Integrated crypto

Robert Connolly robert at linuxfromscratch.org
Thu Apr 8 14:22:02 PDT 2004


On April 6, 2004 02:38 pm, Adrian Fisher wrote:
> Why not use CPU speed clocks?  I know they are known but they very
> rarely match the quoted speed (this system is a P4 2.8 GHz system and
> runs at different speeds regularly (sometimes faster, sometimes slower)
> depending on the temperature of the room, what it is doing, etc.
> Another option is to make use of the speed of cache and memory, while
> these figures can be gained by the system they can be incorporated into
> an algorythym that masks them.
>
> Just an idea.

How much do they vary? if it goes up or down by 1.00 degrees there would only 
be about 100 posibilities, if my math is right. It might help though; 
personaly I think hardware entropy should be sent to /dev/random. The frandom 
guy seems to have a good solution, but he says he's busy, might be a couple 
weeks before he can finish it. Another problem with using cpu temp is that it 
repeats.




More information about the hlfs-dev mailing list