Re: [TLS] Encrypting record headers: practical for TLS 1.3 after all?

[Aaron Zauner <[e-mail]>]

PS:

Aaron Zauner wrote:
> No it's not. It's a very short presentation from a TLS-WG interim
> meeting. The threat-model concerns Akamai's (and other's) current and -
> possibly - future use of TLS. We're not trying to build an Onion routing
> protocol. Given the FUD on the Tor dev list, this is a good thing. While
> the presentation might have flaws from the perspective of an Onion
> routing protocol developer, it reflects the point of view of a lot of
> people/companies on this list, I assume.
> 

I don't think traffic analysis is in the treat model for TLS proper. If
we wanted to circumvent traffic analysis we'd have to introduce noise
and randomness (Pond does a good job there using Tor and other
mechanisms). I don't see how we can engineer a low-latency (now even
0-RTT) network security protocol that will do that in a performant
manner. When time comes and people have 10-40-100GE at home, maybe.
Infiniband would be nice. But that will still leave out use for 3rd
world countries (which still run on XP anyway). This is a technical list
and we should keep politics and FUD aside as best as possible.


Greetings from a very empty Sinai peninsula,
(A)aron