = { 0, G_M >> 2, G_M >> 1, (G_M >> 1) ^ (G_M >> 2) }; u1byte tab_ef[4] = { 0, (G_M >> 1) ^ (G_M >> 2), G_M >> 1, G_M >> 2 }; #define ffm_01(x) (x) #define ffm_5b(x) ((x) ^ ((x) >> 2) ^ tab_5b[(x) & 3]) #define ffm_ef(x) ((x) ^ ((x) >> 1) ^ ((x) >> 2) ^ tab_ef[(x) & 3]) u1byte ror4[16] = { 0, 8, 1, 9, 2, 10, 3, 11, 4, 12, 5, 13, 6, 14, 7, 15 }; u1byte ashx[16] = { 0, 9, 2, 11, 4, 13, 6, 15, 8, 1, 10, 3, 12, 5, 14, 7 }; u1byte qt0[2][16] = { { 8, 1, 7, 13, 6, 15, 3, 2, 0, 11, 5, 9, 14, 12, 10, 4 }, { 2, 8, 11, 13, 15, 7, 6, 14, 3, 1, 9, 4, 0, 10, 12, 5 } }; u1byte qt1[2][16] = { { 14, 12, 11, 8, 1, 2, 3, 5, 15, 4, 10, 6, 7, 0, 9, 13 }, { 1, 14, 2, 11, 4, 12, 3, 7, 6, 13, 10, 5, 15, 9, 0, 8 } }; u1byte qt2[2][16] = { { 11, 10, 5, 14, 6, 13, 9, 0, 12, 8, 15, 3, 2, 4, 7, 1 }, { 4, 12, 7, 5, 1, 6, 9, 10, 0, 14, 13, 8, 2, 11, 3, 15 } }; u1byte qt3[2][16] = { { 13, 7, 15, 4, 1, 2, 6, 14, 9, 11, 3, 0, 8, 5, 12, 10 }, { 11, 9, 5, 1, 12, 3, 13, 14, 6, 4, 7, 15, 2, 0, 8, 10 } }; u1byte qp(const u4byte n, const u1byte x) { u1byte a0, a1, a2, a3, a4, b0, b1, b2, b3, b4; a0 = x >> 4; b0 = x & 15; a1 = a0 ^ b0; b1 = ror4[b0] ^ ashx[a0]; a2 = qt0[n][a1]; b2 = qt1[n][b1]; a3 = a2 ^ b2; b3 = ror4[b2] ^ ashx[a2]; a4 = qt2[n][a3]; b4 = qt3[n][b3]; return (b4 << 4) | a4; }; #ifdef Q_TABLES u4byte qt_gen = 0; u1byte q_tab[2][256]; #define q(n,x) q_tab[n][x] void gen_qtab(void) { u4byte i; for(i = 0; i < 256; ++i) { q(0,i) = qp(0, (u1byte)i); q(1,i) = qp(1, (u1byte)i); } }; #else #define q(n,x) qp(n, x) #endif #ifdef M_TABLE u4byte mt_gen = 0; u4byte m_tab[4][256]; void gen_mtab(void) { u4byte i, f01, f5b, fef; for(i = 0; i < 256; ++i) { f01 = q(1,i); f5b = ffm_5b(f01); fef = ffm_ef(f01); m_tab[0][i] = f01 + (f5b << 8) + (fef << 16) + (fef << 24); m_tab[2][i] = f5b + (fef << 8) + (f01 << 16) + (fef << 24); f01 = q(0,i); f5b = ffm_5b(f01); fef = ffm_ef(f01); m_tab[1][i] = fef + (fef << 8) + (f5b << 16) + (f01 << 24); m_tab[3][i] = f5b + (f01 << 8) + (fef << 16) + (f5b << 24); } }; #define mds(n,x) m_tab[n][x] #else #define fm_00 ffm_01 #define fm_10 ffm_5b #define fm_20 ffm_ef #define fm_30 ffm_ef #define q_0(x) q(1,x) #define fm_01 ffm_ef #define fm_11 ffm_ef #define fm_21 ffm_5b #define fm_31 ffm_01 #define q_1(x) q(0,x) #define fm_02 ffm_5b #define fm_12 ffm_ef #define fm_22 ffm_01 #define fm_32 ffm_ef #define q_2(x) q(1,x) #define fm_03 ffm_5b #define fm_13 ffm_01 #define fm_23 ffm_ef #define fm_33 ffm_5b #define q_3(x) q(0,x) #define f_0(n,x) ((u4byte)fm_0##n(x)) #define f_1(n,x) ((u4byte)fm_1##n(x) << 8) #define f_2(n,x) ((u4byte)fm_2##n(x) << 16) #define f_3(n,x) ((u4byte)fm_3##n(x) << 