[Cryptography] It's GnuTLS's turn: "Critical new bug in crypto library leaves Linux, apps open to drive-by attacks"

[Theodore Ts'o]

On Tue, Jun 03, 2014 at 04:29:16PM -0400, John Ioannidis wrote:
> It makes sense in another perverse sort of way: so long as users are
> susceptible to social engineering (the world's oldest profession),  there
> will always be far easier ways for criminals to victimize users than to
> break the crypto. So why bother fixing the crypto? It's not the most
> pressing problem.

There's actually a bigger problem, which is users can't tell whether
or not a company has good security or not.  (Sure, they can tell after
a massive security failure, ala Target, but that's not the same
thing.)  So from a economic signalling perspective, which makes more
sense?  (a) investing extra money to improve the company's security
(whether this is by paying for more power and CPU cores to do
encrpytion even for internal RPC's, or to spend lots of software
engineering time doing code audits and redesigning vulnerable
systems), or (b) employing marketing specialists to make it _appear_
that your company has really good security, for those few customers
that will actually pay the switching cost and/or pay more for
additional security?

There's a reason why there is so much snake oil being sold by security
companies in the field, and it also explains the race to the bottom by
most CA providers.

Obviously, a company will want to invest just enough in security to
hopefully prevent the really embarassing security breach, but when you
consider how little the market has punished various CA's after some
truly spectacular failures, the economic choices that they have made
seem perfectly rational from a free market point of view --- just as
the long delay in implementing chip and pin for credit cards in the
United States was a perfectly rational choice from an economic point
of view.

Regards,

						- Ted