Re: [quicwg/base-drafts] Saving a round-trip time in the initial handshakes with retry (#2552)

[Erik Sy <[e-mail]>]

A QUIC server that is not validating the source address of a client before proceeding with the connection establishment is a vulnerable system. Adversaries can easily launch attacks to exhaust the server's resources or exploit the server within reflection attacks against other endpoints. 
QUIC deploys mitigations such as limiting the response size. However, it is unclear at this point in time whether these defenses are sufficient.
As the Internet is a hostile place, we will observe that QUIC servers have to make use of stateless retries to validate the client’s source address. However, a trustworthy prediction of the number of connection establishments using a stateless retry seems currently not to be feasible.

The proposal allows compensating for the costs of preventing IP address spoofing across connections to several hostnames. Thus, it provides an improved security versus performance tradeoff compared to the status quo.

As the probability of a stateless retry seems to be higher for resumed TLS sessions, session resumption across SNI values provides another incentive to implement this proposal.

If we agree, that this presents a useful feature for QUIC’s connection establishment, then I prefer to mark it quicv1. The reason is, that it allows saving round-trip during the establishment of initial connections even when the same security requirements are applied as TLS over TCP does. Thus, it provides an additional incentive to deploy QUIC instead of using TLS over TCP.

-- 
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/2552#issuecomment-478550385