/** * Author......: See docs/credits.txt * License.....: MIT */ //shared mem too small //#define NEW_SIMD_CODE #define XSTR(x) #x #define STR(x) XSTR(x) #ifdef KERNEL_STATIC #include STR(INCLUDE_PATH/inc_vendor.h) #include STR(INCLUDE_PATH/inc_types.h) #include STR(INCLUDE_PATH/inc_platform.cl) #include STR(INCLUDE_PATH/inc_common.cl) #include STR(INCLUDE_PATH/inc_hash_md4.cl) #include STR(INCLUDE_PATH/inc_hash_md5.cl) #include STR(INCLUDE_PATH/inc_cipher_rc4.cl) #endif typedef struct krb5asrep { u32 account_info[512]; u32 checksum[4]; u32 edata2[5120]; u32 edata2_len; } krb5asrep_t; DECLSPEC int decrypt_and_check (LOCAL_AS u32 *S, PRIVATE_AS u32 *data, GLOBAL_AS const u32 *edata2, const u32 edata2_len, PRIVATE_AS const u32 *K2, PRIVATE_AS const u32 *checksum, const u64 lid) { rc4_init_128 (S, data, lid); u32 out0[4]; /* 8 first bytes are nonce, then ASN1 structs (DER encoding: TLV) The first byte is always 0x79 (01 1 11001, where 01 = "class=APPLICATION", 1 = "form=constructed", 11001 is application type 25) The next byte is the length: if length < 128 bytes: length is on 1 byte, and the next byte is 0x30 (class=SEQUENCE) else if length <= 256: length is on 2 bytes, the first byte is 0x81, and the third byte is 0x30 (class=SEQUENCE) else if length > 256: length is on 3 bytes, the first byte is 0x82, and the fourth byte is 0x30 (class=SEQUENCE) */ rc4_next_16_global (S, 0, 0, edata2 + 0, out0, lid); if (((out0[2] & 0x00ff80ff) != 0x00300079) && ((out0[2] & 0xFF00FFFF) != 0x30008179) && ((out0[2] & 0x0000FFFF) != 0x00008279 || (out0[3] & 0x000000FF) != 0x00000030)) return 0; rc4_init_128 (S, data, lid); u8 i = 0; u8 j = 0; // init hmac u32 w0[4]; u32 w1[4]; u32 w2[4]; u32 w3[4]; w0[0] = K2[0]; w0[1] = K2[1]; w0[2] = K2[2]; w0[3] = K2[3]; w1[0] = 0; w1[1] = 0; w1[2] = 0; w1[3] = 0; w2[0] = 0; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = 0; md5_hmac_ctx_t ctx; md5_hmac_init_64 (&ctx, w0, w1, w2, w3); int edata2_left; for (edata2_left = edata2_len; edata2_left >= 64; edata2_left -= 64) { j = rc4_next_16_global (S, i, j, edata2, w0, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w1, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w2, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w3, lid); i += 16; edata2 += 4; md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 64); } w0[0] = 0; w0[1] = 0; w0[2] = 0; w0[3] = 0; w1[0] = 0; w1[1] = 0; w1[2] = 0; w1[3] = 0; w2[0] = 0; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = 0; if (edata2_left < 16) { j = rc4_next_16_global (S, i, j, edata2, w0, lid); i += 16; edata2 += 4; truncate_block_4x4_le_S (w0, edata2_left & 0xf); } else if (edata2_left < 32) { j = rc4_next_16_global (S, i, j, edata2, w0, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w1, lid); i += 16; edata2 += 4; truncate_block_4x4_le_S (w1, edata2_left & 0xf); } else if (edata2_left < 48) { j = rc4_next_16_global (S, i, j, edata2, w0, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w1, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w2, lid); i += 16; edata2 += 4; truncate_block_4x4_le_S (w2, edata2_left & 0xf); } else { j = rc4_next_16_global (S, i, j, edata2, w0, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w1, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w2, lid); i += 16; edata2 += 4; j = rc4_next_16_global (S, i, j, edata2, w3, lid); i += 16; edata2 += 4; truncate_block_4x4_le_S (w3, edata2_left & 0xf); } md5_hmac_update_64 (&ctx, w0, w1, w2, w3, edata2_left); md5_hmac_final (&ctx); if (checksum[0] != ctx.opad.h[0]) return 0; if (checksum[1] != ctx.opad.h[1]) return 0; if (checksum[2] != ctx.opad.h[2]) return 0; if (checksum[3] != ctx.opad.h[3]) return 0; return 1; } DECLSPEC void kerb_prepare (PRIVATE_AS const u32 *K, PRIVATE_AS const u32 *checksum, PRIVATE_AS u32 *digest, PRIVATE_AS u32 *K2) { // K1=MD5_HMAC(K,1); with 1 encoded as little indian on 4 bytes (01000000 in hexa); u32 