Module refinery.lib.crypto.aria

"""
Pure-Python implementation of the ARIA block cipher as specified in RFC5794. ARIA is a South Korean
standard block cipher with a substitution-permutation network (SPN) structure. It supports 128, 192,
and 256-bit keys with 12, 14, or 16 rounds respectively.
"""
from __future__ import annotations

import struct

from refinery.lib.crypto import BlockCipher, BufferType, CipherMode

_SB1 = (
    0x63, 0x7C, 0x77, 0x7B, 0xF2, 0x6B, 0x6F, 0xC5, 0x30, 0x01, 0x67, 0x2B, 0xFE, 0xD7, 0xAB, 0x76,
    0xCA, 0x82, 0xC9, 0x7D, 0xFA, 0x59, 0x47, 0xF0, 0xAD, 0xD4, 0xA2, 0xAF, 0x9C, 0xA4, 0x72, 0xC0,
    0xB7, 0xFD, 0x93, 0x26, 0x36, 0x3F, 0xF7, 0xCC, 0x34, 0xA5, 0xE5, 0xF1, 0x71, 0xD8, 0x31, 0x15,
    0x04, 0xC7, 0x23, 0xC3, 0x18, 0x96, 0x05, 0x9A, 0x07, 0x12, 0x80, 0xE2, 0xEB, 0x27, 0xB2, 0x75,
    0x09, 0x83, 0x2C, 0x1A, 0x1B, 0x6E, 0x5A, 0xA0, 0x52, 0x3B, 0xD6, 0xB3, 0x29, 0xE3, 0x2F, 0x84,
    0x53, 0xD1, 0x00, 0xED, 0x20, 0xFC, 0xB1, 0x5B, 0x6A, 0xCB, 0xBE, 0x39, 0x4A, 0x4C, 0x58, 0xCF,
    0xD0, 0xEF, 0xAA, 0xFB, 0x43, 0x4D, 0x33, 0x85, 0x45, 0xF9, 0x02, 0x7F, 0x50, 0x3C, 0x9F, 0xA8,
    0x51, 0xA3, 0x40, 0x8F, 0x92, 0x9D, 0x38, 0xF5, 0xBC, 0xB6, 0xDA, 0x21, 0x10, 0xFF, 0xF3, 0xD2,
    0xCD, 0x0C, 0x13, 0xEC, 0x5F, 0x97, 0x44, 0x17, 0xC4, 0xA7, 0x7E, 0x3D, 0x64, 0x5D, 0x19, 0x73,
    0x60, 0x81, 0x4F, 0xDC, 0x22, 0x2A, 0x90, 0x88, 0x46, 0xEE, 0xB8, 0x14, 0xDE, 0x5E, 0x0B, 0xDB,
    0xE0, 0x32, 0x3A, 0x0A, 0x49, 0x06, 0x24, 0x5C, 0xC2, 0xD3, 0xAC, 0x62, 0x91, 0x95, 0xE4, 0x79,
    0xE7, 0xC8, 0x37, 0x6D, 0x8D, 0xD5, 0x4E, 0xA9, 0x6C, 0x56, 0xF4, 0xEA, 0x65, 0x7A, 0xAE, 0x08,
    0xBA, 0x78, 0x25, 0x2E, 0x1C, 0xA6, 0xB4, 0xC6, 0xE8, 0xDD, 0x74, 0x1F, 0x4B, 0xBD, 0x8B, 0x8A,
    0x70, 0x3E, 0xB5, 0x66, 0x48, 0x03, 0xF6, 0x0E, 0x61, 0x35, 0x57, 0xB9, 0x86, 0xC1, 0x1D, 0x9E,
    0xE1, 0xF8, 0x98, 0x11, 0x69, 0xD9, 0x8E, 0x94, 0x9B, 0x1E, 0x87, 0xE9, 0xCE, 0x55, 0x28, 0xDF,
    0x8C, 0xA1, 0x89, 0x0D, 0xBF, 0xE6, 0x42, 0x68, 0x41, 0x99, 0x2D, 0x0F, 0xB0, 0x54, 0xBB, 0x16,
)

