- From: Mary Ellen Zurko <mzurko@us.ibm.com>
- Date: Fri, 23 Jan 2009 08:55:38 -0500
- To: "Francois Daoust <fd" <fd@w3.org>
- Cc: Web Security Context Working Group <public-wsc-wg@w3.org>
Received on Friday, 23 January 2009 13:56:18 UTC
> 1. What are the main dangers associated with the use of hashed > credentials? Identity spoofing? Hashing doesn't ensure a unique value, does it? So I presume in the backend there's some hash table that tracks state and deals with conflicts by trying some new has. I thought the security properties of hashes were that it would be hard to find a second text that hashes to the same value. Not sure how hard it is to come up with something that hashes to some randomly useful identity. Seems like a danger to me, but ianac. > 2. Are there practical recipes to avoid the dangers (e.g. "encrypt the > client's IP address in the hashed credentials to ensure they cannot be > used by some other client"?) Just encrypt the identity then. Be sure to avoid standard cryptographic mistakes, and protect against reply and changes. Oh wait, that's what that SSL does :-). > 3. Can we consider it a good practice? In some not-highly-sensitive > cases, e.g. for applications that use identity to personalize the > look-and-feel? Never? Well if your security model is "nbd", then it sounds fine to me! But then, which use any cryptography at all?
Received on Friday, 23 January 2009 13:56:18 UTC