[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ms-cryptoapi
Subject:    Re: "guts" questions
From:       Bob Denny <rdenny () DC3 ! COM>
Date:       1996-09-18 13:52:40
[Download RAW message or body]


Yeah, that stuff looks good. One important aspect of generating randomness is
to measure, or at least conservatively estimate, the entropy of the "witches
brew". To generate a clean 1024 bit RSA keypair, you need at least 1024 bits
of entropy. Any less generates an effectively smaller keypair (as far as
cryptanalysts are concerned). The sorts of numbers you listed are random, but
you need to know how much entropy is generated per unit time.

What if you read the net interface bytes/sec read and written a few times
while there is no net traffic? See what I mean? It's very tricky to generate
randomness from numbers like that. It's easier to measure the CHANGE in such
numbers over a time interval, and the number of bits changed. This will come
closer to an entropy estimate.

  -- Bob

On Sep 17, Scott Field <sfield@MICROSOFT.COM> wrote:
> Subject: Re: "guts" questions
> I should provide some more background on what's available via the
> performance counters, to clarify my original statement:
>
> Cache: hits/sec, misses/sec
> Network interface: bytes sent/sec, bytes recd/sec
> Processor: CPU load, Interrupts/sec
> Logical disk: bytes transfer/sec
> Server: files open, logon count
> Thread: context switches/second
> Process: handle counts, private bytes
> Memory: page faults/sec

[prev in list] [next in list] [prev in thread] [next in thread]