~skeeto/public-inbox

Re: Prospecting for Hash Functions (The "theoretical bias limit (bias = ~0.021)" for 32-bit hash functions isn't correct)

Details
Message ID
<FaYUy1VicRNUj9uJlhMuZ6k3mYpv4n2Y5Wh3Jf6Ar9eV750-AMFltNH2Pd1kiQrQYZJvixiY9gUQQlHgpoQXSU0VNvYvVPitUmImwelyuO8=@protonmail.com>
DKIM signature
pass
Download raw message
Hi, I've used your hash-prospector project quite a lot recently,
and I wanted to know the theoretical bias limit for other powers of two.

While working on this I found out that the bias limit of `~0.021`
for 32-bit hash functions is incorrect and doesn't reflect reality.

I tested a totally random hash function that doesn't depend on the hash input
and just returns the output from an RNG (RomuQuad).

I also created an issue in the GitHub repository which explains this further
and offers a working method of calculating the theoretical bias limit:
https://github.com/skeeto/hash-prospector/issues/12

Best regards.
Reply to thread Export thread (mbox)