| ??? 04/10/11 02:26 Read: times |
#181795 - Don't think 1:1 mapping Responding to: ???'s previous message |
Jan Waclawek said:
So, to say, "SRAMs are good sources of physical randomness" is the same as to say "throwing dice is a good source of physical randomness", without taking into account that dice are too often irregular and biased, and that an incorrect method of throwing them may lead to less than random results. Note that you don't need every byte of the SRAM to be a good source of randomness. Even if you only get 1 bit of randomness for every 100 bytes, then you still get 20 bits from a 2kB large SRAM. That is still a quite good result. My earlier experiments with SRAM showed that quite a lot of cells tended to always turn on in a specific state because of internal tolerances. But as long as not all memory cells gets in the same state (zero or one) every time, then there must be cells that are balanced enough to flip either direction. As already noted, they can have very good retention when powered off, so it may not be quick to toggle the power for extra random data. With charges remaining, it doesn't matter if a flip-flop is more likely to start up with a zero or a one - it will start up with the same state it had before the power off. |
| Topic | Author | Date |
| Truly Random Number Generator | 01/01/70 00:00 | |
| Latency Time Problem | 01/01/70 00:00 | |
| this is bad | 01/01/70 00:00 | |
| Don't think 1:1 mapping | 01/01/70 00:00 | |
| understanding | 01/01/70 00:00 | |
| Doesn't matter | 01/01/70 00:00 | |
| Yeah, yeah!! | 01/01/70 00:00 | |
| Way more than 3 | 01/01/70 00:00 | |
| baloney | 01/01/70 00:00 | |
| So easy to make assumptions and crash and burn | 01/01/70 00:00 | |
| Missing the point! | 01/01/70 00:00 | |
| Its just soooo wrong | 01/01/70 00:00 | |
| Randomness - NOT | 01/01/70 00:00 | |
| The key point is | 01/01/70 00:00 | |
Johnson noise versus zener noise... | 01/01/70 00:00 |