The concluding discussion has a lot of great pointers to future work and
things the security community needs to work out. This part is relevant to
this community (and my favorite part of the paper):

Many HTTPS security features expect connections to be end-to-end by mixing
the HTTP and TLS layers, and by implementing HTTPS features in browser code
rather than in TLS libraries. For example, to overcome weaknesses in
existing revocation protocols, Firefox ships with OneCRL [43] and Chrome,
CRLSets [24]. Both of these solutions increase browser security in the
typical end-to-end case. However, these solutions provide no protection in
the presence of a TLS proxy and because the solution is not part of the TLS
protocol itself, TLS libraries do not implement these safe revocation
checks. In a second example, HPKP directives are passed over HTTP rather
than during the TLS handshake. Browsers cannot perform HPKP validation for
proxied connections because they do not have access the destination
certificate and proxies do not perform this validation in practice.

While it is possible for proxies to perform this additional verification,
they are not doing so, and in many cases vendors are struggling to
correctly deploy existing TLS libraries, let alone implement additional
security features. Given this evidence, our community needs to decide what
roles should be carried out by the browser versus TLS implementation. If we
expect browsers to perform this additional verification, proxies need a
mechanism to pass connection details (i.e., server certificate and
cryptographic parameters) to the browser. If we expect proxies to perform
this validation, we need to standardize these validation steps in TLS and
implement them in standard libraries. Unfortunately the current situation,
in which we ignore proxy behavior, results in the worst case scenario where
neither party is performing strict validation.


On Wed, Feb 8, 2017 at 10:26 AM, Gervase Markham <[email protected]> wrote:

> Of interest to those involved in security policy: a paper by a group,
> including Mozillians, about the high levels of prevalence of HTTPS
> interception:
> https://jhalderm.com/pub/papers/interception-ndss17.pdf
>
> Gerv
> _______________________________________________
> dev-security-policy mailing list
> [email protected]
> https://lists.mozilla.org/listinfo/dev-security-policy
>



-- 
konklone.com | @konklone <https://twitter.com/konklone>
_______________________________________________
dev-security-policy mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-security-policy

Reply via email to