Jean-Marc Desperrier wrote:... >Anders Rundgren wrote: >>[...] >> There is though a general weakness in schemes that do not terminate >> the client-side in the SSL channel [...] >> this deficiency is eliminated by >> "targeting" the client-side operation for the site and certificate in >> the server-end. By doing that the receiver (server), can immediately >> detect if the operation has gone through a phishing proxy or not. It is >> an extra test to do but rather simple.
>They are indeed tricks to enhance the situation and try to detect that >the currently connected client is subject to a MITM attack, but it >doesn't resolve the problem entirely. It is still theoretically possible >for the MITM attacker to adapt to this and to modify on the fly the >dynamic content that implements the check to circumvent it. Since the attacker, at least if operating on the network level, will not be able to create a signature associated with the subject, I do not see much utility of that exploit. Or is there something I have missed here? I don't claim having all the answers in this matter... >You could take all weaknesses and solve them one by one, but >you'd end up having implemented a new secure channel on top >of the HTTP connexion. Yes, that is the goal. I believe this is what CardSpace (InfoCards) does as well. Although I like standards, SSL-client-authentication as implemented in for example MSIE is user-unfriendly. There seems to be no guidelines in this area. According to people with more explicit knowledge of the inner workings of SSL that I have, browsers cache passwords and keys to not force reauthentication. Password caching is not needed for the alternative solutions that use a session cookie as the user key. The session cookie stuff also makes logout very straightforward. This appears to be less simple using SSL-client-authentication. That is, if SSL-client-authentication is going to be the sole way of authenticating with a certificate, I believe the browser wendors should do something about it. http://www.w3.org/2005/Security/usability-ws/Overview Anders Rundgren _______________________________________________ dev-tech-crypto mailing list dev-tech-crypto@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-tech-crypto
