Re: [TLS] proposal to encrypt ContentType for TLS 1.3

[Colm MacCárthaigh <[e-mail]>]

On Mon, Jul 14, 2014 at 9:10 AM, Martin Rex <[e-mail]> wrote:
> I meant to say that it is a feature of TLS to be able to distinguish
> handshake phase from application data phase on the outside.

During the week of heartbleed, I wrote a silly kernel module that does
this. I've just put it on GitHub for reference:

https://github.com/colmmacc/nf_conntrack_tls

I wouldn't recommend actually using the module, as it was an emergency
temporary measure borne out of necessity and in general; moving input
processing risks to the kernel is an especially bad idea. But I've
uploaded it as an example of where having the record type in the plain
did help; it allows a middle-layer to do sufficient protocol
validation to protect against a record-level flaw in the application's
record processing (such as Heartbleed).

Even having done that, this still seems like a mis-feature; especially
considering how bad leaking the alert type has been over the years.

-- 
Colm