Messages in this thread | | | Date | Mon, 24 Jan 2000 17:40:17 -0500 | From | Sandy Harris <> | Subject | Re: Intel 810 Random Number Generator |
| |
Gregory Maxwell wrote:
> In this case, we can play the paranoid and just mix it in and not > increase the entropy count. > > If the current 'twisted crc32 mixer thingy' doesn't provide that your > entropy can never go down even with NSA input data, then we need ones > that does.
Any reversible binary operation -- a XOR b, a + b, encrypt a with key b, ... -- provides that.
The attacker doesn't know a before the operation. Afterward, if he knows b and the result and the operation is reversible, then he can recover a. Therefore the result contains at least as much information unknown to the attacker (=at least as much entropy) as a did.
The point of the twisting is to spread the incoming entropy around the pool a bit. Down at the lowest level, it uses XOR to mix so it cannot reduce entropy, whatever the input data.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |