This commit is contained in:
dela
2026-02-26 06:14:36 +08:00
commit 06b67f60d6
20 changed files with 1440 additions and 0 deletions

3
.gitignore vendored Normal file
View File

@@ -0,0 +1,3 @@
/target
CLAUDE.md
/docs

114
Cargo.lock generated Normal file
View File

@@ -0,0 +1,114 @@
# This file is automatically @generated by Cargo.
# It is not intended for manual editing.
version = 4
[[package]]
name = "base64"
version = "0.22.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "72b3254f16251a8381aa12e40e3c4d2f0199f8c6508fbecb9d91f575e0fbb8c6"
[[package]]
name = "hcaptcha-pow"
version = "0.1.0"
dependencies = [
"base64",
"serde",
"serde_json",
]
[[package]]
name = "itoa"
version = "1.0.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "92ecc6618181def0457392ccd0ee51198e065e016d1d527a7ac1b6dc7c1f09d2"
[[package]]
name = "memchr"
version = "2.8.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f8ca58f447f06ed17d5fc4043ce1b10dd205e060fb3ce5b979b8ed8e59ff3f79"
[[package]]
name = "proc-macro2"
version = "1.0.106"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8fd00f0bb2e90d81d1044c2b32617f68fcb9fa3bb7640c23e9c748e53fb30934"
dependencies = [
"unicode-ident",
]
[[package]]
name = "quote"
version = "1.0.44"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "21b2ebcf727b7760c461f091f9f0f539b77b8e87f2fd88131e7f1b433b3cece4"
dependencies = [
"proc-macro2",
]
[[package]]
name = "serde"
version = "1.0.228"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9a8e94ea7f378bd32cbbd37198a4a91436180c5bb472411e48b5ec2e2124ae9e"
dependencies = [
"serde_core",
"serde_derive",
]
[[package]]
name = "serde_core"
version = "1.0.228"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "41d385c7d4ca58e59fc732af25c3983b67ac852c1a25000afe1175de458b67ad"
dependencies = [
"serde_derive",
]
[[package]]
name = "serde_derive"
version = "1.0.228"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d540f220d3187173da220f885ab66608367b6574e925011a9353e4badda91d79"
dependencies = [
"proc-macro2",
"quote",
"syn",
]
[[package]]
name = "serde_json"
version = "1.0.149"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "83fc039473c5595ace860d8c4fafa220ff474b3fc6bfdb4293327f1a37e94d86"
dependencies = [
"itoa",
"memchr",
"serde",
"serde_core",
"zmij",
]
[[package]]
name = "syn"
version = "2.0.117"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e665b8803e7b1d2a727f4023456bbbbe74da67099c585258af0ad9c5013b9b99"
dependencies = [
"proc-macro2",
"quote",
"unicode-ident",
]
[[package]]
name = "unicode-ident"
version = "1.0.24"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e6e4313cd5fcd3dad5cafa179702e2b244f760991f45397d14d4ebf38247da75"
[[package]]
name = "zmij"
version = "1.0.21"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b8848ee67ecc8aedbaf3e4122217aff892639231befc6a1b58d29fff4c2cabaa"

9
Cargo.toml Normal file
View File

@@ -0,0 +1,9 @@
[package]
name = "hcaptcha-pow"
version = "0.1.0"
edition = "2021"
[dependencies]
base64 = "0.22"
serde = { version = "1", features = ["derive"] }
serde_json = "1"

74
src/gift256/encrypt.rs Normal file
View File

@@ -0,0 +1,74 @@
//! GIFT-256 fixsliced encryption (13 rounds)
//! Corresponds to gift256_encrypt_fixsliced @ 0x80048469
//!
//! Round structure: 4 cycles x {SBox+P1+KeyXOR, SBox+DA+KeyXOR, SBox+P2+KeyXOR, SBox+DB+KeyXOR}
//! Last cycle only executes the first type (SBox+P1+KeyXOR)
use super::sbox::sbox_bitsliced;
use super::linear::{linear_p1, linear_p2, diffusion_a, diffusion_b};
use super::interleave::{pack_input, nibble_deinterleave, unpack_output};
/// GIFT-256 encryption.
/// input: 256-bit plaintext (8 u32, little-endian)
/// rk: 120 u32 round keys (from key_schedule)
/// output: 256-bit ciphertext (8 u32)
pub fn encrypt(input: &[u32; 8], rk: &[u32; 120]) -> [u32; 8] {
// === Input packing (bit-interleave + initial round key XOR) ===
let initial_rk: [u32; 8] = rk[0..8].try_into().unwrap();
let mut s = pack_input(input, &initial_rk);
// === 13 rounds main loop (4 cycles) ===
// Round key indexing:
// Initial: rk[0..7]
// Type A: rk[base+8 .. base+15]
// Type DA: rk[base+16 .. base+23]
// Type P2: rk[base+24 .. base+31]
// Type DB: rk[base+32 .. base+39]
// where base = cycle * 32
for cycle in 0..4 {
let base = cycle * 32;
// --- Round Type A: SBox + P1 + KeyXOR ---
sbox_bitsliced(&mut s);
linear_p1(&mut s);
let rk_a: [u32; 8] = rk[base + 8..base + 16].try_into().unwrap();
for j in 0..8 {
s[j] ^= rk_a[j];
}
if cycle == 3 {
break; // Last cycle only has Type A
}
// --- Round Type DA: SBox + nibble-ROT diffusion + KeyXOR ---
sbox_bitsliced(&mut s);
let rk_da: [u32; 8] = rk[base + 16..base + 24].try_into().unwrap();
diffusion_a(&mut s, &rk_da);
// --- Round Type P2: SBox + P2 + KeyXOR ---
sbox_bitsliced(&mut s);
linear_p2(&mut s);
let rk_p2: [u32; 8] = rk[base + 24..base + 32].try_into().unwrap();
for j in 0..8 {
s[j] ^= rk_p2[j];
}
// --- Round Type DB: SBox + byte-ROT8 diffusion + KeyXOR ---
sbox_bitsliced(&mut s);
let rk_db: [u32; 8] = rk[base + 32..base + 40].try_into().unwrap();
diffusion_b(&mut s, &rk_db);
}
// === Output deinterleave ===
for j in 0..8 {
s[j] = nibble_deinterleave(s[j]);
}
// Final SBox
sbox_bitsliced(&mut s);
// Final round key XOR + deinterleave (rk[112..119])
let final_rk: [u32; 8] = rk[112..120].try_into().unwrap();
unpack_output(&s, &final_rk)
}

