The reason the Session header mechanism I propose does not address cookie
scoping is that there is already a perfectly adequate confidentiality and
integrity solution in the hands of the Server: encrypt the cookie contents
and append a MAC to the end.

The only situation where I see cookie stealing as a problem is where the
cookie is a bearer token. Since the client never looks at cookie content
there is no good reason not to encrypt if the data is at all confidential.

Cookie overwriting is a little more complicated as merely being able to
detect overwriting is not the same as being able to prevent a denial of
cookie attack.

If people think the existing scope mechanism in cookies is insufficient, I
am open to extending the Session draft to add an origin cookie type
capability. Or we can redefine the cookie handling as well if that is
needed to ensure backwards compatibility.


On the issue of preventing replay attacks, the session continuation draft I
wrote describes two application scenarios. The first being the Web browser
scenario, the second being Web Services.

The Web service I have developed that uses the draft uses nonces passed in
the protocol to prevent replay attacks. Web Services typically require much
stronger replay attack prevention than Web Browsing. But this is easier to
achieve since some of the checking lives more naturally in the Web Service
protocol layer than the HTTP binding.

I don't see the need to prevent replay attacks in Web Browsing except in
the very special case of forms entry which servers can check themselves by
using a nonce passed in a hidden form field. I should probably add a
section to the draft explaining the reasoning there.
_______________________________________________
websec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/websec

Reply via email to