[Dailydave] information operations efforts and data carving

Dave Aitel dave.aitel at gmail.com
Thu Aug 9 17:09:57 UTC 2018


Previously Unreleased Work:
https://docs.google.com/presentation/d/1tMlJvnUv_Qbh5mx2RYbyuTHTHr9c9ShIKBzz_JDGn_s/edit?usp=sharing
Paper on the 3M Tweets from Clemson:
https://www.cyxtera.com/blog/data-carving-the-internet-research-agency-tweets

So what you see a lot in some papers is this sort of thing (this one is
from the original Clemson paper):
[image: image.png]I always get flashbacks of that XKCD Correlation vs
Causation comic <https://xkcd.com/552/> when I see this type of graphic.
The point of these is to say "Hey, there is increased activity from the
Internet Research Agency (IREA) before the FSB/GRU sends their data to
Wikileaks. And on one hand: Maybe. The flow of information is fun to track.
But in addition to metadata like count(number_of_tweets) you also have the
CONTENT of the Tweets available! There's an infinite number of real world
events you can attach with lines to a histogram to associate with various
curves if you're ignoring the content. Using NLP techniques to analyze the
tweets at various points helps us make better conjectures about the flow of
information through an adversaries IO systems.

I call this kind of thing "data carving" because at some level you are
spelunking about through a massive dataset, like a radiologist taking
slices of imagery from a human body and trying to build systemic knowledge
by staring at them long enough. It's a logic puzzle attached to a
visualization problem attached to natural language parsing and machine
learning science, and if you're doing it right, linguistic, cultural and
domain specific knowledge as well.

My blogpost comes across very much as a Brainspace ad but Brainspace
happens to be a great tool for data carving and basically nobody in our
community has ever heard of it, the way most penetration testers have never
heard of uudeview <http://www.fpx.de/fp/Software/UUDeview/>. I would have
killed for something like Brainspace back when I did more analysis work.

When you do data carving right, you find things that SURPRISE you. We found
lots of examples of attempts to do things remotely like setting up protests
and even joining AN ONLINE ARMY they called the US Freedom Army. Check out
this response on the survivalist website
<https://twitter.com/MiasmRe/status/1027568331656765443> to it.

But if you read our previous work (linked above) we've also seen bot
accounts that are virtually indistinguishable from real people until you go
try to verify them on LinkedIn. We've also found massive client-side
exploitation networks. And it's fun to read the sites like Off Guardian
<https://off-guardian.org/> and see how they all fit into the picture. So
much work in this propaganda edifice. Even if it mostly is creating
poisonous confections it can still have beauty.

Oh, and here's another dark secret: I'm a pretty active advertiser on
Twitter, for various reasons. Twitter charges advertisers per
"interaction". A lot of these fake accounts click a lot of ads. At some
level, they can bribe Twitter to ignore their manipulation of the
conversation and community. If we want to change the effect of Information
Operations we can't be hunting botnets and manipulators - we have to change
the economic incentives. Engaging the policy community on that should
be...fun.

-dave
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.immunityinc.com/pipermail/dailydave/attachments/20180809/1f7dc308/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 52640 bytes
Desc: not available
URL: <http://lists.immunityinc.com/pipermail/dailydave/attachments/20180809/1f7dc308/attachment-0001.png>


More information about the Dailydave mailing list