190
src/gift256/interleave.rs Normal file
View File

@@ -0,0 +1,190 @@
//! GIFT-256 fixsliced bit interleave/deinterleave
//! Used to convert between standard and fixsliced representations
//! Corresponds to gift256_encrypt_fixsliced State 0 (input) and State 3 (output)
/// Pack 8 u32 (standard) into 8 u32 (fixsliced) with initial round key XOR.
/// Corresponds to gift256_encrypt_fixsliced State 0.
///
/// Input read order (from decompiled offsets):
/// input[7]=@0x1c, input[3]=@0x0c, input[6]=@0x18, input[2]=@0x08
/// input[5]=@0x14, input[1]=@0x04, input[4]=@0x10, input[0]=@0x00
pub fn pack_input(input: &[u32; 8], initial_rk: &[u32; 8]) -> [u32; 8] {
let (mut a, mut b, mut c, mut d, mut e, mut f, mut g, mut h) =
(input[7], input[3], input[6], input[2], input[5], input[1], input[4], input[0]);
// Phase 1: 0x55555555 interleave (odd/even bit separation)
let t0 = (a ^ (b >> 1)) & 0x55555555;
a ^= t0;
b ^= t0 << 1;
let t1 = (c ^ (d >> 1)) & 0x55555555;
c ^= t1;
d ^= t1 << 1;
let t2 = (e ^ (f >> 1)) & 0x55555555;
e ^= t2;
f ^= t2 << 1;
let t3 = (g ^ (h >> 1)) & 0x55555555;
g ^= t3;
h ^= t3 << 1;
// Phase 2: 0x33333333 interleave (2-bit group separation)
let t4 = (a ^ (c >> 2)) & 0x33333333;
a ^= t4;
c ^= t4 << 2;
let t5 = (e ^ (g >> 2)) & 0x33333333;
e ^= t5;
g ^= t5 << 2;
let t6 = (b ^ (d >> 2)) & 0x33333333;
b ^= t6;
d ^= t6 << 2;
let t7 = (f ^ (h >> 2)) & 0x33333333;
f ^= t7;
h ^= t7 << 2;
// Phase 3: 0x0F0F0F0F interleave (nibble separation) + initial round key XOR
let t8 = (a ^ (e >> 4)) & 0x0F0F0F0F;
let out7 = initial_rk[7] ^ (a ^ t8);
let out3 = initial_rk[3] ^ ((t8 << 4) ^ e);
let t9 = (c ^ (g >> 4)) & 0x0F0F0F0F;
let out5 = initial_rk[5] ^ (c ^ t9);
let out1 = initial_rk[1] ^ ((t9 << 4) ^ g);
let t10 = (b ^ (f >> 4)) & 0x0F0F0F0F;
let out6 = initial_rk[6] ^ (b ^ t10);
let out2 = initial_rk[2] ^ ((t10 << 4) ^ f);
let t11 = (d ^ (h >> 4)) & 0x0F0F0F0F;
let out4 = initial_rk[4] ^ (d ^ t11);
let out0 = initial_rk[0] ^ ((t11 << 4) ^ h);
[out0, out1, out2, out3, out4, out5, out6, out7]
}
/// Nibble deinterleave.
/// Decompiled: ((uVar4 >> 4 ^ uVar4) & 0xf000f00) * 0x11 ^ uVar4
pub fn nibble_deinterleave(x: u32) -> u32 {
(((x >> 4) ^ x) & 0x0F000F00).wrapping_mul(0x11) ^ x
}
/// Key schedule deinterleave type A (offsets 0x20..0x3C, 0x60..0x7C, 0x1A0..0x1BC)
/// x = ((x ^ x>>4) & 0x030F0C00) * 0x11 ^ x;
/// x = ((x>>2 ^ x) & 0x33003300) * 5 ^ x;
pub fn key_deinterleave_a(x: u32) -> u32 {
let x = (((x ^ (x >> 4)) & 0x030F0C00).wrapping_mul(0x11)) ^ x;
((((x >> 2) ^ x) & 0x33003300).wrapping_mul(5)) ^ x
}
/// Key schedule deinterleave type B (offsets 0x40..0x5C)
/// x = ((x>>4 ^ x) & 0x0F000F00) * 0x11 ^ x;
pub fn key_deinterleave_b(x: u32) -> u32 {
((((x >> 4) ^ x) & 0x0F000F00).wrapping_mul(0x11)) ^ x
}
/// Key schedule deinterleave type C (offsets 0x60..0x7C second group)
/// x = ((x ^ x>>4) & 0x0C0F0300) * 0x11 ^ x;
/// x = ((x>>2 ^ x) & 0x33003300) * 5 ^ x;
pub fn key_deinterleave_c(x: u32) -> u32 {
let x = (((x ^ (x >> 4)) & 0x0C0F0300).wrapping_mul(0x11)) ^ x;
((((x >> 2) ^ x) & 0x33003300).wrapping_mul(5)) ^ x
}
/// Interleave a half-key (4 u32 input -> 8 u32 output).
/// Same 3-stage butterfly as pack_input but on 4 words.
/// Used in key_schedule to interleave the two key halves.
pub fn interleave_key_half(input: &[u32], output: &mut [u32]) {
let (mut a, mut b) = (input[3], input[1]);
let (mut c, mut d) = (input[2], input[0]);
// Phase 1: 0x55555555
let t0 = (a ^ (b >> 1)) & 0x55555555;
a ^= t0;
b ^= t0 << 1;
let t1 = (c ^ (d >> 1)) & 0x55555555;
c ^= t1;
d ^= t1 << 1;
// Phase 2: 0x33333333
let t2 = (a ^ (c >> 2)) & 0x33333333;
a ^= t2;
c ^= t2 << 2;
let t3 = (b ^ (d >> 2)) & 0x33333333;
b ^= t3;
d ^= t3 << 2;
// Phase 3: 0x0F0F0F0F
let t4 = (a ^ (b >> 4)) & 0x0F0F0F0F;
output[7] = a ^ t4;
output[3] = (t4 << 4) ^ b;
let t5 = (c ^ (d >> 4)) & 0x0F0F0F0F;
output[5] = c ^ t5;
output[1] = (t5 << 4) ^ d;
// For the remaining outputs, use the already-interleaved values
let t6 = (a ^ (c >> 4)) & 0x0F0F0F0F;
output[6] = a ^ t6;
output[2] = (t6 << 4) ^ c;
let t7 = (b ^ (d >> 4)) & 0x0F0F0F0F;
output[4] = b ^ t7;
output[0] = (t7 << 4) ^ d;
}
/// Unpack output: reverse bit-interleave + final round key XOR.
/// Reverse of pack_input: apply masks in reverse order (0x0F -> 0x33 -> 0x55).
pub fn unpack_output(s: &[u32; 8], final_rk: &[u32; 8]) -> [u32; 8] {
// XOR with final round keys first
let mut a = s[7] ^ final_rk[7];
let mut b = s[6] ^ final_rk[6];
let mut c = s[5] ^ final_rk[5];
let mut d = s[3] ^ final_rk[3];
let mut e = s[4] ^ final_rk[4];
let mut f = s[2] ^ final_rk[2];
let mut g = s[1] ^ final_rk[1];
let mut h = s[0] ^ final_rk[0];
// Reverse Phase 3: 0x0F0F0F0F
let t8 = (a ^ (e >> 4)) & 0x0F0F0F0F;
a ^= t8;
e ^= t8 << 4;
let t9 = (c ^ (g >> 4)) & 0x0F0F0F0F;
c ^= t9;
g ^= t9 << 4;
let t10 = (b ^ (f >> 4)) & 0x0F0F0F0F;
b ^= t10;
f ^= t10 << 4;
let t11 = (d ^ (h >> 4)) & 0x0F0F0F0F;
d ^= t11;
h ^= t11 << 4;
// Reverse Phase 2: 0x33333333
let t4 = (a ^ (c >> 2)) & 0x33333333;
a ^= t4;
c ^= t4 << 2;
let t5 = (e ^ (g >> 2)) & 0x33333333;
e ^= t5;
g ^= t5 << 2;
let t6 = (b ^ (d >> 2)) & 0x33333333;
b ^= t6;
d ^= t6 << 2;
let t7 = (f ^ (h >> 2)) & 0x33333333;
f ^= t7;
h ^= t7 << 2;
// Reverse Phase 1: 0x55555555
let t0 = (a ^ (b >> 1)) & 0x55555555;
a ^= t0;
b ^= t0 << 1;
let t1 = (c ^ (d >> 1)) & 0x55555555;
c ^= t1;
d ^= t1 << 1;
let t2 = (e ^ (f >> 1)) & 0x55555555;
e ^= t2;
f ^= t2 << 1;
let t3 = (g ^ (h >> 1)) & 0x55555555;
g ^= t3;
h ^= t3 << 1;
// Output in original order (reverse of input mapping)
[h, f, d, b, g, e, c, a]
}

