Re: [quicwg/base-drafts] Long Header Packets and Routing Connection IDs (#2834)

[Kazuho Oku <[e-mail]>]

@marten-seemann 
>>> My inclination is that server infrastructure SHOULD NOT route based on the CID in an Initial or 0-RTT packet, even if the client has switched to the one you gave it
>> 
>> needs "unless the server has a way to test if CID is server-generated."
> 
> This check only works if you assume that a client's connection ID is generated randomly, right?
> This assumption doesn't hold in the case of an attacker, who, as I've pointed out earlier, could reuse a connection ID that it obtained on an earlier connection to target the specific backend server it was connected to.

It is true that an attacker can forge a CID it has seen previously, and reuse that to mount DoS attacks. 

However, the issue is not related to long header packets (or how a connection is assigned to a particular node behind a load balancer). An attacker can for example, forge a CID of a short header packet and create multiple copies of short header packets.

-- 
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/2834#issuecomment-505229947