Re: Draft finding - "Transitioning the Web to HTTPS"

Ah. If that's the nature of the issue (and thank you for repeating it again
after Domenic had explained it) i.e. the issue is "whom to trust", then
that could go down all the way down (or up) to the physical world where you
have a human at the end of the trust chain. Let's say I trust the CA but
some state actor snuck in an employee into a critical position within the
CA or (I'm not an expert) in some other way we end up with a compromised
system because of a single person or because the adversary has the means to
undermine the system. Then it can be said that all security systems in
place are imperfect, and that issue transcends ethics, and that it's about
self interest. If it's in the interest of the most powerful nation on earth
to get some information they will get it. We all know that. So maybe the
orientation for the public (and historians) is that no system that exists
in the physical world can be 100% secure unless the information that is
being secured is never intended to be retrieved. Someone destroys the key.
Even then ... quantum computers etc. So I'm sorry to be exceedingly banal
but I think the communication problem around Web Security (and there is one
for sure) is to state clearly that No Amount of Security Will Ever Get You
Privacy From Powerful Adversaries. End of story. I mean maybe each document
about security should have that as a footnote.

I'm over my own limit for ridiculousness but do think the way the topic is
communicated should be inclusive of the fact that the foundation of
security on the web is only there for the average criminal, and that no web
communication can be hidden from state actors, at least not for longer than
it takes them to execute a targeted operation.

Hurray for humanity.

On Wed, Dec 17, 2014 at 3:36 PM, Chris Palmer <palmer@google.com> wrote:
>
> On Wed, Dec 17, 2014 at 8:46 AM, Marc Fawzi <marc.fawzi@gmail.com> wrote:
>
> > Btw, on a related subject, stuff like "signed scripts" which were
> proposed
> > on this list by an independent developer (with the conclusion being that
> > signing a script at least assures that it's not be altered) might be
> part of
> > a more perfect foundation. The argument I heard here against Web Crypto
> over
> > HTTP (or more comprehensively stuff like OpenPGP.js which used by Google
> for
> > its End-to-End security plugin) for client-to-server secure exchange is
> that
> > MITM can alter the script, but a signed script would solve that, so
> > regardless of whether you use a CA or not you should be able to get
> pretty
> > good privacy, right? (assuming signed scripts or signed Chrome/Firefox
> > hosted apps)
>
> Think about this. What would the root of trust for script signatures
> be? Perhaps script execution environments could be born with the
> public keys of trusted third parties that vouch for the identities of
> script authors...
>
>
> If you are referring to Sub-Resource Integrity (SRI), at least the
> top-level page that includes the resources has to be served over
> HTTPS, so that the SHA-256 hashes for the sub-resources are at least
> minimally trustworthy. So you haven't really avoided the secure
> transport requirement for WebCrypto.
>
> (Of course, I argue that even the sub-resources must be served over
> secure transport, even for/especially for SRI. But that's a whole
> other thread.)
>