_SB2 = (
    0xE2, 0x4E, 0x54, 0xFC, 0x94, 0xC2, 0x4A, 0xCC, 0x62, 0x0D, 0x6A, 0x46, 0x3C, 0x4D, 0x8B, 0xD1,
    0x5E, 0xFA, 0x64, 0xCB, 0xB4, 0x97, 0xBE, 0x2B, 0xBC, 0x77, 0x2E, 0x03, 0xD3, 0x19, 0x59, 0xC1,
    0x1D, 0x06, 0x41, 0x6B, 0x55, 0xF0, 0x99, 0x69, 0xEA, 0x9C, 0x18, 0xAE, 0x63, 0xDF, 0xE7, 0xBB,
    0x00, 0x73, 0x66, 0xFB, 0x96, 0x4C, 0x85, 0xE4, 0x3A, 0x09, 0x45, 0xAA, 0x0F, 0xEE, 0x10, 0xEB,
    0x2D, 0x7F, 0xF4, 0x29, 0xAC, 0xCF, 0xAD, 0x91, 0x8D, 0x78, 0xC8, 0x95, 0xF9, 0x2F, 0xCE, 0xCD,
    0x08, 0x7A, 0x88, 0x38, 0x5C, 0x83, 0x2A, 0x28, 0x47, 0xDB, 0xB8, 0xC7, 0x93, 0xA4, 0x12, 0x53,
    0xFF, 0x87, 0x0E, 0x31, 0x36, 0x21, 0x58, 0x48, 0x01, 0x8E, 0x37, 0x74, 0x32, 0xCA, 0xE9, 0xB1,
    0xB7, 0xAB, 0x0C, 0xD7, 0xC4, 0x56, 0x42, 0x26, 0x07, 0x98, 0x60, 0xD9, 0xB6, 0xB9, 0x11, 0x40,
    0xEC, 0x20, 0x8C, 0xBD, 0xA0, 0xC9, 0x84, 0x04, 0x49, 0x23, 0xF1, 0x4F, 0x50, 0x1F, 0x13, 0xDC,
    0xD8, 0xC0, 0x9E, 0x57, 0xE3, 0xC3, 0x7B, 0x65, 0x3B, 0x02, 0x8F, 0x3E, 0xE8, 0x25, 0x92, 0xE5,
    0x15, 0xDD, 0xFD, 0x17, 0xA9, 0xBF, 0xD4, 0x9A, 0x7E, 0xC5, 0x39, 0x67, 0xFE, 0x76, 0x9D, 0x43,
    0xA7, 0xE1, 0xD0, 0xF5, 0x68, 0xF2, 0x1B, 0x34, 0x70, 0x05, 0xA3, 0x8A, 0xD5, 0x79, 0x86, 0xA8,
    0x30, 0xC6, 0x51, 0x4B, 0x1E, 0xA6, 0x27, 0xF6, 0x35, 0xD2, 0x6E, 0x24, 0x16, 0x82, 0x5F, 0xDA,
    0xE6, 0x75, 0xA2, 0xEF, 0x2C, 0xB2, 0x1C, 0x9F, 0x5D, 0x6F, 0x80, 0x0A, 0x72, 0x44, 0x9B, 0x6C,
    0x90, 0x0B, 0x5B, 0x33, 0x7D, 0x5A, 0x52, 0xF3, 0x61, 0xA1, 0xF7, 0xB0, 0xD6, 0x3F, 0x7C, 0x6D,
    0xED, 0x14, 0xE0, 0xA5, 0x3D, 0x22, 0xB3, 0xF8, 0x89, 0xDE, 0x71, 0x1A, 0xAF, 0xBA, 0xB5, 0x81,
)

_SB3 = [0] * 256
_SB4 = [0] * 256
for _i in range(256):
    _SB3[_SB1[_i]] = _i
    _SB4[_SB2[_i]] = _i

_CK1 = (0x517CC1B7, 0x27220A94, 0xFE13ABE8, 0xFA9A6EE0)
_CK2 = (0x6DB14ACC, 0x9E21C820, 0xFF28B1D5, 0xEF5DE2B0)
_CK3 = (0xDB92371D, 0x2126E970, 0x03249775, 0x04E8C90E)


def _xor128(a: tuple[int, ...], b: tuple[int, ...]) -> tuple[int, int, int, int]:
    return (a[0] ^ b[0], a[1] ^ b[1], a[2] ^ b[2], a[3] ^ b[3])


def _sub_layer_odd(t: tuple[int, ...]) -> tuple[int, int, int, int]:
    x0, x1, x2, x3 = t
    r0 = (_SB1[(x0 >> 24) & 0xFF] << 24) | (_SB2[(x0 >> 16) & 0xFF] << 16) | (_SB3[(x0 >> 8) & 0xFF] << 8) | _SB4[x0 & 0xFF]
    r1 = (_SB1[(x1 >> 24) & 0xFF] << 24) | (_SB2[(x1 >> 16) & 0xFF] << 16) | (_SB3[(x1 >> 8) & 0xFF] << 8) | _SB4[x1 & 0xFF]
    r2 = (_SB1[(x2 >> 24) & 0xFF] << 24) | (_SB2[(x2 >> 16) & 0xFF] << 16) | (_SB3[(x2 >> 8) & 0xFF] << 8) | _SB4[x2 & 0xFF]
    r3 = (_SB1[(x3 >> 24) & 0xFF] << 24) | (_SB2[(x3 >> 16) & 0xFF] << 16) | (_SB3[(x3 >> 8) & 0xFF] << 8) | _SB4[x3 & 0xFF]
    return (r0, r1, r2, r3)