162
src/gift256/key_schedule.rs Normal file
View File

@@ -0,0 +1,162 @@
//! GIFT-256 fixsliced key schedule
//! Corresponds to gift256_key_schedule @ 0x80057559
//! Expands a 256-bit key into 120 u32 (480 bytes) of round keys
use super::sbox::sbox_bitsliced;
use super::interleave::{key_deinterleave_a, key_deinterleave_b, key_deinterleave_c, interleave_key_half};
/// NOT compensation table: offsets (u32 indices) that must be inverted at the end.
/// Corresponds to the mass of ks[offset] ^= 0xFFFFFFFF at the end of the decompilation.
const NOT_OFFSETS: [usize; 56] = [
0x08, 0x09, 0x0D, 0x0E,
0x10, 0x11, 0x15, 0x16,
0x18, 0x19, 0x1D, 0x1E,
0x20, 0x21, 0x25, 0x26,
0x28, 0x29, 0x2D, 0x2E,
0x30, 0x31, 0x35, 0x36,
0x38, 0x39, 0x3D, 0x3E,
0x40, 0x41, 0x45, 0x46,
0x48, 0x49, 0x4D, 0x4E,
0x50, 0x51, 0x55, 0x56,
0x58, 0x59, 0x5D, 0x5E,
0x60, 0x61, 0x65, 0x66,
0x68, 0x69, 0x6D, 0x6E,
0x70, 0x71, 0x75, 0x76,
];
/// GF extension function.
/// Decompiled:
/// uVar4 = (src >> param3 | src << (32-param3)) & 0x03030303 ^ dst;
/// result = uVar4 ^ (uVar4 & 0x03030303) << 6
/// ^ (uVar4 & 0x0F0F0F0F) << 4
/// ^ (uVar4 & 0x3F3F3F3F) << 2;
fn gf_expand(src: u32, dst: u32, rot: u32) -> u32 {
let x = (src.rotate_right(rot) & 0x03030303) ^ dst;
x ^ ((x & 0x03030303) << 6) ^ ((x & 0x0F0F0F0F) << 4) ^ ((x & 0x3F3F3F3F) << 2)
}
/// Key rotate: copy ks[base+0..7] forward to ks[base+8..15] with rotation.
/// Corresponds to gift256_key_rotate @ 0x80060a8f
///
/// After tracing the CFF state machine:
/// ks[base+15] = ks[base+7]
/// ks[base+14] = ks[base+6]
/// ...
/// ks[base+9] = ks[base+1]
/// ks[base+8] = ks[base+0]
fn key_rotate(ks: &mut [u32], base: usize) {
for i in (0..8).rev() {
ks[base + 8 + i] = ks[base + i];
}
}
/// LFSR expansion: perform GF expansion on 8 key words.
/// Corresponds to gift256_key_lfsr_expand @ 0x8005a060
/// rot = 14 or 6 (rotation amount)
fn key_lfsr_expand(ks: &mut [u32], base: usize, rot: u32) {
for i in 0..8 {
let src_idx = base + i;
let dst_idx = base + i - 16; // base is always >= 16
ks[src_idx] = gf_expand(ks[src_idx], ks[dst_idx], rot);
}
}
/// Full key schedule: expand 256-bit key into 120 u32 round keys.
pub fn key_schedule(key: &[u8; 32]) -> [u32; 120] {
let mut ks = [0u32; 120];
// === Step 1: Bit-interleave packing (State 3) ===
// Read key as 8 u32 (little-endian)
let mut kw = [0u32; 8];
for i in 0..8 {
kw[i] = u32::from_le_bytes([key[i * 4], key[i * 4 + 1], key[i * 4 + 2], key[i * 4 + 3]]);
}
// Interleave: two halves of 4 words each
interleave_key_half(&kw[0..4], &mut ks[0..8]);
interleave_key_half(&kw[4..8], &mut ks[8..16]);
// === Step 2: 12-step round key expansion (State 4+5, 6 iterations) ===
let mut round = 8usize;
let mut rk_const_offset = 0x10usize; // u32 index
for _step in 0..6 {
// --- Sub-step A (State 4) ---
key_rotate(&mut ks, round);
round += 8;
// S-Box on the newly written 8 words
let mut sbox_buf: [u32; 8] = ks[round - 8..round].try_into().unwrap();
sbox_bitsliced(&mut sbox_buf);
ks[round - 8..round].copy_from_slice(&sbox_buf);
// NOT compensation (4 specific offsets within this block)
ks[round - 8] ^= 0xFFFFFFFF;
ks[round - 7] ^= 0xFFFFFFFF;
ks[round - 3] ^= 0xFFFFFFFF;
ks[round - 2] ^= 0xFFFFFFFF;
// Round constant injection
ks[rk_const_offset] ^= 0xC000;
// LFSR expand with rot=14
key_lfsr_expand(&mut ks, round, 14);
// --- Sub-step B (State 5) ---
key_rotate(&mut ks, round);
round += 8;
let mut sbox_buf: [u32; 8] = ks[round - 8..round].try_into().unwrap();
sbox_bitsliced(&mut sbox_buf);
ks[round - 8..round].copy_from_slice(&sbox_buf);
ks[round - 8] ^= 0xFFFFFFFF;
ks[round - 7] ^= 0xFFFFFFFF;
ks[round - 3] ^= 0xFFFFFFFF;
ks[round - 2] ^= 0xFFFFFFFF;
// LFSR expand with rot=6
key_lfsr_expand(&mut ks, round, 6);
rk_const_offset += 17; // iVar24 += 0x44 bytes = 17 u32 indices
}
// === Step 3: Post-processing (State 2 + State 1) ===
// Deinterleave with 3 different mask patterns, each block covers 32 u32
for block in 0..3 {
let base = block * 32;
// Type A: offsets +8..+15
for j in 0..8 {
ks[base + 8 + j] = key_deinterleave_a(ks[base + 8 + j]);
}
// Type B: offset +16
ks[base + 16] = key_deinterleave_b(ks[base + 16]);
// Type B: offsets +17..+23
for j in 1..8 {
ks[base + 16 + j] = key_deinterleave_b(ks[base + 16 + j]);
}
// Type C: offsets +24..+31
for j in 0..8 {
ks[base + 24 + j] = key_deinterleave_c(ks[base + 24 + j]);
}
}
// Final round key deinterleave (ks[104..111] = offsets 0x1A0..0x1BC)
for j in 0..8 {
ks[104 + j] = key_deinterleave_a(ks[104 + j]);
}
// Initial NOT
ks[8] ^= 0xFFFFFFFF;
// Final NOT compensation
for &off in &NOT_OFFSETS {
ks[off] ^= 0xFFFFFFFF;
}
ks
}

