1
0
mirror of https://github.com/hashcat/hashcat.git synced 2024-11-16 04:49:24 +00:00
hashcat/OpenCL/m25400-pure.cl

445 lines
8.3 KiB
Common Lisp
Raw Normal View History

/**
* Author......: See docs/credits.txt
* License.....: MIT
*/
// TODO use user password as input for md5 of o_digest if no owner password is set
// TODO dynamically add user password including padding to the RC4 input for the computation of the pdf o-value
#ifdef KERNEL_STATIC
#include "inc_vendor.h"
#include "inc_types.h"
#include "inc_platform.cl"
#include "inc_common.cl"
#include "inc_hash_md5.cl"
#endif
#define COMPARE_S "inc_comp_single.cl"
#define COMPARE_M "inc_comp_multi.cl"
CONSTANT_VK u32a padding[8] =
{
0x5e4ebf28,
0x418a754e,
0x564e0064,
0x0801faff,
0xb6002e2e,
0x803e68d0,
0xfea90c2f,
0x7a695364
};
typedef struct pdf
{
int V;
int R;
int P;
int enc_md;
u32 id_buf[8];
u32 u_buf[32];
u32 o_buf[32];
int id_len;
int o_len;
int u_len;
u32 rc4key[2];
u32 rc4data[2];
} pdf_t;
typedef struct pdf14_tmp
{
u32 digest[4];
u32 out[4];
} pdf14_tmp_t;
typedef struct
{
u8 S[256];
u32 wtf_its_faster;
} RC4_KEY;
DECLSPEC void swap (LOCAL_AS RC4_KEY *rc4_key, const u8 i, const u8 j)
{
u8 tmp;
tmp = rc4_key->S[i];
rc4_key->S[i] = rc4_key->S[j];
rc4_key->S[j] = tmp;
}
DECLSPEC void rc4_init_16 (LOCAL_AS RC4_KEY *rc4_key, const u32 *data)
{
u32 v = 0x03020100;
u32 a = 0x04040404;
LOCAL_AS u32 *ptr = (LOCAL_AS u32 *) rc4_key->S;
#ifdef _unroll
#pragma unroll
#endif
for (u32 i = 0; i < 64; i++)
{
*ptr++ = v; v += a;
}
u32 j = 0;
#ifdef _unroll
#pragma unroll
#endif
for (u32 i = 0; i < 16; i++)
{
u32 idx = i * 16;
u32 v;
v = data[0];
j += rc4_key->S[idx] + (v >> 0); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 8); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
v = data[1];
j += rc4_key->S[idx] + (v >> 0); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 8); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
v = data[2];
j += rc4_key->S[idx] + (v >> 0); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 8); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
v = data[3];
j += rc4_key->S[idx] + (v >> 0); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 8); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
}
}
DECLSPEC u8 rc4_next_16 (LOCAL_AS RC4_KEY *rc4_key, u8 i, u8 j, const u32 *in, u32 *out)
{
#ifdef _unroll
#pragma unroll
#endif
for (u32 k = 0; k < 4; k++)
{
u32 xor4 = 0;
u8 idx;
i += 1;
j += rc4_key->S[i];
swap (rc4_key, i, j);
idx = rc4_key->S[i] + rc4_key->S[j];
xor4 |= rc4_key->S[idx] << 0;
i += 1;
j += rc4_key->S[i];
swap (rc4_key, i, j);
idx = rc4_key->S[i] + rc4_key->S[j];
xor4 |= rc4_key->S[idx] << 8;
i += 1;
j += rc4_key->S[i];
swap (rc4_key, i, j);
idx = rc4_key->S[i] + rc4_key->S[j];
xor4 |= rc4_key->S[idx] << 16;
i += 1;
j += rc4_key->S[i];
swap (rc4_key, i, j);
idx = rc4_key->S[i] + rc4_key->S[j];
xor4 |= rc4_key->S[idx] << 24;
out[k] = in[k] ^ xor4;
}
return j;
}
KERNEL_FQ void m25400_init (KERN_ATTR_TMPS_ESALT (pdf14_tmp_t, pdf_t))
{
/**
* base
*/
const u64 gid = get_global_id (0);
//const u64 lid = get_local_id (0);
if (gid >= gid_max) return;
u32 w0[4];
w0[0] = pws[gid].i[ 0];
w0[1] = pws[gid].i[ 1];
w0[2] = pws[gid].i[ 2];
w0[3] = pws[gid].i[ 3];
u32 w1[4];
w1[0] = pws[gid].i[ 4];
w1[1] = pws[gid].i[ 5];
w1[2] = pws[gid].i[ 6];
w1[3] = pws[gid].i[ 7];
const u32 pw_len = pws[gid].pw_len;
/**
* shared
*/
//LOCAL_AS RC4_KEY rc4_keys[64];
//LOCAL_AS RC4_KEY *rc4_key = &rc4_keys[lid];
u32 P = esalt_bufs[DIGESTS_OFFSET].P;
u32 id_buf[12];
id_buf[ 0] = esalt_bufs[DIGESTS_OFFSET].id_buf[0];
id_buf[ 1] = esalt_bufs[DIGESTS_OFFSET].id_buf[1];
id_buf[ 2] = esalt_bufs[DIGESTS_OFFSET].id_buf[2];
id_buf[ 3] = esalt_bufs[DIGESTS_OFFSET].id_buf[3];
id_buf[ 4] = esalt_bufs[DIGESTS_OFFSET].id_buf[4];
id_buf[ 5] = esalt_bufs[DIGESTS_OFFSET].id_buf[5];
id_buf[ 6] = esalt_bufs[DIGESTS_OFFSET].id_buf[6];
id_buf[ 7] = esalt_bufs[DIGESTS_OFFSET].id_buf[7];
id_buf[ 8] = 0;
id_buf[ 9] = 0;
id_buf[10] = 0;
id_buf[11] = 0;
u32 rc4data[2];
rc4data[0] = padding[0];
rc4data[1] = padding[1];
/**
* main init
*/
u32 w0_t[4];
u32 w1_t[4];
u32 w2_t[4];
u32 w3_t[4];
// max length supported by pdf11 is 32
w0_t[0] = padding[0];
w0_t[1] = padding[1];
w0_t[2] = padding[2];
w0_t[3] = padding[3];
w1_t[0] = padding[4];
w1_t[1] = padding[5];
w1_t[2] = padding[6];
w1_t[3] = padding[7];
w2_t[0] = 0;
w2_t[1] = 0;
w2_t[2] = 0;
w2_t[3] = 0;
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = 0;
w3_t[3] = 0;
switch_buffer_by_offset_le (w0_t, w1_t, w2_t, w3_t, pw_len);
// add password
// truncate at 32 is wanted, not a bug!