def _sub_layer_even(t: tuple[int, ...]) -> tuple[int, int, int, int]:
    x0, x1, x2, x3 = t
    r0 = (_SB3[(x0 >> 24) & 0xFF] << 24) | (_SB4[(x0 >> 16) & 0xFF] << 16) | (_SB1[(x0 >> 8) & 0xFF] << 8) | _SB2[x0 & 0xFF]
    r1 = (_SB3[(x1 >> 24) & 0xFF] << 24) | (_SB4[(x1 >> 16) & 0xFF] << 16) | (_SB1[(x1 >> 8) & 0xFF] << 8) | _SB2[x1 & 0xFF]
    r2 = (_SB3[(x2 >> 24) & 0xFF] << 24) | (_SB4[(x2 >> 16) & 0xFF] << 16) | (_SB1[(x2 >> 8) & 0xFF] << 8) | _SB2[x2 & 0xFF]
    r3 = (_SB3[(x3 >> 24) & 0xFF] << 24) | (_SB4[(x3 >> 16) & 0xFF] << 16) | (_SB1[(x3 >> 8) & 0xFF] << 8) | _SB2[x3 & 0xFF]
    return (r0, r1, r2, r3)


def _diff_layer(t: tuple[int, ...]) -> tuple[int, int, int, int]:
    x0, x1, x2, x3 = t
    y00 = (x0 >> 0x18) & 0xFF
    y01 = (x0 >> 0x10) & 0xFF
    y02 = (x0 >> 0x08) & 0xFF
    y03 = (x0        ) & 0xFF # noqa
    y04 = (x1 >> 0x18) & 0xFF
    y05 = (x1 >> 0x10) & 0xFF
    y06 = (x1 >> 0x08) & 0xFF
    y07 = (x1        ) & 0xFF # noqa
    y08 = (x2 >> 0x18) & 0xFF
    y09 = (x2 >> 0x10) & 0xFF
    y10 = (x2 >> 0x08) & 0xFF
    y11 = (x2        ) & 0xFF # noqa
    y12 = (x3 >> 0x18) & 0xFF
    y13 = (x3 >> 0x10) & 0xFF
    y14 = (x3 >> 0x08) & 0xFF
    y15 = (x3        ) & 0xFF # noqa
    return (
        (y03 ^ y04 ^ y06 ^ y08 ^ y09 ^ y13 ^ y14) << 0x18 # noqa
      | (y02 ^ y05 ^ y07 ^ y08 ^ y09 ^ y12 ^ y15) << 0x10 # noqa
      | (y01 ^ y04 ^ y06 ^ y10 ^ y11 ^ y12 ^ y15) << 0x08 # noqa
      | (y00 ^ y05 ^ y07 ^ y10 ^ y11 ^ y13 ^ y14),
        (y00 ^ y02 ^ y05 ^ y08 ^ y11 ^ y14 ^ y15) << 0x18
      | (y01 ^ y03 ^ y04 ^ y09 ^ y10 ^ y14 ^ y15) << 0x10
      | (y00 ^ y02 ^ y07 ^ y09 ^ y10 ^ y12 ^ y13) << 0x08
      | (y01 ^ y03 ^ y06 ^ y08 ^ y11 ^ y12 ^ y13),
        (y00 ^ y01 ^ y04 ^ y07 ^ y10 ^ y13 ^ y15) << 0x18
      | (y00 ^ y01 ^ y05 ^ y06 ^ y11 ^ y12 ^ y14) << 0x10
      | (y02 ^ y03 ^ y05 ^ y06 ^ y08 ^ y13 ^ y15) << 0x08
      | (y02 ^ y03 ^ y04 ^ y07 ^ y09 ^ y12 ^ y14),
        (y01 ^ y02 ^ y06 ^ y07 ^ y09 ^ y11 ^ y12) << 0x18
      | (y00 ^ y03 ^ y06 ^ y07 ^ y08 ^ y10 ^ y13) << 0x10
      | (y00 ^ y03 ^ y04 ^ y05 ^ y09 ^ y11 ^ y14) << 0x08
      | (y01 ^ y02 ^ y04 ^ y05 ^ y08 ^ y10 ^ y15),
    )


