The bug... More info.

[Peter Lebbing]

On 14/04/14 21:27, Robert J. Hansen wrote:
> Given the bug was introduced in March of 2012, that would mean the bug would
>  have had to been discovered, an exploit tested, a product weaponized

In /this specific instance/, I believe these three can indeed be the product of,
well, mere hours.

I don't think it's unreasonable to suppose it might very well be :) the NSA is
reading through all patches that go into the OpenSSL stable tree; it's not that
much work for a big organisation and you might catch some low-hanging fruit.
This specific patch was extremely low-hanging! We take an unchecked 16-bit
length from a packet we receive from the internet, and use that to copy a block
of newly-allocated data to the return packet! Holy crap. The OpenSSL developer
that accepted this patch (that was submitted by an outsider) was not having a
good day, and I think he would never write this code. For a programmer that has
been working on a security project for quite a while, I believe it becomes just
a reflex while writing such code: you take a length argument from user-supplied
data, you do sanity checks on it.

Similarly, I think people at the NSA, trained at reading code for possible
exploits, might have actually squirted coffee when they read through this patch.
And it's really easily exploited, so yes, I think you can weaponize in a mere
afternoon.

> a product distributed to end-users, and deployed by end-users against 
> targets

I'm not going to assess these few points though, they are less obvious. However,
I think you don't need much distribution. You can just send the heartbeats
yourself, and read other people's data from the process's memory. Sifting
through that memory for interesting stuff is more complicated, but doesn't need
to be done on the spot.

> Further, the use of "at least" two years is meant to imply it could have
> been substantially longer -- but it could not have been more than two years
> and a month.

That indeed is pure sensationalism.

The news report might have been completely made up. However, I think it might
still be true, given how big a target OpenSSL is, and how easy to spot the bug
was. Too bad none of the good guys actually spotted it. Such is life,
unfortunately. People make mistakes. Sometimes pretty big, dumb mistakes. Even
the smartest people.

Peter.

-- 
I use the GNU Privacy Guard (GnuPG) in combination with Enigmail.
You can send me encrypted mail if you want some privacy.
My key is available at <http://digitalbrains.com/2012/openpgp-key-peter>