> conveying a cryptographic hash of
> the password from the rendered Blackboard application login form in the
> browser to the Blackboard server (as in, a hash instead of the actual
> password).  Then the server validates that the hash is valid.

You've just described a client-side HMAC implementation, if I
understood correctly, which has no additional security value over a
plaintext password assuming no further computation happens to the
credential on the server other than comparison with a known value.  In
either case the value sent over the wire is sufficient alone to
perform authentication.  I suppose if this is an enterprise credential
not having it stored in the clear on the client might have security
value to _other_ applications.  In that case, however, I'd argue that
something is horribly wrong with your browser if it stores cleartext
passwords.

> I tend to advise against this kind of fanciness as beside the point. > Either 
> you trust SSL or you don't, and if you don't, the implications are
> really bad in ways that a little hashing of passwords aren't going to fix.

+1

> I can state from personal experience that this kind of thing
> discourages (but doesn't completely prevent!) integration via credential
> replay through the login form.

I'm fairly confident that the present implementation of the LT
parameter on the login form is sufficient to guard against replay
attacks.  I invite folks to play with 3.4.10 and see if they can
demonstrate a reasonable replay attack.

M

-- 
You are currently subscribed to [email protected] as: 
[email protected]
To unsubscribe, change settings or access archives, see 
http://www.ja-sig.org/wiki/display/JSG/cas-dev

Reply via email to