RE: Getting our definitions of encryption straight for the HTTP/2 security discussion

> Greetings again. Over the past weeks, people are sometimes talking past each other when they say they want to "always encrypt" HTTP/2 traffic. In specific, many people have used the term "opportunistic encryption" in very different ways without knowing it.
>
> To help people at least understand what each other might be saying in the future, I created a page with some definitions that hopefully everyone can use. Comments are welcome.
>
> http://trac.tools.ietf.org/wg/httpbis/trac/wiki/encryption-definitons

> --Paul Hoffman


“Authenticated Encryption” is a poor choice of term for server-authenticated TLS. “Authenticated Encryption” is already a well-defined term of art in the crypto community with a different meaning. There is even an Authenticated Encryption RFC [RFC 5116].

Perhaps better terms would be:
* Server-authenticated encryption — client authenticates server’s identity
* Better-than-nothing encryption — no server authentication 
* Best-effort encryption — client tries server-authenticated encryption, but falls back to better-than-nothing encryption if necessary
* Opportunistic encryption — client tries best-effort encryption, but falls back to no encryption if necessary

P.S. Better-than-nothing security (BTNS) is an unauthenticated mode of IPsec defined in RFC 5386.

--
James Manger

Received on Wednesday, 20 November 2013 23:55:08 UTC