[aklug] Re: Tor + Firefox

From: Arthur Corliss <acorliss@nevaeh-linux.org>
Date: Tue Feb 14 2012 - 15:44:55 AKST

On Tue, 14 Feb 2012, Christopher Howard wrote:

> Wikipedia's cert is not self-signed, but is signed by DigiCert Inc. At
> present, I do not accept any self-signed certificates.

Isn't that just a tad silly? For any site that you're merely a
reader on (no logons, just browsing) how does rejecting self-signed
certs make you any safer? CAs offer one marginally useful service:
authentication of parties. If all you care about is the ability to
peruse a site without a third party being able to monitor specific
requests (never mind the obvious fact that CA-signed or not, they
always know who you're connecting to, barring proxy services) then
self-signed is perfectly fine. You still have transport encryption.

> I trust the global CA because it is practical to do so and still
> maintain reasonable security expectations. Finding a certificate that
> has been assigned by a global CA gives me a reasonable assurance that
> the host I am communicating with is one operated by the person or
> group that owns the domain I am attempting to contact. This is, of
> course, because global CAs when issuing certificates will, at the
> least, do a domain registration validation to ensure that the customer
> has control over the domain he purports to own. Furthermore, global
> CAs interest in preserving their reputations as CAs.

Not to pile on, but given the recent CAs that have been hacked, how wise
is it to trust a global CA? CAs offer a great service in theory, but
it's the implementation that'll kill you. And I wouldn't be surprised
if the "global" CAs are lax in many security aspects.

Reminds me of a conversation I had on a securityfocus mailing list with a
chucklehead who kept his root CA in a non-networked vm on his desktop. He
assumed that because the vm was non-networked it was secure.... even
though it was executed and stored on a mobile, networked laptop. It's
that kind of idiocy that runs rampant throughout all the industries.

And keep in mind that the CAs aren't in the business of security, they're
in the business of making money, trying to sell a security service. If
they can cut corners to increase their margins with only moderate risk
and liability, you can be guaranteed they will.

> A self-signed certificate is not inherently insecure as far as the SSL
> protocol itself is concerned, it just says very little about the
> origin. One may accept self-signed certificates, but ideally only
> after one has applied some reasonable method to ensure origin. For
> example, if gentoo.org provided a self-signed certificate, I could
> call the Gentoo foundation and ask them what the fingerprint of their
> certificate was, and then make an exception for that specific
> certificate. Though, I would be very annoyed with the Gentoo
> foundation for making me go to all the trouble.

Using your example, that's just silly. Perhaps not if you were logging
into their systems and authoring content, but not just to read their
content. I would also assert that for many organizations using certs
for non-commercial purposes it's also unreasonable to demand that they
incur the expense of dealing with the global jack wagons.

Ideally, gentoo would do what I do: run your own CA. That way they
can generate whatever certs they need for any purpose and you only
need to import one root CA cert to validate all the certs in use.
It's a reasonable compromise.

> As to default browser certificates, of course there is a certain
> amount of risk there that could be lessened by education, though I
> imagine that for practical reasons the typical user is just as safe
> accepting the browser default pack as he is trying to maintain his
> own. In the case of Firefox, all the issues regarding which
> certificates should be accepted or dropped are discussed publicly on
> the dev-security-policy mailing list
> <https://lists.mozilla.org/listinfo/dev-security-policy>. So you could
> listen in and then delete or add individual certs from the pack if you
> happened to disagree with a final decision.
>
> So that, I think, is workable enough philosophy.

It works, but even they've demonstrated that it leaves you at risk in many
circumstances. Just keep your eyes open, and understand that the mozilla
committee isn't validating the IT solutions and security of any of the
CAs out there. And when it fails, as it has, they're going to be
*re*active, not proactive, and remove root CA certs only after their
assumptions of their reliability has been proven wrong. Not before.

         --Arthur Corliss
           Live Free or Die
---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.
Received on Tue Feb 14 15:45:07 2012

This archive was generated by hypermail 2.1.8 : Tue Feb 14 2012 - 15:45:07 AKST