Comments on: The Myth of Self-Signed Security http://www.adamfrisby.com/blog/2009/07/the-myth-of-self-signed-security/ ZOMGWTFHAI Tue, 15 Jun 2010 00:36:58 +0000 http://wordpress.org/?v=2.8.4 hourly 1 By: Will Dowling http://www.adamfrisby.com/blog/2009/07/the-myth-of-self-signed-security/comment-page-1/#comment-8315 Will Dowling Tue, 04 Aug 2009 03:17:41 +0000 http://www.adamfrisby.com/blog/?p=388#comment-8315 The interesting thing about sslsniff isn't that it's a self-signed certificate - it's actually a certificate signed by one of the trusted CA's, that has then had additional signatures put in the chain. Web browser implementations incorrectly check that there is *a* signature it trusts, but doesn't do full inspection of the certificate to begin with. The whole trust model is a bit broken to begin with anyway however, because there are really only two trust levels: absolute trust and none at all; and is it like your web browser actively checks certificate revocations from the CA? I'll give you a hint - they don't. I disagree that self-signed certificates are useless in general, but only if you already have a trusted copy of your CA's public certificate on your machine. For example, up until recently I used to self-sign my IMAP/SSL certificate because I can trust it once and explicitly remember the one you trust. This is how SSH works - hosts generate their host key once off and use that for all future communication: you build the trust when you first connect. Interesting projects for the doubters include the Australian based CACert, which offers web-of-trust style verification and an equivalent level of verification as low-end SSL providers. The only problem is that their CA certificate isn't trusted in any browsers or OS platforms - though there's a 10 year old ticket at Mozilla for changing this (and a huge debate surrounding it). Ultimately, whilst we can keep using bigger and bigger keys (to keep ahead of brute forcing by a comfortable margin) - the weakness with SSL is in how it's implemented. The interesting thing about sslsniff isn’t that it’s a self-signed certificate – it’s actually a certificate signed by one of the trusted CA’s, that has then had additional signatures put in the chain. Web browser implementations incorrectly check that there is *a* signature it trusts, but doesn’t do full inspection of the certificate to begin with.

The whole trust model is a bit broken to begin with anyway however, because there are really only two trust levels: absolute trust and none at all; and is it like your web browser actively checks certificate revocations from the CA? I’ll give you a hint – they don’t.

I disagree that self-signed certificates are useless in general, but only if you already have a trusted copy of your CA’s public certificate on your machine. For example, up until recently I used to self-sign my IMAP/SSL certificate because I can trust it once and explicitly remember the one you trust. This is how SSH works – hosts generate their host key once off and use that for all future communication: you build the trust when you first connect.

Interesting projects for the doubters include the Australian based CACert, which offers web-of-trust style verification and an equivalent level of verification as low-end SSL providers. The only problem is that their CA certificate isn’t trusted in any browsers or OS platforms – though there’s a 10 year old ticket at Mozilla for changing this (and a huge debate surrounding it).

Ultimately, whilst we can keep using bigger and bigger keys (to keep ahead of brute forcing by a comfortable margin) – the weakness with SSL is in how it’s implemented.

]]>