w0[4]; u32 w1[4]; u32 w2[4]; u32 w3[4]; w0[0] = K[0]; w0[1] = K[1]; w0[2] = K[2]; w0[3] = K[3]; w1[0] = 0; w1[1] = 0; w1[2] = 0; w1[3] = 0; w2[0] = 0; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = 0; md5_hmac_ctx_t ctx1; md5_hmac_init_64 (&ctx1, w0, w1, w2, w3); w0[0] = 8; w0[1] = 0; w0[2] = 0; w0[3] = 0; w1[0] = 0; w1[1] = 0; w1[2] = 0; w1[3] = 0; w2[0] = 0; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = 0; md5_hmac_update_64 (&ctx1, w0, w1, w2, w3, 4); md5_hmac_final (&ctx1); w0[0] = ctx1.opad.h[0]; w0[1] = ctx1.opad.h[1]; w0[2] = ctx1.opad.h[2]; w0[3] = ctx1.opad.h[3]; w1[0] = 0; w1[1] = 0; w1[2] = 0; w1[3] = 0; w2[0] = 0; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = 0; md5_hmac_ctx_t ctx; md5_hmac_init_64 (&ctx, w0, w1, w2, w3); w0[0] = checksum[0]; w0[1] = checksum[1]; w0[2] = checksum[2]; w0[3] = checksum[3]; w1[0] = 0; w1[1] = 0; w1[2] = 0; w1[3] = 0; w2[0] = 0; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = 0; md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 16); md5_hmac_final (&ctx); digest[0] = ctx.opad.h[0]; digest[1] = ctx.opad.h[1]; digest[2] = ctx.opad.h[2]; digest[3] = ctx.opad.h[3]; K2[0] = ctx1.opad.h[0]; K2[1] = ctx1.opad.h[1]; K2[2] = ctx1.opad.h[2]; K2[3] = ctx1.opad.h[3]; } KERNEL_FQ void m18200_mxx (KERN_ATTR_VECTOR_ESALT (krb5asrep_t)) { /** * modifier */ const u64 lid = get_local_id (0); const u64 gid = get_global_id (0); if (gid >= GID_CNT) return; /** * base */ const u32 pw_len = pws[gid].pw_len; u32x w[64] = { 0 }; for (u32 i = 0, idx = 0; i < pw_len; i += 4, idx += 1) { w[idx] = pws[gid].i[idx]; } LOCAL_VK u32 S[64 * FIXED_LOCAL_SIZE]; u32 checksum[4]; checksum[0] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[0]; checksum[1] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[1]; checksum[2] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[2]; checksum[3] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[3]; /** * loop */ u32x w0l = w[0]; for (u32 il_pos = 0; il_pos < IL_CNT; il_pos += VECT_SIZE) { const u32x w0r = words_buf_r[il_pos / VECT_SIZE]; const u32x w0 = w0l | w0r; w[0] = w0; md4_ctx_t ctx; md4_init (&ctx); md4_update_utf16le (&ctx, w, pw_len); md4_final (&ctx); u32 digest[4]; u32 K2[4]; kerb_prepare (ctx.h, checksum, digest, K2); if (decrypt_and_check (S, digest, esalt_bufs[DIGESTS_OFFSET_HOST].edata2, esalt_bufs[DIGESTS_OFFSET_HOST].edata2_len, K2, checksum, lid) == 1) { if (hc_atomic_inc (&hashes_shown[DIGESTS_OFFSET_HOST]) == 0) { mark_hash (plains_buf, d_return_buf, SALT_POS_HOST, DIGESTS_CNT, 0, DIGESTS_OFFSET_HOST + 0, gid, il_pos, 0, 0); } } } } KERNEL_FQ void m18200_sxx (KERN_ATTR_VECTOR_ESALT (krb5asrep_t)) { /** * modifier */ const u64 lid = get_local_id (0); const u64 gid = get_global_id (0); if (gid >= GID_CNT) return; /** * base */ const u32 pw_len = pws[gid].pw_len; u32x w[64] = { 0 }; for (u32 i = 0, idx = 0; i < pw_len; i += 4, idx += 1) { w[idx] = pws[gid].i[idx]; } LOCAL_VK u32 S[64 * FIXED_LOCAL_SIZE]; u32 checksum[4]; checksum[0] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[0]; checksum[1] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[1]; checksum[2] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[2]; checksum[3] = esalt_bufs[DIGESTS_OFFSET_HOST].checksum[3]; /** * loop */ u32x w0l = w[0]; for (u32 il_pos = 0; il_pos < IL_CNT; il_pos += VECT_SIZE) { const u32x w0r = words_buf_r[il_pos / VECT_SIZE]; const u32x w0 = w0l | w0r; w[0] = w0; md4_ctx_t ctx; md4_init (&ctx); md4_update_utf16le (&ctx, w, pw_len); md4_final (&ctx); u32 digest[4]; u32 K2[4]; kerb_prepare (ctx.h, checksum, digest, K2); if (decrypt_and_check (S, digest, esalt_bufs[DIGESTS_OFFSET_HOST].edata2, esalt_bufs[DIGESTS_OFFSET_HOST].edata2_len, K2, checksum, lid) == 1) { if (hc_atomic_inc (&hashes_shown[DIGESTS_OFFSET_HOST]) == 0) { mark_hash (plains_buf, d_return_buf, SALT_POS_HOST, DIGESTS_CNT, 0, DIGESTS_OFFSET_HOST + 0, gid, il_pos, 0, 0); } } } }