145
src/gift256/linear.rs Normal file
View File

@@ -0,0 +1,145 @@
//! GIFT-256 linear diffusion layers
//! P1 and P2 are standalone functions, DA and DB are inlined in encrypt
use crate::util::nibble_half_swap;
/// P1: byte-ROT2 permutation (even rounds)
/// Corresponds to gift256_linear_P1_byteROT2 @ 0x800172b2
///
/// Core op: rot = ROR(x, 2) & 0xC0C0C0C0 | ROR(x, 10) & 0x3F3F3F3F
/// Processing order: s[6], s[7], s[5], s[4], s[1], s[2], s[0], s[3]
/// Extra: s[7]'s tmp value (t7) is XORed into s[0], s[4], s[3], s[1]
pub fn linear_p1(s: &mut [u32; 8]) {
let byte_rot2 = |x: u32| -> u32 {
(x.rotate_left(22) & 0x3F3F3F3F) | (x.rotate_left(30) & 0xC0C0C0C0)
};
let r6 = byte_rot2(s[6]);
let r7 = byte_rot2(s[7]);
let t6 = s[6] ^ r6;
let t7 = s[7] ^ r7;
s[7] = r7 ^ t6 ^ nibble_half_swap(t7);
let r5 = byte_rot2(s[5]);
let t5 = s[5] ^ r5;
s[6] = r6 ^ t5 ^ nibble_half_swap(t6);
let r4 = byte_rot2(s[4]);
let t4 = s[4] ^ r4;
s[5] = r5 ^ t4 ^ nibble_half_swap(t5);
let r1 = byte_rot2(s[1]);
let t1 = r1 ^ s[1];
let r2 = byte_rot2(s[2]);
let t2 = s[2] ^ r2;
s[2] = r2 ^ t1 ^ nibble_half_swap(t2);
let r0 = byte_rot2(s[0]);
let t0 = r0 ^ s[0];
s[0] = nibble_half_swap(t0) ^ r0 ^ t7; // NOTE: XOR with t7
let r3 = byte_rot2(s[3]);
let t3 = s[3] ^ r3;
s[4] = r4 ^ t3 ^ nibble_half_swap(t4) ^ t7;
s[3] = t2 ^ nibble_half_swap(t3) ^ r3 ^ t7;
s[1] = t0 ^ nibble_half_swap(t1) ^ r1 ^ t7;
}
/// P2: byte-ROT6 permutation (odd rounds)
/// Corresponds to gift256_linear_P2_byteROT6 @ 0x80006e1f
/// Identical structure to P1, only rotation amount changes:
/// rot = ROR(x, 6) & 0xFCFCFCFC | ROR(x, 14) & 0x03030303
pub fn linear_p2(s: &mut [u32; 8]) {
let byte_rot6 = |x: u32| -> u32 {
(x.rotate_left(18) & 0x03030303) | (x.rotate_left(26) & 0xFCFCFCFC)
};
let r6 = byte_rot6(s[6]);
let r7 = byte_rot6(s[7]);
let t6 = s[6] ^ r6;
let t7 = s[7] ^ r7;
s[7] = r7 ^ t6 ^ nibble_half_swap(t7);
let r5 = byte_rot6(s[5]);
let t5 = s[5] ^ r5;
s[6] = r6 ^ t5 ^ nibble_half_swap(t6);
let r4 = byte_rot6(s[4]);
let t4 = s[4] ^ r4;
s[5] = r5 ^ t4 ^ nibble_half_swap(t5);
let r1 = byte_rot6(s[1]);
let t1 = r1 ^ s[1];
let r2 = byte_rot6(s[2]);
let t2 = s[2] ^ r2;
s[2] = r2 ^ t1 ^ nibble_half_swap(t2);
let r0 = byte_rot6(s[0]);
let t0 = r0 ^ s[0];
s[0] = nibble_half_swap(t0) ^ r0 ^ t7;
let r3 = byte_rot6(s[3]);
let t3 = s[3] ^ r3;
s[4] = r4 ^ t3 ^ nibble_half_swap(t4) ^ t7;
s[3] = t2 ^ nibble_half_swap(t3) ^ r3 ^ t7;
s[1] = t0 ^ nibble_half_swap(t1) ^ r1 ^ t7;
}
/// DA: nibble-ROT4/12/20/28 diffusion (inlined in encrypt)
/// Corresponds to gift256_encrypt_fixsliced State 2 first part
///
/// Core op:
/// rotated = (x << 0x14 | x >> 0xc) & 0x0F0F0F0F | (x << 0x1c | x >> 4) & 0xF0F0F0F0
/// tmp = x ^ rotated
/// new = rotated ^ key ^ SWAP16(tmp) ^ cross_xor_pattern
pub fn diffusion_a(s: &mut [u32; 8], rk: &[u32; 8]) {
let nibble_rot = |x: u32| -> u32 {
(x.rotate_left(20) & 0x0F0F0F0F) | (x.rotate_left(28) & 0xF0F0F0F0)
};
let swap16 = |x: u32| -> u32 { x.rotate_left(16) };
let mut r = [0u32; 8];
let mut t = [0u32; 8];
for i in 0..8 {
r[i] = nibble_rot(s[i]);
t[i] = s[i] ^ r[i];
}
// Cross-XOR pattern (derived from decompilation)
s[0] = t[7] ^ r[0] ^ rk[0] ^ swap16(t[0]);
s[2] = r[2] ^ rk[2] ^ swap16(t[2]) ^ t[1];
s[5] = r[5] ^ rk[5] ^ swap16(t[5]) ^ t[4];
s[1] = rk[1] ^ swap16(t[1]) ^ t[0] ^ r[1] ^ t[7];
s[3] = r[3] ^ rk[3] ^ swap16(t[3]) ^ t[2] ^ t[7];
s[4] = rk[4] ^ swap16(t[4]) ^ t[3] ^ r[4] ^ t[7];
s[6] = r[6] ^ rk[6] ^ swap16(t[6]) ^ t[5];
s[7] = rk[7] ^ swap16(t[7]) ^ t[6] ^ r[7];
}
/// DB: byte-ROT8 diffusion (inlined in encrypt)
/// Corresponds to gift256_encrypt_fixsliced State 2 last part
///
/// Core op:
/// rotated = x << 0x18 | x >> 8 (= ROR(x, 8))
/// tmp = rotated ^ x
/// new = rotated ^ key ^ SWAP16(tmp) ^ cross_xor_pattern
pub fn diffusion_b(s: &mut [u32; 8], rk: &[u32; 8]) {
let swap16 = |x: u32| -> u32 { x.rotate_left(16) };
let mut r = [0u32; 8];
let mut t = [0u32; 8];
for i in 0..8 {
r[i] = s[i].rotate_right(8);
t[i] = r[i] ^ s[i];
}
// Cross-XOR pattern (same structure as DA but with byte-ROT8)
s[0] = r[0] ^ rk[0] ^ swap16(t[0]) ^ t[7];
s[2] = r[2] ^ rk[2] ^ swap16(t[2]) ^ t[1];
s[1] = rk[1] ^ swap16(t[1]) ^ t[0] ^ r[1] ^ t[7];
s[3] = r[3] ^ rk[3] ^ swap16(t[3]) ^ t[2] ^ t[7];
s[4] = rk[4] ^ swap16(t[4]) ^ t[3] ^ r[4] ^ t[7];
s[5] = r[5] ^ rk[5] ^ swap16(t[5]) ^ t[4];
s[6] = r[6] ^ rk[6] ^ swap16(t[6]) ^ t[5];
s[7] = rk[7] ^ swap16(t[7]) ^ t[6] ^ r[7];
}

5
src/gift256/mod.rs Normal file
View File

@@ -0,0 +1,5 @@
pub mod sbox;
pub mod encrypt;
pub mod key_schedule;
pub mod linear;
pub mod interleave;

