tfhe-csprng 0.9.0

Cryptographically Secure PRNG used in the TFHE-rs library.
Documentation
use crate::generators::aes_ctr::{
    AesBlockCipher, AesKey, AES_CALLS_PER_BATCH, BYTES_PER_AES_CALL, BYTES_PER_BATCH,
};
use std::arch::x86_64::{
    __m128i, _mm_aesenc_si128, _mm_aesenclast_si128, _mm_aeskeygenassist_si128, _mm_shuffle_epi32,
    _mm_slli_si128, _mm_store_si128, _mm_xor_si128,
};
use std::mem::transmute;

/// An aes block cipher implementation which uses `aesni` instructions.
#[derive(Clone)]
pub struct AesniBlockCipher {
    // The set of round keys used for the aes encryption
    round_keys: [__m128i; 11],
}

impl AesBlockCipher for AesniBlockCipher {
    fn new(key: AesKey) -> AesniBlockCipher {
        let aes_detected = is_x86_feature_detected!("aes");
        let sse2_detected = is_x86_feature_detected!("sse2");

        if !(aes_detected && sse2_detected) {
            panic!(
                "The AesniBlockCipher requires both aes and sse2 x86 CPU features.\n\
                aes feature available: {aes_detected}\nsse2 feature available: {sse2_detected}\n\
                Please consider enabling the SoftwareRandomGenerator with the `software-prng` feature",
            )
        }

        // SAFETY: we checked for aes and sse2 availability
        let round_keys = unsafe { generate_round_keys(key) };
        AesniBlockCipher { round_keys }
    }

    fn generate_batch(&mut self, data: [u128; AES_CALLS_PER_BATCH]) -> [u8; BYTES_PER_BATCH] {
        #[target_feature(enable = "sse2,aes")]
        unsafe fn implementation(
            this: &AesniBlockCipher,
            data: [u128; AES_CALLS_PER_BATCH],
        ) -> [u8; BYTES_PER_BATCH] {
            si128arr_to_u8arr(aes_encrypt_many(
                u128_to_si128(data[0]),
                u128_to_si128(data[1]),
                u128_to_si128(data[2]),
                u128_to_si128(data[3]),
                u128_to_si128(data[4]),
                u128_to_si128(data[5]),
                u128_to_si128(data[6]),
                u128_to_si128(data[7]),
                &this.round_keys,
            ))
        }
        // SAFETY: we checked for aes and sse2 availability in `Self::new`
        unsafe { implementation(self, data) }
    }

    fn generate_next(&mut self, data: u128) -> [u8; BYTES_PER_AES_CALL] {
        unsafe { transmute(aes_encrypt_one(u128_to_si128(data), &self.round_keys)) }
    }
}

#[target_feature(enable = "sse2,aes")]
unsafe fn generate_round_keys(key: AesKey) -> [__m128i; 11] {
    let key = u128_to_si128(key.0);
    let mut keys: [__m128i; 11] = [u128_to_si128(0); 11];
    aes_128_key_expansion(key, &mut keys);
    keys
}

#[inline(always)]
fn aes_encrypt_one(message: __m128i, keys: &[__m128i; 11]) -> __m128i {
    unsafe {
        let mut tmp_1 = _mm_xor_si128(message, keys[0]);

        for key in keys.iter().take(10).skip(1) {
            tmp_1 = _mm_aesenc_si128(tmp_1, *key);
        }

        _mm_aesenclast_si128(tmp_1, keys[10])
    }
}

// Uses aes to encrypt many values at once. This allows a substantial speedup (around 30%)
// compared to the naive approach.
#[allow(clippy::too_many_arguments)]
#[inline(always)]
fn aes_encrypt_many(
    message_1: __m128i,
    message_2: __m128i,
    message_3: __m128i,
    message_4: __m128i,
    message_5: __m128i,
    message_6: __m128i,
    message_7: __m128i,
    message_8: __m128i,
    keys: &[__m128i; 11],
) -> [__m128i; 8] {
    unsafe {
        let mut tmp_1 = _mm_xor_si128(message_1, keys[0]);
        let mut tmp_2 = _mm_xor_si128(message_2, keys[0]);
        let mut tmp_3 = _mm_xor_si128(message_3, keys[0]);
        let mut tmp_4 = _mm_xor_si128(message_4, keys[0]);
        let mut tmp_5 = _mm_xor_si128(message_5, keys[0]);
        let mut tmp_6 = _mm_xor_si128(message_6, keys[0]);
        let mut tmp_7 = _mm_xor_si128(message_7, keys[0]);
        let mut tmp_8 = _mm_xor_si128(message_8, keys[0]);

        for key in keys.iter().take(10).skip(1) {
            tmp_1 = _mm_aesenc_si128(tmp_1, *key);
            tmp_2 = _mm_aesenc_si128(tmp_2, *key);
            tmp_3 = _mm_aesenc_si128(tmp_3, *key);
            tmp_4 = _mm_aesenc_si128(tmp_4, *key);
            tmp_5 = _mm_aesenc_si128(tmp_5, *key);
            tmp_6 = _mm_aesenc_si128(tmp_6, *key);
            tmp_7 = _mm_aesenc_si128(tmp_7, *key);
            tmp_8 = _mm_aesenc_si128(tmp_8, *key);
        }

        tmp_1 = _mm_aesenclast_si128(tmp_1, keys[10]);
        tmp_2 = _mm_aesenclast_si128(tmp_2, keys[10]);
        tmp_3 = _mm_aesenclast_si128(tmp_3, keys[10]);
        tmp_4 = _mm_aesenclast_si128(tmp_4, keys[10]);
        tmp_5 = _mm_aesenclast_si128(tmp_5, keys[10]);
        tmp_6 = _mm_aesenclast_si128(tmp_6, keys[10]);
        tmp_7 = _mm_aesenclast_si128(tmp_7, keys[10]);
        tmp_8 = _mm_aesenclast_si128(tmp_8, keys[10]);

