[cryptography] Crypto Vulns

shawn wilson ag4ve.us at gmail.com
Sun Mar 8 12:38:37 EDT 2015


On Mar 7, 2015 9:11 PM, "coderman" <coderman at gmail.com> wrote:
>
> On 3/7/15, Dave Horsfall <dave at horsfall.org> wrote:
> > On Sat, 7 Mar 2015, Kevin wrote:
> >
> >> > No 1 vulnerability of crypto is the user
> >> > 2nd passphrases
> >> > 3rd overconfidence
> >> > 4th trust in the producer
> >> > 5th believing backdoors are No. 1
> >>
> >> I don't agree that the user should be first on that list unless you are
> >> talking about poor implementation.
> >
> > How would you arrange them, then?  I seem to recall that Enigma was
broken
> > largely due to sloppy user practices e.g. weak message key, re-use of
> > keys, repeating same message with a weaker scheme, etc.  Used properly,
> > Enigma would've been unbreakable at the time.
>
>
> 1. failed software and security engineering. [#'s 1, 2, 4 above all
> reduce to this error.]

I strongly agree with this. For example, people are told to use a password
managers for each site and most people end up with the same password across
hundreds of sites - is that a user failure or one of software? I copy and
paste passwords between pgp files and browsers all the time but I don't
expect my mom to.

> 2. overconfidence [believing backdoors or nation state attacks are
> your weakness is overconfidence in the rest of your threat model]

Well kinda (not necessarily "overconfidence" but the example). How about
this: would the creator of gnupg be getting >$100k per year (I think it's
renewed in 5 years - I suspect it will be) without the NSA things? Point
being, jumping at shadows can cause productive fear (until you die of a
heart attack).

> 3. complacency [if everything else is in place, letting habit slide to
> convenience, then to compromise, will result in sorrow.]
>

Orgs with otherwise pretty damn secure software setups do education next to
teach their people how not to mess up again (this is generally done after a
pentest). However, your average organization isn't going to do this - your
average person can't do this. So I wonder whether we really want to change
habits or make software that learns to conform to the user while staying
secure.

> some would say that truly strong, usable crypto systems with integrity
> for the common public are impossible. i would retort that just because
> we don't know how to build them yet, does not mean they won't exist in
> the future. :P
>

We're starting to build them - take Proton Mail for example. No need to
know pgp, generate a key, verify keys (I don't use it so IDK how they
handle trust). The keys are local to you. It seems there might be
shortcomings with this but I'll give them "pretty good".  And this is just
one example of how you can take a pretty sophisticated software and make it
so that end users can deal with it and aren't likely to leak data and the
like.

OTOH, systems like Active Directory that are hard to setup, not scalable,
allow downgrading of hashes, and have issues like PtH central to the
protocol. Again, not something you can blame a user for - just a badly
designed system. We can do better - should expect better.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.randombit.net/pipermail/cryptography/attachments/20150308/4e4b5f28/attachment.html>


More information about the cryptography mailing list