90
src/gift256/sbox.rs Normal file
View File

@@ -0,0 +1,90 @@
//! GIFT-256 bitsliced S-Box
//! Corresponds to gift256_sbox_bitsliced @ 0x8000394c
//! Pure AND/XOR Boolean network, 32 S-Box instances in parallel
//!
//! Directly transcribed from Ghidra decompilation.
//! WARNING: Variables are reassigned multiple times. Do NOT optimize or merge.
/// Bitsliced S-Box: 8 u32 words processed in parallel.
/// Each bit position across s[0]..s[7] forms one 8-bit S-Box input/output.
///
/// Register mapping (encrypted_mem_read_u32 offsets):
/// s[0]=@0x00, s[1]=@0x04, s[2]=@0x08, s[3]=@0x0c
/// s[4]=@0x10, s[5]=@0x14, s[6]=@0x18, s[7]=@0x1c
pub fn sbox_bitsliced(s: &mut [u32; 8]) {
// Direct mapping from decompilation variable names:
// uVar1=@0x1c=s[7], uVar2=@0x04=s[1], uVar3=@0x10=s[4], uVar4=@0x08=s[2]
// uVar5=@0x0c=s[3], uVar6=@0x18=s[6], uVar7=@0x14=s[5], uVar(last)=@0x00=s[0]
let (a, b, c, d, e, f, g, h) = (s[7], s[1], s[2], s[3], s[4], s[5], s[6], s[0]);
let t1 = a ^ b; // uVar8 = uVar1 ^ uVar2
let t2 = e ^ c; // uVar14 = uVar3 ^ uVar4
let t3 = t1 ^ t2; // uVar9
let t4 = d ^ t3 ^ g; // uVar10
let t5 = e ^ a; // uVar15 = uVar3 ^ uVar1
let t6 = t4 ^ t5; // uVar11
let fg = f ^ g; // uVar7 = s[5] ^ s[6]
let t7 = t6 ^ fg; // uVar12
let ag = h ^ fg; // uVar25 = s[0] ^ (f^g)
let bx = b ^ ag; // uVar2 updated
let cd_mix = c ^ d ^ t3; // uVar26
let fg_cd = fg ^ cd_mix; // uVar7 subsequent
// Nonlinear core
let n1 = (t6 & t5) ^ (t2 & (fg_cd ^ t6)); // uVar16
let n2 = (t7 & t1) ^ t7 ^ (bx & ag) ^ t1 ^ n1; // uVar5
let ac = a ^ c; // uVar4 updated
let n1_ext = n1 ^ (cd_mix & t3) ^ t4 ^ ((h ^ cd_mix) & (ac ^ bx)); // uVar16 updated
let p = n2 & n1_ext; // uVar10
let q = (ac & fg_cd) ^ (t6 & t5); // uVar22
let r = (h & (e ^ ag)) ^ ac ^ fg_cd ^ (cd_mix & t3) ^ q; // uVar17
let qx = q ^ t7 ^ ((a ^ ag) & (h ^ t6)) ^ (t7 & t1) ^ a; // uVar22 updated
let m = (p ^ r) & (qx ^ n2); // uVar13
let n1x = r ^ n1_ext; // uVar16 = r ^ n1_ext
let rx = r ^ ((qx ^ p) & n1x); // uVar17 updated
let sx = (p ^ m) & qx; // uVar18
let n1x = (rx & (sx ^ qx ^ p)) ^ n1x; // uVar16 final
let sx = sx ^ n2; // uVar18 ^= uVar5
let qx = qx ^ m; // uVar22 ^= uVar13
let u = sx ^ qx; // uVar19
let v = n1x ^ u; // uVar5(final) = uVar16 ^ uVar19
let w = qx ^ rx; // uVar20
// Diffusion outputs
let o1 = ((v ^ w) & t2) ^ (w & t5); // uVar10
let o2 = o1 ^ (u & (ac ^ bx)); // uVar13
let o3 = o2 ^ (t3 & sx); // uVar9
let o4 = t7 & (rx ^ n1x); // uVar12
let o5 = o4 ^ (n1x & ag); // uVar23
let o6 = ((v ^ w) & (fg_cd ^ t6)) ^ o5; // uVar21
let o7 = o6 ^ (w & t6); // uVar14
// Output assignments (encrypted_mem_write_u32):
s[7] = o3 ^ o7; // @0x1c = uVar9 ^ uVar14
let o8 = (rx ^ n1x) & t1; // uVar8
let o9 = (qx & h) ^ (cd_mix & sx); // uVar24
let o6 = o8 ^ o9 ^ (fg_cd & v) ^ o6; // uVar21 updated
let o10 = o6 ^ (rx & (a ^ ag)); // uVar1
s[5] = (ac & v) ^ (w & t5) ^ o7 ^ o10; // @0x14
s[6] = o3 ^ o7; // @0x18 (same as s[7])
let o3x = (u & (h ^ cd_mix)) ^ (cd_mix & sx) ^ o5 ^ o3; // uVar9 updated
s[4] = o3x; // @0x10
s[2] = o2 ^ (qx & (e ^ ag)) ^ o10; // @0x08
let o4x = o9 ^ (rx & (h ^ t6)) ^ o4; // uVar12 updated
let o11 = o4x ^ o1 ^ (n1x & bx); // uVar1
s[1] = o11 ^ o6; // @0x04
s[0] = o11 ^ o8; // @0x00
s[3] = o4x ^ o3x; // @0x0c
}

65
src/hash/finalize.rs Normal file
View File

@@ -0,0 +1,65 @@
//! Hash finalization
//! Corresponds to hash_finalize @ 0x8000a792
use super::inner_compress::inner_compress;
use super::message::{process_message, u32x4_from_bytes};
/// Finalize: MD length padding + final compression + XOR truncation.
///
/// Parameters (from decompilation):
/// round_keys_and_state: 0x200 bytes (round keys + hash state)
/// chaining: 16 bytes chaining value
/// msg: message bytes
///
/// Returns: 16-byte digest
pub fn finalize(
hash_state: &[u32; 8],
chaining: &[u8; 16],
msg: &[u8],
) -> [u8; 16] {
let mut state = *hash_state;
// 1. Append 0x01 marker
process_message(&mut state, &[0x01]);
// 2. Process remaining message
process_message(&mut state, msg);
// 3. Length encoding (5-byte little-endian bit length)
let bit_len = msg.len() as u64;
let byte_bits = bit_len.wrapping_mul(8); // convert to bits
let mut len_block = [0u8; 16];
len_block[0] = (byte_bits & 0xFF) as u8;
len_block[1] = ((byte_bits >> 8) & 0xFF) as u8;
len_block[2] = ((byte_bits >> 16) & 0xFF) as u8;
len_block[3] = ((byte_bits >> 24) & 0xFF) as u8;
len_block[4] = ((byte_bits >> 32) & 0xFF) as u8;
// 4. Final compression
let block = u32x4_from_bytes(&len_block);
inner_compress(&mut state, &block);
// 5. State copy (Feistel output arrangement)
// copy state[4..5] -> state[0..1], state[6..7] -> state[2..3]
state[0] = state[4];
state[1] = state[5];
state[2] = state[6];
state[3] = state[7];
// 6. XOR truncation: output[15-i] = state_bytes[i] ^ chaining[i]
let state_bytes = state_to_bytes(&state);
let mut output = [0u8; 16];
for i in 0..16 {
output[15 - i] = state_bytes[i] ^ chaining[i];
}
output
}
fn state_to_bytes(s: &[u32; 8]) -> [u8; 32] {
let mut out = [0u8; 32];
for i in 0..8 {
out[i * 4..i * 4 + 4].copy_from_slice(&s[i].to_le_bytes());
}
out
}

