On Fri, 2016-08-05 at 10:04 +0200, Niels Möller wrote:
> Nikos Mavrogiannopoulos <[email protected]> writes:
> 
> > 
> > Yes. Although if this is only for the versions prior to using the
> > prepare function, this is not a significant threat (the private key
> > computations are typically done on trusted values by the server).
> One scanario is a web hosting provider that handles private server
> keys
> provided by untrusted customers. No idea how common that is, but one
> wouldn't want one customer to crash the webserver also used by
> others.

right.

> > What is more important for older versions of gnutls are the public
> > key
> > operations such as ecdsa_verify(), dsa_verify() and rsa_encrypt().
> 
> Now I'm confused, I hope I didn't introduce any mpz_powm_sec calls on
> the code paths operating on public keys only? I don't think we have
> to care too much about obscure use cases where the supposedly public
> exponent actually needs to be well protected.

Correct, it seems I replied prior to thinking.

regards,
Nikos

_______________________________________________
nettle-bugs mailing list
[email protected]
http://lists.lysator.liu.se/mailman/listinfo/nettle-bugs

Reply via email to