Re: [P2PSIP] Re: HIP pros and cons

[Eric Rescorla <[e-mail]>]

At Fri, 21 Dec 2007 11:51:02 +0200 (EET),
Miika Komu wrote:
> 
> On Tue, 18 Dec 2007, Eric Rescorla wrote:
> 
> Eric,
> 
> > At Tue, 18 Dec 2007 19:45:49 +0200 (EET),
> > Miika Komu wrote:
> >>
> >> On Tue, 18 Dec 2007, Brian Rosen wrote:
> >>
> >> Hi Brian,
> >>
> >> thanks for the feedback. It seems like the alternative b is mandatory for
> >> p2p-sip. Otherwise, then I don't see how you are going to avoid the use of
> >> cryptographic certificates without compromising dht identifier security
> >> and revocation.
> >>
> >> In alternative b, it is still possible to use host generated dht key
> >> identifiers.
> >
> > Again, not if you want to use them as the locations in the DHT,
> > because then you get chosen location attacks.
> 
> Yes but it costs. I don't see the difference here with host vs. enrollment 
> server generated keys because the end-host can enroll several times until 
> it is satisfied with the id. What makes the difference is the cost.

For the nth time, yes, but it requires a separate query to the 
enrollment server, which the enrollment server can throttle.
This doesn't work if the enrollment server lets the peer
propose the peer-id, since the peer can mount an offline 
attack on the location. Again, this is all in my slides from
YVR.


> Since I took the time to describe a detailed, step-by-step approach with 
> host generated identifiers, could you do the same with enrollment 
> server generated keys? Thanks.

I don't propose using enrollment server generated keys. I propose
separating cryptographic keys from peer-ids and using certificates
to bind them together. This is described in fairly complete detail
draft-bryan-p2psip-reload-02.


> P.S. If you want to reference some work, please give the authors and 
> titles to the publication or URL instead of some acronyms. Thanks.

I generally try to. If there's a particular one that wasn't clear
lemme know and I'll send you a full reference.

-Ekr


_______________________________________________
P2PSIP mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/p2psip