138
src/hash/inner_compress.rs Normal file
View File

@@ -0,0 +1,138 @@
//! Inner compression: GF(2^4) matrix multiplication + LFSR diffusion
//! Corresponds to inner_compress_gf2x4 @ 0x80007084
//!
//! Processes 128-bit message blocks, updates 256-bit state
use crate::util::partial_bitrev_shr1;
/// GF(2^4) nibble-parallel multiplication.
/// Decompiled (State 0):
/// a1 = a & 0x11111111; a2 = a & 0x22222222; a4 = a & 0x44444444; a8 = a & 0x88888888;
/// b1 = b & 0x11111111; b2 = b & 0x22222222; b4 = b & 0x44444444; b8 = b & 0x88888888;
/// r8 = (a1*b8 ^ b4*a2 ^ a8*b1 ^ a4*b2) & 0x88888888
/// r4 = (b8*a8 ^ a1*b4 ^ a4*b1 ^ a2*b2) & 0x44444444
/// r1 = (b8*a2 ^ a4*b4 ^ b1*a1 ^ b2*a8) & 0x11111111
/// r2 = (a4*b8 ^ b4*a8 ^ b1*a2 ^ b2*a1) & 0x22222222
fn gf2x4_mul(a: u32, b: u32) -> u32 {
let a1 = a & 0x11111111;
let a2 = a & 0x22222222;
let a4 = a & 0x44444444;
let a8 = a & 0x88888888;
let b1 = b & 0x11111111;
let b2 = b & 0x22222222;
let b4 = b & 0x44444444;
let b8 = b & 0x88888888;
let r8 = (a1.wrapping_mul(b8) ^ b4.wrapping_mul(a2) ^ a8.wrapping_mul(b1) ^ a4.wrapping_mul(b2)) & 0x88888888;
let r4 = (b8.wrapping_mul(a8) ^ a1.wrapping_mul(b4) ^ a4.wrapping_mul(b1) ^ a2.wrapping_mul(b2)) & 0x44444444;
let r1 = (b8.wrapping_mul(a2) ^ a4.wrapping_mul(b4) ^ b1.wrapping_mul(a1) ^ b2.wrapping_mul(a8)) & 0x11111111;
let r2 = (a4.wrapping_mul(b8) ^ b4.wrapping_mul(a8) ^ b1.wrapping_mul(a2) ^ b2.wrapping_mul(a1)) & 0x22222222;
r8 | r4 | r1 | r2
}
/// LFSR forward feedback: x << 31 ^ x << 30 ^ x << 25
#[inline]
fn lfsr_fwd(x: u32) -> u32 {
(x << 31) ^ (x << 30) ^ (x << 25)
}
/// LFSR backward feedback: x >> 2 ^ x >> 1 ^ x >> 7
#[inline]
fn lfsr_bwd(x: u32) -> u32 {
(x >> 2) ^ (x >> 1) ^ (x >> 7)
}
/// Inner compression function.
/// state: 256-bit (8 u32), uses Feistel structure on upper/lower halves
/// block: 128-bit message block (4 u32)
pub fn inner_compress(state: &mut [u32; 8], block: &[u32; 4]) {
// 1. XOR message into state upper half (note: reverse order)
state[7] ^= block[3];
state[6] ^= block[2];
state[5] ^= block[1];
state[4] ^= block[0];
// 2. Prepare working values
let (s0, s1, s2, s3) = (state[4], state[5], state[6], state[7]);
let (m0, m1, m2, m3) = (block[0], block[1], block[2], block[3]);
// Precompute state-side values
let s01 = s0 ^ s1;
let s23 = s2 ^ s3;
let s02 = s0 ^ s2;
let s13 = s1 ^ s3;
let s0123 = s01 ^ s23;
// Precompute message-side values
let m01 = m0 ^ m1;
let m23 = m2 ^ m3;
let m02 = m0 ^ m2;
let m13 = m1 ^ m3;
let m0123 = m01 ^ m23;
// Bit-reversed versions
let rs0 = partial_bitrev_shr1(s0);
let rs1 = partial_bitrev_shr1(s1);
let rs2 = partial_bitrev_shr1(s2);
let rs3 = partial_bitrev_shr1(s3);
let rs01 = partial_bitrev_shr1(s01);
let rs23 = partial_bitrev_shr1(s23);
let rs02 = partial_bitrev_shr1(s02);
let rs13 = partial_bitrev_shr1(s13);
let rs0123 = partial_bitrev_shr1(s0123);
// 3. GF(2^4) multiplications (18 products)
let p0 = gf2x4_mul(s0, m0);
let p1 = gf2x4_mul(s1, m1);
let p2 = gf2x4_mul(s2, m2);
let p3 = gf2x4_mul(s3, m3);
let p01 = gf2x4_mul(s01, m01);
let p23 = gf2x4_mul(s23, m23);
let p02 = gf2x4_mul(s02, m02);
let p13 = gf2x4_mul(s13, m13);
let p0123 = gf2x4_mul(s0123, m0123);
let rp0 = gf2x4_mul(rs0, m0);
let rp1 = gf2x4_mul(rs1, m1);
let rp2 = gf2x4_mul(rs2, m2);
let rp3 = gf2x4_mul(rs3, m3);
let rp01 = gf2x4_mul(rs01, m01);
let rp23 = gf2x4_mul(rs23, m23);
let rp02 = gf2x4_mul(rs02, m02);
let rp13 = gf2x4_mul(rs13, m13);
let rp0123 = gf2x4_mul(rs0123, m0123);
// 4. Karatsuba-style recombination
// Row 0: p0
// Row 1: p01 ^ p0 ^ p1
// Row 2: p02 ^ p0 ^ p2
// Row 3: p0123 ^ p01 ^ p02 ^ p13 ^ p23 ^ p0 ^ p3
let k0 = p0;
let k1 = p01 ^ p0 ^ p1;
let k2 = p02 ^ p0 ^ p2;
let k3 = p0123 ^ p01 ^ p02 ^ p13 ^ p23 ^ p0 ^ p3;
let rk0 = rp0;
let rk1 = rp01 ^ rp0 ^ rp1;
let rk2 = rp02 ^ rp0 ^ rp2;
let rk3 = rp0123 ^ rp01 ^ rp02 ^ rp13 ^ rp23 ^ rp0 ^ rp3;
// 5. LFSR diffusion + mixing
let out0 = k0 ^ lfsr_bwd(rk0) ^ lfsr_fwd(k0);
let out1 = k1 ^ lfsr_bwd(rk1) ^ lfsr_fwd(k1);
let out2 = k2 ^ lfsr_bwd(rk2) ^ lfsr_fwd(k2);
let out3 = k3 ^ lfsr_bwd(rk3) ^ lfsr_fwd(k3);
// 6. Feistel update: shift right half
state[4] = state[0];
state[5] = state[1];
state[6] = state[2];
state[7] = state[3];
// Write new left half
state[0] = out0;
state[1] = out1;
state[2] = out2;
state[3] = out3;
}

34
src/hash/message.rs Normal file
View File

@@ -0,0 +1,34 @@
//! Message block processing for the hash function.
//! Breaks messages into 16-byte blocks and calls inner_compress.
use super::inner_compress::inner_compress;
/// Process a message by breaking it into 16-byte blocks.
/// Each block is converted to [u32; 4] (little-endian) and compressed into state.
pub fn process_message(state: &mut [u32; 8], msg: &[u8]) {
let mut offset = 0;
while offset + 16 <= msg.len() {
let block = u32x4_from_bytes(&msg[offset..offset + 16]);
inner_compress(state, &block);
offset += 16;
}
// Handle remaining bytes (partial block, zero-padded)
if offset < msg.len() {
let mut padded = [0u8; 16];
let remaining = msg.len() - offset;
padded[..remaining].copy_from_slice(&msg[offset..]);
let block = u32x4_from_bytes(&padded);
inner_compress(state, &block);
}
}
/// Convert 16 bytes to [u32; 4] (little-endian).
pub fn u32x4_from_bytes(b: &[u8]) -> [u32; 4] {
[
u32::from_le_bytes([b[0], b[1], b[2], b[3]]),
u32::from_le_bytes([b[4], b[5], b[6], b[7]]),
u32::from_le_bytes([b[8], b[9], b[10], b[11]]),
u32::from_le_bytes([b[12], b[13], b[14], b[15]]),
]
}

