CVE-2014-3466: GnuTLS buffer overflow

[anon1385]

https://bugzilla.redhat.com/show_bug.cgi?id=1101932

[rikacomet]

If I remember correctly, something similar demonstrated by someone(a bounty hunter) a few months back, and was featured on HN front page as well.

I remember it was maybe related to Facebook, and not to TLS/SSL specifically. Very similar.. sending excessively long session id values.

I wonder if excessively long session id values can break something else as well?

[farness → rikacomet]

Excessively long data is a cornerstone of security vulnerabilities.

> I wonder if excessively long session id values can break something else as well?

Yes, with p~=1.

[rikacomet → farness]

Indeed it is, but what I was curious to know more about the particular case of session IDs.

[azet]

PoC (not weaponized and ugly code due to lack of time): https://github.com/azet/CVE-2014-3466_PoC

hf.

[0x0]

At least it's a client only vulnerability. Hope your servers don't make outbound connections! :)

[SoftwareMaven → 0x0]

Because clients don't get re-routed through malicious redirects to unexpected servers. The consolation is that GnuTLS isn't used for popular web browsers.