Re: [quicwg/base-drafts] Authenticating connection IDs (#3439)

[Christian Huitema <[e-mail]>]

@martinthomson it seems that we have agreement on two points a bit of discussion on the third:

1) The spec should describe the "munged retry" attack. It can be used to play games with server farms, either by attackers in the middle, or by attackers as client. It can produce a DOS attack of some sort despite the "retry" protection mechanisms.

2) The spec should describe how servers can protect themselves by tying the "Retry-CID" to the retry token, and in fact recommend that they do that.

3) Adding "retry-CID" as a transport parameter would give assurance to clients that the server is protected against the attack, and allow clients to detect attacks if the server is not protected.

I am not sure that (3) is really needed. I do not really understand the client stake there, except in the general terms of defeating attackers in the middle and encouraging good server implementations. OTOH, if the server does (2), it is not very onerous to also do (3).

But then, what prevents servers that do not care from taking shortcuts, not verify the token, and copy any incoming post-retry DCID in the transport parameter?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/3439#issuecomment-595508210