69
src/hash/mmo.rs Normal file
View File

@@ -0,0 +1,69 @@
//! Matyas-Meyer-Oseas compression: H_i = E_K(0) ⊕ double(extract(E_K(0)))
//! Corresponds to mmo_compress @ 0x8004fbac
use crate::gift256::encrypt::encrypt;
/// MMO compression state
pub struct MmoState {
pub round_keys: [u32; 120],
pub chaining: [u8; 16],
}
/// MMO compression function.
/// Decompiled flow:
/// 1. Encrypt all-zero plaintext: gift256_encrypt(ciphertext, round_keys, zeros)
/// 2. Extract bytes 0x20..0x2F (16 bytes) from ciphertext buffer
/// 3. Assemble as two u64 (big-endian): hi=bytes[0x20..0x27], lo=bytes[0x28..0x2F]
/// 4. GF(2^128) doubling: {hi,lo} <<= 1, if carry: hi ^= (carry<<62) | (carry<<57)
/// 5. Write into chaining value, copy round keys
pub fn mmo_compress(round_keys: &[u32; 120]) -> MmoState {
// 1. Encrypt all-zero plaintext
let plaintext = [0u32; 8];
let ciphertext = encrypt(&plaintext, round_keys);
// 2. Convert ciphertext to bytes (little-endian u32 layout)
let ct_bytes = ciphertext_to_bytes(&ciphertext);
// 3. Build hi and lo u64 from bytes at offset 0x20 (= first 16 bytes of ciphertext)
// The ciphertext buffer in the decompilation starts at the plaintext+ciphertext stack area
// offset 0x20 from the base = the ciphertext portion (bytes 0..15)
let hi = u64::from_be_bytes([
ct_bytes[0], ct_bytes[1], ct_bytes[2], ct_bytes[3],
ct_bytes[4], ct_bytes[5], ct_bytes[6], ct_bytes[7],
]);
let lo = u64::from_be_bytes([
ct_bytes[8], ct_bytes[9], ct_bytes[10], ct_bytes[11],
ct_bytes[12], ct_bytes[13], ct_bytes[14], ct_bytes[15],
]);
// 4. GF(2^128) doubling: shift left by 1
let carry = hi >> 63;
let mut hi = (hi << 1) | (lo >> 63);
let lo = lo << 1;
// Feedback polynomial: if carry bit set, XOR with reduction polynomial
// Decompiled: (uVar16 << 0x38 & 0x8000000000000000 | uVar18 << 0x3e | uVar18 << 0x39)
// where uVar18 = carry bit
if carry != 0 {
hi ^= (1u64 << 62) | (1u64 << 57);
}
// 5. Write chaining value (big-endian)
let mut chaining = [0u8; 16];
chaining[0..8].copy_from_slice(&hi.to_be_bytes());
chaining[8..16].copy_from_slice(&lo.to_be_bytes());
MmoState {
round_keys: *round_keys,
chaining,
}
}
fn ciphertext_to_bytes(ct: &[u32; 8]) -> [u8; 32] {
let mut bytes = [0u8; 32];
for i in 0..8 {
let b = ct[i].to_le_bytes();
bytes[i * 4..i * 4 + 4].copy_from_slice(&b);
}
bytes
}

4
src/hash/mod.rs Normal file
View File

@@ -0,0 +1,4 @@
pub mod mmo;
pub mod inner_compress;
pub mod message;
pub mod finalize;

6
src/lib.rs Normal file
View File

@@ -0,0 +1,6 @@
pub mod util;
pub mod pcg;
pub mod sbox;
pub mod gift256;
pub mod hash;
pub mod solver;

109
src/main.rs Normal file
View File

@@ -0,0 +1,109 @@
//! hCaptcha PoW solver CLI
//! Parses JWT challenge -> calls solver -> outputs base64 nonce
use base64::{engine::general_purpose::STANDARD, Engine};
use serde::Deserialize;
mod pcg;
mod sbox;
mod gift256;
mod hash;
mod solver;
mod util;
#[derive(Deserialize)]
#[allow(dead_code)]
struct JwtPayload {
#[serde(default)]
f: u32,
#[serde(default)]
s: u32,
#[serde(default)]
t: String,
d: String,
#[serde(default)]
l: String,
#[serde(default)]
i: String,
#[serde(default)]
e: u64,
#[serde(default)]
n: String,
#[serde(default)]
c: u32,
}
fn main() {
// 1. Read JWT from command line
let jwt_req = std::env::args().nth(1).expect("Usage: hcaptcha-pow <jwt>");
// 2. Decode JWT payload (base64url, no signature verification)
let parts: Vec<&str> = jwt_req.split('.').collect();
if parts.len() < 2 {
eprintln!("Invalid JWT format");
std::process::exit(1);
}
let payload_b64 = parts[1];
let payload_bytes = base64::engine::general_purpose::URL_SAFE_NO_PAD
.decode(payload_b64)
.expect("Invalid base64 in JWT payload");
let payload: JwtPayload = serde_json::from_slice(&payload_bytes).expect("Invalid JSON in JWT payload");
println!("Algorithm: {}", payload.n);
println!("Difficulty: {}", payload.c);
println!("Expiration: {}", payload.e);
// 3. Decode challenge data from `d` field
let challenge_data = STANDARD.decode(&payload.d).unwrap_or_else(|_| {
// Try with padding adjustment
let padded = format!("{}==", payload.d.trim_end_matches('='));
STANDARD.decode(&padded).expect("Invalid base64 in challenge data")
});
println!("Challenge data: {} bytes", challenge_data.len());
// 4. Parse challenge (needs at least 49 bytes: 32 key + 16 target + 1 extra)
if challenge_data.len() < 49 {
eprintln!("Challenge data too short: {} bytes (need >= 49)", challenge_data.len());
std::process::exit(1);
}
let mut key_material = [0u8; 32];
key_material.copy_from_slice(&challenge_data[0..32]);
let mut target_hash = [0u8; 16];
target_hash.copy_from_slice(&challenge_data[32..48]);
let extra_byte = challenge_data[48];
let challenge = solver::Challenge {
key_material,
target_hash,
extra_byte,
};
// 5. Solve
let seed = std::time::SystemTime::now()
.duration_since(std::time::UNIX_EPOCH)
.unwrap()
.as_nanos() as u64;
let max_iter = if payload.c > 0 { payload.c } else { 1_000_000 };
println!("Solving with max {} iterations...", max_iter);
let solution = solver::solve(&challenge, max_iter, seed);
// 6. Output
match solution {
Some(sol) => {
let nonce_b64 = STANDARD.encode(sol.nonce);
println!("Found solution in {} iterations", sol.iterations);
println!("n={}", nonce_b64);
}
None => {
eprintln!("No solution found within {} iterations", max_iter);
std::process::exit(1);
}
}
}

49
src/pcg.rs Normal file
View File

@@ -0,0 +1,49 @@
//! PCG-XSH-RR-64/32 pseudo-random number generator
//! Corresponds to pow_main_dispatch State 0x43
//! Evidence: multiplier 0x5851f42d4c957f2d, inc = seed << 1 | 1
const PCG_MULTIPLIER: u64 = 0x5851f42d4c957f2d;
pub struct PcgRng {
state: u64,
inc: u64,
}
impl PcgRng {
/// Initialize PCG from seed.
/// Decompiled:
/// uVar36 = uVar36 << 1 | 1; // inc = seed << 1 | 1
/// uVar35 = uVar36 * 0x5851f42d4c957f2e + uVar35 * 0x5851f42d4c957f2d;
pub fn new(seed: u64) -> Self {
let inc = (seed << 1) | 1; // always odd
let state = seed
.wrapping_mul(PCG_MULTIPLIER)
.wrapping_add(inc.wrapping_mul(PCG_MULTIPLIER.wrapping_add(1)));
Self { state, inc }
}
/// Generate one 32-bit random number.
/// Decompiled:
/// uVar35 = uVar35 * 0x5851f42d4c957f2d + uVar36;
/// uVar37 = (uint)(uVar35 >> 0x20);
/// uVar6 = uVar37 >> 0xd ^ (uint)(uVar35 >> 0x1b);
/// uVar37 = uVar37 >> 0x1b;
/// output = uVar6 >> uVar37 | uVar6 << (0x20 - uVar37);
fn next_u32(&mut self) -> u32 {
self.state = self.state.wrapping_mul(PCG_MULTIPLIER).wrapping_add(self.inc);
let hi = (self.state >> 32) as u32;
let xorshifted = (hi >> 13) ^ ((self.state >> 27) as u32);
let rot = hi >> 27;
xorshifted.rotate_right(rot)
}
/// Generate 12-byte nonce.
/// Corresponds to the 12-iteration loop in State 0x43.
pub fn generate_nonce(&mut self) -> [u8; 12] {
let mut nonce = [0u8; 12];
for i in 0..12 {
nonce[i] = self.next_u32() as u8;
}
nonce
}
}

