Hacker News Reader

Comment by caffeinewriter | Hacker News Reader

I'm really hesitant to even consider this without it being run through the Diehard Tests[0], since from my understanding, "True Randomness" should be cryptographically secure should this be used in a CSPRNG.

[0]: https://en.wikipedia.org/wiki/Diehard_tests

View on HN

rng-tools includes a command for testing random number generator output. I tried this with a FreeBSD port from https://github.com/waitman/rngtest

Results aren't great:

    rngtest: bits received from input: 300032
    rngtest: FIPS 140-2 successes: 0
    rngtest: FIPS 140-2 failures: 15
    rngtest: FIPS 140-2(2001-10-10) Monobit: 15
    rngtest: FIPS 140-2(2001-10-10) Poker: 15
    rngtest: FIPS 140-2(2001-10-10) Runs: 15
    rngtest: FIPS 140-2(2001-10-10) Long run: 0
    rngtest: FIPS 140-2(2001-10-10) Continuous run: 0
    rngtest: input channel speed: (min=236.690; avg=245.396; max=251.742)bits/s

And with more data:

After nearly 8 hours:

    rngtest: bits received from input: 6762248
    rngtest: FIPS 140-2 successes: 156
    rngtest: FIPS 140-2 failures: 182
    rngtest: FIPS 140-2(2001-10-10) Monobit: 182
    rngtest: FIPS 140-2(2001-10-10) Poker: 174
    rngtest: FIPS 140-2(2001-10-10) Runs: 160
    rngtest: FIPS 140-2(2001-10-10) Long run: 0
    rngtest: FIPS 140-2(2001-10-10) Continuous run: 0

Yikes.

A reference from the diehard tests page. The dieharder tests:

http://www.phy.duke.edu/~rgb/General/dieharder.php

A relatively recent (2013) library implementation of the diehard tests plus additional tests.

Some things to note:

The OP implementation is going to be VERY dependent on OS/language/etc. For example even if it works in C it won't necessarily work in JVM. It probably will not work on a microcontroller.

Reply on news.ycombinator.com