The key pinning is a very good step in the direction of web security.
However, it does raises the operational bar for web sites and is not
helpful with new installations.

Yet the observation here is that the reliable security is especially
important for sites that authenticate the users. So why not to use the
passwords that users have to verify the certificates?

SRP [1] or J-PAKE [2] protocols allows *mutual* authentication of both
client and server based on the shared secret. RFC5054 [3] even
describes how to use SRP in TLS Authentication.

The problem with these protocols is that they assumes a side channel
for the distribution of the shared secret. In practice this translates
into supporting both certificate-based HTTPS for the initial user
registration and the new protocols for authenticated user sessions.
This is rather costly operationally and perhaps the reason while the
protocols remains outside mainstream browser/server implementations.

Hence the suggestion to allow to use SRP/J-PAKE *only* for certificate
verification. The idea is to provide hard to spoof yet customizable
GUI in the browser where the user can assume that entering a password
there is always safe. The browser then performs secure authentication
using ordinary HTTP PUT requests. If it is OK, it uses the newly
established encryption channel to verify the certificate for the
subsequent connection. After that a normal HTTPS connection resumes.

Compared with RFC5054 the key difference is that the server side in
this case can be implemented as a part of web application without any
changes in the server software/hardware to support the new protocol.
This is possible as the authentication happens only once per session
so computationally expensive J-PAKE/SRP exchange happens rarely and
deployment of accelerators can be avoided even on busy sites.

[1] http://srp.stanford.edu/
[2] http://en.wikipedia.org/wiki/Password_Authenticated_Key_Exchange_by_Juggling
[3] http://tools.ietf.org/html/rfc5054
_______________________________________________
websec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/websec

Reply via email to