30
src/sbox.rs Normal file
View File

@@ -0,0 +1,30 @@
//! Custom polynomial S-Box: S(x) = 192x^6 + 224x^5 + 120x^4 + 200x^3 + 150x^2 + 65x + 147 (mod 256)
//! Corresponds to pow_main_dispatch State 0x0E/0x0F
//! Evidence: constants 0xE0, 0x78, -0x40, 0x86838DC8, 0x96, 0x41, 0x93
/// Apply polynomial S-Box to each byte of a 32-byte buffer.
/// Decompiled:
/// iVar4 = iVar1 * iVar1; // x^2
/// iVar10 = iVar4 * iVar4; // x^4
/// result = iVar1 * iVar10 * 0xe0 // 224*x^5
/// + iVar10 * 0x78 // 120*x^4
/// + iVar10 * iVar4 * -0x40 // -64*x^6 = 192*x^6 (mod 256)
/// + ((iVar1 * 0x86838DC8 + 0x96) * iVar1 + 0x41) * iVar1
/// + 0x93;
pub fn apply_polynomial_sbox(buf: &mut [u8; 32]) {
for b in buf.iter_mut() {
let x = *b as u32;
let x2 = x.wrapping_mul(x);
let x4 = x2.wrapping_mul(x2);
let result = x.wrapping_mul(x4).wrapping_mul(0xE0) // 224*x^5
.wrapping_add(x4.wrapping_mul(0x78)) // 120*x^4
.wrapping_add(x4.wrapping_mul(x2).wrapping_mul(0xFFFF_FFC0)) // 192*x^6 (-64 mod 2^32)
.wrapping_add(
x.wrapping_mul(0x86838DC8_u32).wrapping_add(0x96) // 200*x + 150
.wrapping_mul(x).wrapping_add(0x41) // *x + 65
.wrapping_mul(x) // -> 200x^3+150x^2+65x
)
.wrapping_add(0x93); // + 147
*b = result as u8;
}
}

97
src/solver.rs Normal file
View File

@@ -0,0 +1,97 @@
//! Top-level PoW solver
//! Corresponds to pow_main_dispatch (Yb) solve path (0xABAB270C)
use crate::pcg::PcgRng;
use crate::sbox::apply_polynomial_sbox;
use crate::gift256;
use crate::hash;
/// PoW solution result
pub struct PowSolution {
pub nonce: [u8; 12],
pub iterations: u32,
}
/// Challenge data parsed from JWT `d` field
pub struct Challenge {
pub key_material: [u8; 32], // 32-byte key material (before S-Box)
pub target_hash: [u8; 16], // 16-byte target hash
pub extra_byte: u8, // 33rd byte
}
/// Main solve function.
///
/// Full flow:
/// 1. PCG generates 12-byte nonce
/// 2. S-Box polynomial substitution (32-byte key material)
/// 3. GIFT-256 key schedule -> 480 bytes round keys
/// 4. MMO compression
/// 5. Assemble hash input: [nonce_u32_0, nonce_u32_1, nonce_u32_2, 0x01000000]
/// 6. GIFT-256 encrypt + hash_finalize -> 128-bit digest
/// 7. Constant-time 16-byte comparison
/// 8. Match -> return nonce; no match -> regenerate nonce
pub fn solve(challenge: &Challenge, max_iterations: u32, seed: u64) -> Option<PowSolution> {
let mut rng = PcgRng::new(seed);
// Pre-compute key schedule and MMO (these don't depend on the nonce)
let mut key_data = challenge.key_material;
apply_polynomial_sbox(&mut key_data);
let round_keys = gift256::key_schedule::key_schedule(&key_data);
let mmo_state = hash::mmo::mmo_compress(&round_keys);
// Initialize hash state from MMO output
// The hash state is derived from the round keys + chaining value
let base_state = [0u32; 8];
// State initialized to zeros, will be populated by the hash process
for iter in 0..max_iterations {
// 1. Generate 12-byte nonce
let nonce = rng.generate_nonce();
// 2. Assemble hash input block
let nonce_u32_0 = u32::from_le_bytes([nonce[0], nonce[1], nonce[2], nonce[3]]);
let nonce_u32_1 = u32::from_le_bytes([nonce[4], nonce[5], nonce[6], nonce[7]]);
let nonce_u32_2 = u32::from_le_bytes([nonce[8], nonce[9], nonce[10], nonce[11]]);
let hash_input = [nonce_u32_0, nonce_u32_1, nonce_u32_2, 0x01000000u32];
// 3. GIFT-256 encrypt
let encrypted = gift256::encrypt::encrypt(
&[hash_input[0], hash_input[1], hash_input[2], hash_input[3], 0, 0, 0, 0],
&round_keys,
);
// 4. Finalize hash
// Build state from encrypted output
let mut hash_state = base_state;
for i in 0..8 {
hash_state[i] = encrypted[i];
}
let digest = hash::finalize::finalize(
&hash_state,
&mmo_state.chaining,
&nonce,
);
// 5. Compare with target
if constant_time_eq(&digest, &challenge.target_hash) {
return Some(PowSolution {
nonce,
iterations: iter + 1,
});
}
}
None
}
/// Constant-time comparison (16 bytes)
fn constant_time_eq(a: &[u8; 16], b: &[u8; 16]) -> bool {
let mut result = 0u8;
for i in 0..16 {
result |= a[i] ^ b[i];
}
result == 0
}

47
src/util.rs Normal file
View File

@@ -0,0 +1,47 @@
//! Bit-manipulation helpers, directly mapped from repeated WASM decompilation patterns.
/// 32-bit right rotate
#[inline(always)]
pub fn ror32(x: u32, n: u32) -> u32 {
x.rotate_right(n)
}
/// 32-bit left rotate
#[inline(always)]
pub fn rol32(x: u32, n: u32) -> u32 {
x.rotate_left(n)
}
/// 32-bit byte swap (endian conversion)
/// Decompiled pattern: x << 0x18 | (x & 0xff00) << 8 | x >> 8 & 0xff00 | x >> 0x18
#[inline(always)]
pub fn bswap32(x: u32) -> u32 {
x.swap_bytes()
}
/// Full bit reversal: bswap → nibble_swap → bit_pair_swap → bit_swap
/// Used in inner_compress_gf2x4
#[inline]
pub fn bitrev32(x: u32) -> u32 {
let x = bswap32(x);
let x = ((x >> 4) & 0x0F0F0F0F) | ((x & 0x0F0F0F0F) << 4);
let x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
((x >> 1) & 0x55555555) | ((x & 0x55555555) << 1)
}
/// Partial bit reversal (mask 0x55555554 instead of 0x55555555), then >> 1
/// Used in inner_compress_gf2x4
#[inline]
pub fn partial_bitrev_shr1(x: u32) -> u32 {
let x = bswap32(x);
let x = ((x >> 4) & 0x0F0F0F0F) | ((x & 0x0F0F0F0F) << 4);
let x = ((x >> 2) & 0x33333333) | ((x & 0x33333333) << 2);
(((x >> 1) & 0x55555554) | ((x & 0x55555555) << 1)) >> 1
}
/// Nibble half-swap (used in linear diffusion layers P1 and P2)
/// (x << 12 | x >> 20) & 0x0F0F0F0F | (x << 20 | x >> 12) & 0xF0F0F0F0
#[inline]
pub fn nibble_half_swap(x: u32) -> u32 {
(x.rotate_left(12) & 0x0F0F0F0F) | (x.rotate_left(20) & 0xF0F0F0F0)
}