From: Carl Ellison <cme@acm.org>

...

...

...

Steve Shear <azur@netcom.com> writes:

[much cut]

...

I've been charged with developing an Internet service which needs to assure its clients of anonymity. However, we fear some clients may abuse the service and we wish to prevent the abusers from re-enrollment if terminated for misbehavior. (In your example, it would be the person(s) trying to discover the service host via flood).

My thought was to base enrollment on some sort of 'blinding' of their certified signature (e.g., from Verisign) which produces a unique result for each signature but prevents the service from reconstructing the signature itself (and thereby reveal the client's identity). I'm calling this negative authentication.

The mistake is to think of using ID certificates (like those from Verisign) in the first place. They don't mean anything.

You want an authorization certificate, such as produced by SPKI. You need to know what a key is authorized to do, not what name is associated with the key.

(Sorry about quoting so much, but I liked Steve Shear's succinct problem statement.) I don't see how authorization certificates solve this problem. How would you determine if someone was qualified to receive an authorization certificate? And what would you do to make them stop using the service if they abuse it, and to stop them from getting new authorization certificates? Thanks, Hal