Re: [quicwg/base-drafts] handling of coalesced packets with decryption errors creates DoS opportunity (#2308)

[Marten Seemann <[e-mail]>]

> Doesn't decryption failure happen also when the packet gets corrupt while being transmitted?

I assumed those to be **very** rare, so treating the whole datagram as corrupted might be a valid response. I don't have any numbers about that though. Maybe we can ask @ianswett if he can share some measurements?

> Does that matter, considering the fact that an attacker can force AEAD operation for every 21 bytes by using a short header packet (1 type + 4 byte space + 16 byte AEAD tag)?

The assumption here is that sending a lot of small packet is more expensive than sending one large packet of the same payload size. First, there's UDP and IP overhead that we're not counting here, and second (and probably more importantly), you only need a single syscall.

Looking at the numbers, I think I messed up my calculation of how many coalesced packets fit into a single datagram, since I failed to account for the AEAD tag. This brings the number QUIC packets per datagram down to about 35.

-- 
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/2308#issuecomment-451822218