Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
???
04/10/11 11:06
Read: times


 
#181798 - Doesn't matter
Responding to: ???'s previous message
You think too much about a bit being more often zero or more often one. That doesn't really matter. You can normalize your randomness. But to be able to perform any normalization, you first have to have bits that does "toss the coin". Even coins that gives 99% heads up are suitable for creating randomness - it just takes more throws to get one bit of randomness while a perfectly matched coin would give one bit/toss.

Note that most encryption don't need high amounts of continuous entryphy. For AES, you get good encryption with a 128-bit key, and extreme encryption with a 256-bit key. With public-key encryption on the other hand, you need huge keys because of the design. When you have your initial AES key, you can manage quite well before the need to replace the AES key with a new one, just to limit the amount of information that gets encrypted with a single key.

But if you can get 1 bit per 100 bytes, then a 32kB RAM can give 320 bits of high-quality entrophy - too much for anyone to scan through. And the article is not about encryption. It's about getting a good seed for a pseudo-random generator. For lots of applications, it's enough that you can turn on a device 100 times, and get it to perform differently every time. For a game, it's enough that you get enough entrophy that players can't learn the sequences and beat their opponents by being able to figure out what next choice offered by the game will be.

For systems that needs large amounts of entrophy, it's obviously better to work with noise since only the bandwidth and sensitivity of the device will limit the amount of entrophy that can be produced every second. But the suggestion here is to get a reasonable amount of entrophy without adding extra hardware. It works, as long as people aren't foolish enough to think they can take the first 16 bytes of RAM contents and get 128 bits of perfect entrophy. But even something as simple as a 32-bit CRC evaluated over a reasonable amount of the RAM will produce a quite good entrophy for seeding a pseudo-random generator. With a large RAM and md5, you will get an even better entrophy. Both alternatives will manage well irrespective of bits being most often zero or most often one. Each non-fixed bit will permuate the bits of the output.

List of 15 messages in thread
TopicAuthorDate
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      

Back to Subject List