Timing Cryptanalysis Attack

[Eric Young]

On Mon, 11 Dec 1995, Anonymous wrote:
> pck at netcom.com (Paul C. Kocher) writes:
> I just read this paper, and while it is somewhat interesting, I
> don't think the walls of cryptography are in any danger of
> crumbling.
...
> So while this is a very nice piece of work, and certainly of
> theoretical interest, I don't think it will modify the way in
> which people are advised to utilize cryptographic software, or
> cause companies like Netscape of RSADSI to shed any tears.

Read the SKIP spec (SKIP is Sun's IP level encryption protocol).  It uses
Diffle-Hellman certificates.  That means fixed secret DH keys being used
in routers.  It is hard to thing of a better target for this type of
attack.  I have not done a complete read of the SKIP specification (only a
quick scan) so I could be wrong about SKIP but DH certificates sound like
a very very bad idea.  The other source for attack would be any networked
service that is on a local network.  Single user machines are far better
targes than multi-user systems.  That Web server sitting idle not doing
much, repeatedly hit it with https requests and if you are on a local
network, you should be able to get very good timing information. 

I for one will probably add a flag for conditional compilation of my 
bignumber library so that it will take constant time.  This may be a %10 
slow down (using small windows exponentiation) which is trivial compared 
to the %30 speedup I will probably get when I implement a faster mod 
function :-).

eric
--
Eric Young                  | Signature removed since it was generating
AARNet: eay at mincom.oz.au    | more followups than the message contents :-)