def _rotr128(v: tuple[int, ...], n: int) -> tuple[int, int, int, int]:
    val = (v[0] << 96) | (v[1] << 64) | (v[2] << 32) | v[3]
    mask = (1 << 128) - 1
    n %= 128
    val = ((val >> n) | (val << (128 - n))) & mask
    return (
        (val >> 96) & 0xFFFFFFFF,
        (val >> 64) & 0xFFFFFFFF,
        (val >> 32) & 0xFFFFFFFF,
        val & 0xFFFFFFFF,
    )


def _fo(d: tuple[int, ...], rk: tuple[int, ...]) -> tuple[int, int, int, int]:
    return _diff_layer(_sub_layer_odd(_xor128(d, rk)))


def _fe(d: tuple[int, ...], rk: tuple[int, ...]) -> tuple[int, int, int, int]:
    return _diff_layer(_sub_layer_even(_xor128(d, rk)))


def _aria_key_schedule(key: bytes):
    key_len = len(key)
    if key_len == 16:
        num_rounds = 12
        ck1, ck2, ck3 = _CK1, _CK2, _CK3
    elif key_len == 24:
        num_rounds = 14
        ck1, ck2, ck3 = _CK2, _CK3, _CK1
    elif key_len == 32:
        num_rounds = 16
        ck1, ck2, ck3 = _CK3, _CK1, _CK2
    else:
        raise ValueError(F'Invalid key size: {key_len}')

    padded = key + b'\x00' * (32 - key_len)
    kl = struct.unpack('>4I', padded[:16])
    kr = struct.unpack('>4I', padded[16:32])

    w0 = kl
    w1 = _fo(w0, ck1)
    w1 = _xor128(w1, kr)
    w2 = _fe(w1, ck2)
    w2 = _xor128(w2, w0)
    w3 = _fo(w2, ck3)
    w3 = _xor128(w3, w1)

    ek: list[tuple[int, int, int, int]] = []

    for n in (19, 31, 67):
        ek.append(_xor128(w0, _rotr128(w1, n)))
        ek.append(_xor128(w1, _rotr128(w2, n)))
        ek.append(_xor128(w2, _rotr128(w3, n)))
        ek.append(_xor128(w3, _rotr128(w0, n)))

    ek.append(_xor128(w0, _rotr128(w1, 97)))

    if num_rounds >= 14:
        ek.append(_xor128(w1, _rotr128(w2, 97)))
        ek.append(_xor128(w2, _rotr128(w3, 97)))

    if num_rounds >= 16:
        ek.append(_xor128(w3, _rotr128(w0, 97)))
        ek.append(_xor128(w0, _rotr128(w1, 109)))

    return ek[:num_rounds + 1], num_rounds


class ARIA(BlockCipher):
    """
    Pure-Python ARIA block cipher implementation.
    """
    block_size = 16
    key_size = frozenset({16, 24, 32})

    _ek: list[tuple[int, int, int, int]]
    _dk: list[tuple[int, int, int, int]]
    _num_rounds: int

    @property
    def key(self):
        return self._key_bytes

    @key.setter
    def key(self, key: bytes):
        self._key_bytes = key
        ek, nr = _aria_key_schedule(key)
        self._ek = ek
        self._num_rounds = nr
        # Decryption round keys: dk[i] = A(ek[N-i]) for 1..N-1, dk[0] = ek[N], dk[N] = ek[0]
        self._dk = dk = [ek[nr]]
        for i in range(1, nr):
            dk.append(_diff_layer(ek[nr - i]))
        dk.append(ek[0])

    def block_encrypt(self, data: BufferType) -> BufferType:
        t = struct.unpack('>4I', bytes(data))
        ek = self._ek
        nr = self._num_rounds
        for i in range(nr - 1):
            if i % 2 == 0:
                t = _fo(t, ek[i])
            else:
                t = _fe(t, ek[i])
        # Last round: no diffusion layer
        if (nr - 1) % 2 == 0:
            t = _xor128(_sub_layer_odd(_xor128(t, ek[nr - 1])), ek[nr])
        else:
            t = _xor128(_sub_layer_even(_xor128(t, ek[nr - 1])), ek[nr])
        return struct.pack('>4I', *t)

    def block_decrypt(self, data: BufferType) -> BufferType:
        t = struct.unpack('>4I', bytes(data))
        dk = self._dk
        nr = self._num_rounds
        for i in range(nr - 1):
            if i % 2 == 0:
                t = _fo(t, dk[i])
            else:
                t = _fe(t, dk[i])
        if (nr - 1) % 2 == 0:
            t = _xor128(_sub_layer_odd(_xor128(t, dk[nr - 1])), dk[nr])
        else:
            t = _xor128(_sub_layer_even(_xor128(t, dk[nr - 1])), dk[nr])
        return struct.pack('>4I', *t)

    def __init__(self, key: BufferType, mode: CipherMode | None):
        super().__init__(key, mode)