5 ms·
Why not seed a few more things in to the random numbers like temperature and noise? Both should be hard to measure exactly the same outside the casing, hard to
by anotheryou 8y ago
Why not seed a few more things in to the random numbers like temperature and noise?
Both should be hard to measure exactly the same outside the casing, hard to control and when done right should only add randomness.
Bringing heaters and timed, stealthy noise-generators should be difficult :) Also past noise should influence the future randomness.
- hashkb 8y agoYou could measure those things someplace far away (secret), or in a random office underneath the casino.
- function_seven 8y agoIf you’re taking the LSBs of those readings, measuring it using another instrument at any distance from the machine would be useless. (Audio) noise and temperature are very noisy inputs.
- maxerickson 8y agoThe phone is recovering the internal state of the PRNG, different initialization wouldn't stop that.
- LeoPanthera 8y agoThis is certainly not my field of expertise, but I would guess that regularly re-seeding the RNG with environmental information would prevent that attack.
- alkonaut 8y agoIt’s trivial to make it impossible to predict the machine output. Just source even a low quality entropy source. But now take that machine and source code and prove to gambling authorities what its properties are. That’s now almost impossible instead. You have made a machine that may behave differently under evaluation.
- smallstepforman 8y agoThe affected PRNG (90's design) was reseeded every 100ms and only had 96 bit resolution. These days the PRNG is reseeded every 10 ms (+-5ms sliding interval) and use a minimum of 1024 bit resolution.