// add padding
w0_t[0] |= w0[0];
w0_t[1] |= w0[1];
w0_t[2] |= w0[2];
w0_t[3] |= w0[3];
w1_t[0] |= w1[0];
w1_t[1] |= w1[1];
w1_t[2] |= w1[2];
w1_t[3] |= w1[3];
w2_t[0] = 0x80;
w2_t[1] = 0;
w2_t[2] = 0;
w2_t[3] = 0;
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = 32 * 8;
w3_t[3] = 0;
u32 digest[4];
digest[0] = MD5M_A;
digest[1] = MD5M_B;
digest[2] = MD5M_C;
digest[3] = MD5M_D;
md5_transform (w0_t, w1_t, w2_t, w3_t, digest);
tmps[gid].digest[0] = digest[0];
tmps[gid].digest[1] = digest[1];
tmps[gid].digest[2] = digest[2];
tmps[gid].digest[3] = digest[3];
tmps[gid].out[0] = rc4data[0];
tmps[gid].out[1] = rc4data[1];
tmps[gid].out[2] = 0;
tmps[gid].out[3] = 0;
}
KERNEL_FQ void m25400_loop (KERN_ATTR_TMPS_ESALT (pdf14_tmp_t, pdf_t))
{
/**
* base
*/
const u64 gid = get_global_id (0);
const u64 lid = get_local_id (0);
if (gid >= gid_max) return;
/**
* shared
*/
LOCAL_VK RC4_KEY rc4_keys[64];
LOCAL_AS RC4_KEY *rc4_key = &rc4_keys[lid];
/**
* loop
*/
u32 digest[4];
digest[0] = tmps[gid].digest[0];
digest[1] = tmps[gid].digest[1];
digest[2] = tmps[gid].digest[2];
digest[3] = tmps[gid].digest[3];
u32 out[4];
out[0] = tmps[gid].out[0];
out[1] = tmps[gid].out[1];
out[2] = tmps[gid].out[2];
out[3] = tmps[gid].out[3];
for (u32 i = 0, j = loop_pos; i < loop_cnt; i++, j++)
{
if (j < 50)
{
u32 w0_t[4];
u32 w1_t[4];
u32 w2_t[4];
u32 w3_t[4];
w0_t[0] = digest[0];
w0_t[1] = digest[1];
w0_t[2] = digest[2];
w0_t[3] = digest[3];
w1_t[0] = 0x80;
w1_t[1] = 0;
w1_t[2] = 0;
w1_t[3] = 0;
w2_t[0] = 0;
w2_t[1] = 0;
w2_t[2] = 0;
w2_t[3] = 0;
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = 16 * 8;
w3_t[3] = 0;
digest[0] = MD5M_A;
digest[1] = MD5M_B;
digest[2] = MD5M_C;
digest[3] = MD5M_D;
md5_transform (w0_t, w1_t, w2_t, w3_t, digest);
}
else
{
const u32 x = j - 50;
const u32 xv = x << 0
| x << 8
| x << 16
| x << 24;
u32 tmp[4];
tmp[0] = digest[0] ^ xv;
tmp[1] = digest[1] ^ xv;
tmp[2] = digest[2] ^ xv;
tmp[3] = digest[3] ^ xv;
rc4_init_16 (rc4_key, tmp);
rc4_next_16 (rc4_key, 0, 0, out, out);
}
}
tmps[gid].digest[0] = digest[0];
tmps[gid].digest[1] = digest[1];
tmps[gid].digest[2] = digest[2];
tmps[gid].digest[3] = digest[3];
tmps[gid].out[0] = out[0];
tmps[gid].out[1] = out[1];
tmps[gid].out[2] = out[2];
tmps[gid].out[3] = out[3];
}
KERNEL_FQ void m25400_comp (KERN_ATTR_TMPS_ESALT (pdf14_tmp_t, pdf_t))
{
/**
* modifier
*/
const u64 gid = get_global_id (0);
if (gid >= gid_max) return;
const u64 lid = get_local_id (0);
/**
* digest
*/
const u32 r0 = tmps[gid].out[0];
const u32 r1 = tmps[gid].out[1];
const u32 r2 = 0;
const u32 r3 = 0;
#define il_pos 0
#ifdef KERNEL_STATIC
#include COMPARE_M
#endif
}