24) #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)) #endif u4byte h_fun(const u4byte x, const u4byte key[]) { u4byte b0, b1, b2, b3; #ifndef M_TABLE u4byte m5b_b0, m5b_b1, m5b_b2, m5b_b3; u4byte mef_b0, mef_b1, mef_b2, mef_b3; #endif b0 = byte(x, 0); b1 = byte(x, 1); b2 = byte(x, 2); b3 = byte(x, 3); switch(k_len) { case 4: b0 = q(1, b0) ^ byte(key[3],0); b1 = q(0, b1) ^ byte(key[3],1); b2 = q(0, b2) ^ byte(key[3],2); b3 = q(1, b3) ^ byte(key[3],3); case 3: b0 = q(1, b0) ^ byte(key[2],0); b1 = q(1, b1) ^ byte(key[2],1); b2 = q(0, b2) ^ byte(key[2],2); b3 = q(0, b3) ^ byte(key[2],3); case 2: b0 = q(0,q(0,b0) ^ byte(key[1],0)) ^ byte(key[0],0); b1 = q(0,q(1,b1) ^ byte(key[1],1)) ^ byte(key[0],1); b2 = q(1,q(0,b2) ^ byte(key[1],2)) ^ byte(key[0],2); b3 = q(1,q(1,b3) ^ byte(key[1],3)) ^ byte(key[0],3); } #ifdef M_TABLE return mds(0, b0) ^ mds(1, b1) ^ mds(2, b2) ^ mds(3, b3); #else b0 = q(1, b0); b1 = q(0, b1); b2 = q(1, b2); b3 = q(0, b3); m5b_b0 = ffm_5b(b0); m5b_b1 = ffm_5b(b1); m5b_b2 = ffm_5b(b2); m5b_b3 = ffm_5b(b3); mef_b0 = ffm_ef(b0); mef_b1 = ffm_ef(b1); mef_b2 = ffm_ef(b2); mef_b3 = ffm_ef(b3); b0 ^= mef_b1 ^ m5b_b2 ^ m5b_b3; b3 ^= m5b_b0 ^ mef_b1 ^ mef_b2; b2 ^= mef_b0 ^ m5b_b1 ^ mef_b3; b1 ^= mef_b0 ^ mef_b2 ^ m5b_b3; return b0 | (b3 << 8) | (b2 << 16) | (b1 << 24); #endif }; #ifdef MK_TABLE #ifdef ONE_STEP u4byte mk_tab[4][256]; #else u1byte sb[4][256]; #endif #define q20(x) q(0,q(0,x) ^ byte(key[1],0)) ^ byte(key[0],0) #define q21(x) q(0,q(1,x) ^ byte(key[1],1)) ^ byte(key[0],1) #define q22(x) q(1,q(0,x) ^ byte(key[1],2)) ^ byte(key[0],2) #define q23(x) q(1,q(1,x) ^ byte(key[1],3)) ^ byte(key[0],3) #define q30(x) q(0,q(0,q(1, x) ^ byte(key[2],0)) ^ byte(key[1],0)) ^ byte(key[0],0) #define q31(x) q(0,q(1,q(1, x) ^ byte(key[2],1)) ^ byte(key[1],1)) ^ byte(key[0],1) #define q32(x) q(1,q(0,q(0, x) ^ byte(key[2],2)) ^ byte(key[1],2)) ^ byte(key[0],2) #define q33(x) q(1,q(1,q(0, x) ^ byte(key[2],3)) ^ byte(key[1],3)) ^ byte(key[0],3) #define q40(x) q(0,q(0,q(1, q(1, x) ^ byte(key[3],0)) ^ byte(key[2],0)) ^ byte(key[1],0)) ^ byte(key[0],0) #define q41(x) q(0,q(1,q(1, q(0, x) ^ byte(key[3],1)) ^ byte(key[2],1)) ^ byte(key[1],1)) ^ byte(key[0],1) #define q42(x) q(1,q(0,q(0, q(0, x) ^ byte(key[3],2)) ^ byte(key[2],2)) ^ byte(key[1],2)) ^ byte(key[0],2) #define q43(x) q(1,q(1,q(0, q(1, x) ^ byte(key[3],3)) ^ byte(key[2],3)) ^ byte(key[1],3)) ^ byte(key[0],3) gen_mk_tab(u4byte key[]) { u4byte i; u1byte by; switch(k_len) { case 2: for(i = 0; i < 256; ++i) { by = (u1byte)i; #ifdef ONE_STEP mk_tab[0][i] = mds(0, q20(by)); mk_tab[1][i] = mds(1, q21(by)); mk_tab[2][i] = mds(2, q22(by)); mk_tab[3][i] = mds(3, q23(by)); #else