        [tmp_1, tmp_2, tmp_3, tmp_4, tmp_5, tmp_6, tmp_7, tmp_8]
    }
}

fn aes_128_assist(temp1: __m128i, temp2: __m128i) -> __m128i {
    let mut temp3: __m128i;
    let mut temp2 = temp2;
    let mut temp1 = temp1;
    unsafe {
        temp2 = _mm_shuffle_epi32(temp2, 0xff);
        temp3 = _mm_slli_si128(temp1, 0x4);
        temp1 = _mm_xor_si128(temp1, temp3);
        temp3 = _mm_slli_si128(temp3, 0x4);
        temp1 = _mm_xor_si128(temp1, temp3);
        temp3 = _mm_slli_si128(temp3, 0x4);
        temp1 = _mm_xor_si128(temp1, temp3);
        temp1 = _mm_xor_si128(temp1, temp2);
    }
    temp1
}

#[inline(always)]
fn aes_128_key_expansion(key: __m128i, keys: &mut [__m128i; 11]) {
    let (mut temp1, mut temp2): (__m128i, __m128i);
    temp1 = key;
    unsafe {
        _mm_store_si128(keys.as_mut_ptr(), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x01);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(1), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x02);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(2), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x04);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(3), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x08);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(4), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x10);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(5), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x20);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(6), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x40);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(7), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x80);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(8), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x1b);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(9), temp1);
        temp2 = _mm_aeskeygenassist_si128(temp1, 0x36);
        temp1 = aes_128_assist(temp1, temp2);
        _mm_store_si128(keys.as_mut_ptr().offset(10), temp1);
    }
}

#[inline(always)]
fn u128_to_si128(input: u128) -> __m128i {
    unsafe { transmute(input) }
}

#[allow(unused)] // to please clippy when tests are not activated
fn si128_to_u128(input: __m128i) -> u128 {
    unsafe { transmute(input) }
}

#[inline(always)]
fn si128arr_to_u8arr(input: [__m128i; 8]) -> [u8; BYTES_PER_BATCH] {
    unsafe { transmute(input) }
}

#[cfg(test)]
mod test {
    use super::*;

    // Test vector for aes128, from the FIPS publication 197
    const CIPHER_KEY: u128 = u128::from_be(0x000102030405060708090a0b0c0d0e0f);
    const KEY_SCHEDULE: [u128; 11] = [
        u128::from_be(0x000102030405060708090a0b0c0d0e0f),
        u128::from_be(0xd6aa74fdd2af72fadaa678f1d6ab76fe),
        u128::from_be(0xb692cf0b643dbdf1be9bc5006830b3fe),
        u128::from_be(0xb6ff744ed2c2c9bf6c590cbf0469bf41),
        u128::from_be(0x47f7f7bc95353e03f96c32bcfd058dfd),
        u128::from_be(0x3caaa3e8a99f9deb50f3af57adf622aa),
        u128::from_be(0x5e390f7df7a69296a7553dc10aa31f6b),
        u128::from_be(0x14f9701ae35fe28c440adf4d4ea9c026),
        u128::from_be(0x47438735a41c65b9e016baf4aebf7ad2),
        u128::from_be(0x549932d1f08557681093ed9cbe2c974e),
        u128::from_be(0x13111d7fe3944a17f307a78b4d2b30c5),
    ];
    const PLAINTEXT: u128 = u128::from_be(0x00112233445566778899aabbccddeeff);
    const CIPHERTEXT: u128 = u128::from_be(0x69c4e0d86a7b0430d8cdb78070b4c55a);

    #[test]
    fn test_generate_key_schedule() {
        // Checks that the round keys are correctly generated from the sample key from FIPS
        let key = u128_to_si128(CIPHER_KEY);
        let mut keys: [__m128i; 11] = [u128_to_si128(0); 11];
        aes_128_key_expansion(key, &mut keys);
        for (expected, actual) in KEY_SCHEDULE.iter().zip(keys.iter()) {
            assert_eq!(*expected, si128_to_u128(*actual));
        }
    }

    #[test]
    fn test_encrypt_many_messages() {
        // Checks that encrypting many plaintext at the same time gives the correct output.
        let message = u128_to_si128(PLAINTEXT);
        let key = u128_to_si128(CIPHER_KEY);
        let mut keys: [__m128i; 11] = [u128_to_si128(0); 11];
        aes_128_key_expansion(key, &mut keys);
        let ciphertexts = aes_encrypt_many(
            message, message, message, message, message, message, message, message, &keys,
        );
        for ct in &ciphertexts {
            assert_eq!(CIPHERTEXT, si128_to_u128(*ct));
        }
    }

    #[test]
    fn test_encrypt_one_message() {
        let message = u128_to_si128(PLAINTEXT);
        let key = u128_to_si128(CIPHER_KEY);
        let mut keys: [__m128i; 11] = [u128_to_si128(0); 11];
        aes_128_key_expansion(key, &mut keys);
        let ciphertext = aes_encrypt_one(message, &keys);
        assert_eq!(CIPHERTEXT, si128_to_u128(ciphertext));
    }
}