Nonce position and test vectors PRNG

dchest · 2016-05-29T23:42:47Z

Hello!

I'm porting New Hope to JavaScript and want to make sure my code behaves exactly the same and generates the same test vectors as reference implementation. I found the following nits:

In helprec nonce byte is put at the end 00000000B (https://github.com/tpoeppelmann/newhope/blob/master/ref/error_correction.c#L81), while in poly_getnoise it's in the beginning B00000000 (https://github.com/tpoeppelmann/newhope/blob/master/ref/poly.c#L105). Since nonce bytes are different anyway, I see no reason for this (well, other than to make sure it's really-really different).
PRNG in testvectors.c is very strange: it seems like there's confusion between uint32 (out) and unsigned char (x) https://github.com/tpoeppelmann/newhope/blob/master/ref/test/testvectors.c#L51 It kinda works, since randombytes always gets 32 bytes, but will write past the buffer if length is not divisible by 4. I propose replacing the whole construction with ChaCha-based PRNG:

static unsigned char key[32] = { 3,1,4,1,5,9,2,6,5,3,5,8,9,7,9,3,2,3,8,4,6,2,6,4,3,3,8,3,2,7,9,5 } ;
static unsigned char nonce[8];
static unsigned char out[64];
static int outleft = 0;

void randombytes(unsigned char *x,unsigned long long xlen)
{
  while (xlen > 0) {
    if (outleft == 0) {
      if (!++nonce[0]) if (!++nonce[1]) if (!++nonce[2]) ++nonce[3];
      crypto_stream_chacha20(out, 64, nonce, key);
      outleft = 64;
    }
    *x = out[--outleft];
    ++x;
    --xlen;
  }
}

Also, I could build reference implementation on OS X only if I removed -msse2avx -march=corei7-avx from CFLAGS. It seems like they are not needed for ref, so perhaps should be removed?

Thank you!

The text was updated successfully, but these errors were encountered:

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Nonce position and test vectors PRNG #1

Nonce position and test vectors PRNG #1

Nonce position and test vectors PRNG #1

Nonce position and test vectors PRNG #1

Comments

Uh oh!