sb[0][i] = q20(by); sb[1][i] = q21(by); sb[2][i] = q22(by); sb[3][i] = q23(by); #endif } break; case 3: for(i = 0; i < 256; ++i) { by = (u1byte)i; #ifdef ONE_STEP mk_tab[0][i] = mds(0, q30(by)); mk_tab[1][i] = mds(1, q31(by)); mk_tab[2][i] = mds(2, q32(by)); mk_tab[3][i] = mds(3, q33(by)); #else sb[0][i] = q30(by); sb[1][i] = q31(by); sb[2][i] = q32(by); sb[3][i] = q33(by); #endif } break; case 4: for(i = 0; i < 256; ++i) { by = (u1byte)i; #ifdef ONE_STEP mk_tab[0][i] = mds(0, q40(by)); mk_tab[1][i] = mds(1, q41(by)); mk_tab[2][i] = mds(2, q42(by)); mk_tab[3][i] = mds(3, q43(by)); #else sb[0][i] = q40(by); sb[1][i] = q41(by); sb[2][i] = q42(by); sb[3][i] = q43(by); #endif } } }; # ifdef ONE_STEP # define g0_fun(x) ( mk_tab[0][byte(x,0)] ^ mk_tab[1][byte(x,1)] \ ^ mk_tab[2][byte(x,2)] ^ mk_tab[3][byte(x,3)] ) # define g1_fun(x) ( mk_tab[0][byte(x,3)] ^ mk_tab[1][byte(x,0)] \ ^ mk_tab[2][byte(x,1)] ^ mk_tab[3][byte(x,2)] ) # else # define g0_fun(x) ( mds(0, sb[0][byte(x,0)]) ^ mds(1, sb[1][byte(x,1)]) \ ^ mds(2, sb[2][byte(x,2)]) ^ mds(3, sb[3][byte(x,3)]) ) # define g1_fun(x) ( mds(0, sb[0][byte(x,3)]) ^ mds(1, sb[1][byte(x,0)]) \ ^ mds(2, sb[2][byte(x,1)]) ^ mds(3, sb[3][byte(x,2)]) ) # endif #else #define g0_fun(x) h_fun(x,s_key) #define g1_fun(x) h_fun(rotl(x,8),s_key) #endif /* The (12,8) Reed Soloman code has the generator polynomial g(x) = x^4 + (a + 1/a) * x^3 + a * x^2 + (a + 1/a) * x + 1 where the coefficients are in the finite field GF(2^8) with a modular polynomial a^8 + a^6 + a^3 + a^2 + 1. To generate the remainder we have to start with a 12th order polynomial with our eight input bytes as the coefficients of the 4th to 11th terms. That is: m[7] * x^11 + m[6] * x^10 ... + m[0] * x^4 + 0 * x^3 +... + 0 We then multiply the generator polynomial by m[7] * x^7 and subtract it - xor in GF(2^8) - from the above to eliminate the x^7 term (the artihmetic on the coefficients is done in GF(2^8). We then multiply the generator polynomial by x^6 * coeff(x^10) and use this to remove the x^10 term. We carry on in this way until the x^4 term is removed so that we are left with: r[3] * x^3 + r[2] * x^2 + r[1] 8 x^1 + r[0] which give the resulting 4 bytes of the remainder. This is equivalent to the matrix multiplication in the Twofish description but much faster to implement. */ #define G_MOD 0x0000014d u4byte mds_rem(u4byte p0, u4byte p1) { u4byte i, t, u; for(i = 0; i < 8; ++i) { t = p1 >> 24; // get most significant coefficient p1 = (p1 << 8) | (p0 >> 24); p0 <<= 8; // shift others up // multiply t by a (the primitive element - i.e. left shift) u = (t << 1); if(t & 0x80) // subtract modular polynomial on overflow u ^= G_MOD; p1 ^= t ^ (u << 16); // remove t * (a * x^2 + 1) u ^= (t >> 1); // form u = a * t + t / a = t * (a + 1 / a); if(t & 0x01) // add the modular polynomial on underflow u ^= G_MOD >> 1; p1 ^= (u << 24) | (u << 8); // remove t * (a + 1/a) * (x^3 + x) } return p1; }; /* initialise the key schedule from the user supplied key */ u4byte *set_key(const u4byte in_key[], const u4byte key_len) { u4byte i, a, b, me_key[4], mo_key[4]; #ifdef Q_TABLES if(!qt_gen) { gen_qtab(); qt_gen = 1; } #endif #ifdef M_TABLE if(!mt_gen) { gen_mtab(); mt_gen = 1; } #endif k_len = key_len / 64; /* 2, 3 or 4 */ for(i = 0; i < k_len; ++i) { a = in_key[i + i]; me_key[i] = a; b = in_key[i + i + 1]; mo_key[i] = b; s_key[k_len - i - 1] = mds_rem(a, b); } for(i = 0; i < 40; i += 2) { a = 0x01010101 * i; b = a + 0x01010101; a = h_fun(a, me_key); b = rotl(h_fun(b, mo_key), 8); l_key[i] = a + b; l_key[i + 1] = rotl(a + 2 * b, 9); } #ifdef MK_TABLE gen_mk_tab(s_key); #endif return l_key; }; /* encrypt a block of text */ #define f_rnd(i) \ t1 = g1_fun(blk[1]); t0 = g0_fun(blk[0]); \ blk[2] = rotr(blk[2] ^ (t0 + t1 + l_key[4 * (i) + 8]), 1); \ blk[3] = rotl(blk[3], 1) ^ (t0 + 2 * t1 + l_key[4 * (i) + 9]); \ t1 = g1_fun(blk[3]); t0 = g0_fun(blk[2]); \ blk[0] = rotr(blk[0] ^ (t0 + t1 + l_key[4 * (i) + 10]), 1); \ blk[1] = rotl(blk[1], 1) ^ (t0 + 2 * t1 + l_key[4 * (i) + 11]) void encrypt(const u4byte in_blk[4], u4byte out_blk[]) { u4byte t0, t1, blk[4]; blk[0] = in_blk[0] ^ l_key[0]; blk[1] = in_blk[1] ^ l_key[1]; blk[2] = in_blk[2] ^ l_key[2]; blk[3] = in_blk[3] ^ l_key[3]; f_rnd(0); f_rnd(1); f_rnd(2); f_rnd(3); f_rnd(4); f_rnd(5); f_rnd(6); f_rnd(7); out_blk[0] = blk[2] ^ l_key[4]; out_blk[1] = blk[3] ^ l_key[5]; out_blk[2] = blk[0] ^ l_key[6]; out_blk[3] = blk[1] ^ l_key[7]; }; /* decrypt a block of text */ #define i_rnd(i) \ t1 = g1_fun(blk[1]); t0 = g0_fun(blk[0]); \ blk[2] = rotl(blk[2], 1) ^ (t0 + t1 + l_key[4 * (i) + 10]); \ blk[3] = rotr(blk[3] ^ (t0 + 2 * t1 + l_key[4 * (i) + 11]), 1); \ t1 = g1_fun(blk[3]); t0 = g0_fun(blk[2]); \ blk[0] = rotl(blk[0], 1) ^ (t0 + t1 + l_key[4 * (i) + 8]); \ blk[1] = rotr(blk[1] ^ (t0 + 2 * t1 + l_key[4 * (i) + 9]), 1) void decrypt(const u4byte in_blk[4], u4byte out_blk[4]) { u4byte t0, t1, blk[4]; blk[0] = in_blk[0] ^ l_key[4]; blk[1] = in_blk[1] ^ l_key[5]; blk[2] = in_blk[2] ^ l_key[6]; blk[3] = in_blk[3] ^ l_key[7]; i_rnd(7); i_rnd(6); i_rnd(5); i_rnd(4); i_rnd(3); i_rnd(2); i_rnd(1); i_rnd(0); out_blk[0] = blk[2] ^ l_key[0]; out_blk[1] = blk[3] ^ l_key[1]; out_blk[2] = blk[0] ^ l_key[2]; out_blk[3] = blk[1] ^ l_key[3]; };