Re: Fwd: [ PRIVACY Forum ] Surveillance via bogus SSL certificates

--- On Sat, 4/3/10, Dave Howe <DaveHowe@gmx.co.uk> wrote:

> From: Dave Howe <DaveHowe@gmx.co.uk>
> Subject: Re: Fwd: [ PRIVACY Forum ]  Surveillance via bogus SSL      certificates
> To: "Email List - Cypherpunks" <cypherpunks@al-qaeda.net>
> Date: Saturday, April 3, 2010, 4:19 PM
> Rayservers wrote:
> > I have proposed that we strip out ALL outside
> certificate authorities from an
> > open source browser, and distribute such... and to
> practice what I preach, I
> > just went into FF and nuked the bunch - and whee, I
> can connect, verify the cert
> > and login :). The USER - a la monkey sphere - has to
> decide if she trusts the
> > Certificate Authority - who the hell are they anyway?
> And to answer my own
> > rhetorical question - those that issue the highest
> TRUST certificates to
> > licensed scammers a.k.a. the banks. I do not trust a
> single one of the
> > recommendations of official CAs. If I am forced, like
> one has to in this world -
> > to visit a bank website, I can figure out how much I
> distrust them all by
> > myself. All I want to know is "am I visiting the same
> site again"... and a "self
> > signed" cert is all I need, "ssh style". And yes, I
> love the monkeysphere
> > approach which would add meaningful levels of trust to
> that choice. And no -
> > there is no difference in my trust level if the cert
> says "self signed" or
> > "fairysign super duper" perhaps the former is better!
> - at least fairysign
> > cannot go off and bless the MITM - especially of any
> sites I run!
> 
> Its a nice theory, but doesn't cover first-visit scenarios,
> nor the
> yearly rekey grind of giving CAs (large amounts of) money
> for the
> results of a fairly easy math problem.


The first visit scenario is definitely an issue. that brings it to the other question - why cannot CA's issue certificates to sites say like 10 years or 20 years and get the corresponding money for that. Most certificates issued by CA's usually have 2-3 years validity. Incase of a significant mathematical breakthrough the CA should provide an alternate secure certifying mechanism if the breakthrough occurred within the service period (10/20 years). The question is why do popular https sites not go for certificates that expire in 10/20 years if it helps security?



Another question, this one is specific to gmail - which the entire session is on https.

when i click a pdf in my gmail to be opened with google docs, the certificate is signed by google(used a third part browser plugin to check this). that is fine, however my browser never alerts me as a potential untrusted certificate and if want to add it as an exception. does that mean google is an intermediate CA or what does that mean?


Thank you,
Sarad AV



> 
> What I would prefer is some parallel system where person
> 'x', who I
> trust, may or may not have visited site 'y', and may or may
> not have
> signed the then certificate, the signature for which (with
> its date of
> providence) is then stored *on the site* for me to access
> though a
> well-known url. That way, I can look with suspicion at
> sites which do
> not have such a certificate, investigate myself if they are
> serving the
> certificate I am expecting to see (and how do I do that? I
> have tried in
> the past phoning companies to obtain their website public
> key for
> independent verification; most don't know what one is, a
> few have even
> said they can't disclose that as it is *priviledged
> information*....)
> 
> But, who do I trust for that, who do *you* trust for that,
> and will
> those people be wiling to give up a significant slice of
> time every year
> revisiting websites after their certificates are renewed,
> and facing the
> same hurdles I did (the complete ignorance of most
> companies as to how
> their websites' certificate works and unwillingness to
> supply an
> accurate fingerprint over the phone).