All Tech Considered
Mon January 30, 2012
What The FBI Wants In A Social Media Monitoring App
Originally published on Tue January 31, 2012 12:02 am
According to the document, the bureau is looking for a mapping app — or a "geospatial alert and analysis mapping application" — that, among other things, helps it search "publicly available" sources like Facebook and Twitter for national security threats.
Some other items on the FBI's functionality wish list include:
- "... instant notifications of breaking events, incidents, and emerging threats that have been vetted and meet the defined search parameters."
- "Ability to immediately access geospatial maps" that plot "US Domestic terrorist data"; "global terrorist data"; "US Embassy, consulate and military installations around the world"; weather conditions and forecasts; and "video feeds from traffic cameras."
- "Ability to instantly search and monitor key words and strings in all 'publicly available' tweets across the Twitter Site and any other 'publicly available' social networking sites/forums (i .e. Facebook, MySpace, etc.)."
- "Ability to immediately translate into English, tweets and any other open forum publicly available social media captured in a foreign language."
- "Ability to geo-locate the open source social media 'search' by setting a radius by both miles and kilometers (i.e. 5 miles, 10 miles, 50 miles radius) that will allow the user to quickly narrow the search to a specific area/region/location."
- The ability to "geospatially locate bad actors or groups and analyze their movements, vulnerabilities, limitations, and possible adverse actions"
- The ability to "develop pattern-of-life matrices to support law enforcement planning and enforcement operations"
- "... reference documents such as a dictionary of 'tweet' lingo"
Sean Gourley has worked with defense agencies in the past and now heads the intelligence firm Quid. He gives NPR's Audie Cornish one example for how the agency might use the app to monitor breaking news.
"If there's an attack that's just been carried out in north Afghanistan that they weren't aware of, there might be reports of that on the social media channels that they're watching," Gourley says. "[People] might be tweeting, 'I heard a loud bang,' or someone says, 'Maybe there's a bomb around the corner' or there are these kinds of reports. Now, each of these pieces kind of starts to form a little bit of a mosaic and they start to combine this mosaic back together to say, 'We can be pretty sure that something happened here and here's what we think it is.' "
But the FBI also specifies that it wants to use the app to "predict future actions taken by bad actors." According to Gourley, that involves creating profiles of known bad actors based on their social media presence.
"Then what they can do with that is say, 'Here's the kind of profile of somebody that we'd be potentially interested in, even if we don't know that they're already a bad actor,' " Gourley says. "I should say that this stuff is all very experimental at the moment, and by no means does it exist today."
In an email statement to NPR, the FBI said, "The application will not focus on specific persons or protected groups, but on words that relate to 'events' and 'crisis,' and activities constituting violations of federal criminal law or threats to national security."
Then there's the question of why the FBI — an agency that's usually reluctant when it comes to sharing its social media tactics — would publicly lay out its intelligence-gathering plans and then ask for civilian help in executing them. Gourley theorizes that it has something to do with the fact that these days it's easier to find qualified mathematics Ph.D.'s in Silicon Valley than it is to find them in the federal government.
"What that means is the top solutions to these kinds of problems don't actually lie within the government anymore; they actually start to lie in the startup companies," Gourley says. "So increasingly the government starts to turn to these groups to say, 'Can you help us solve these types of problems?' "
MELISSA BLOCK, HOST:
The FBI raised eyebrows in the tech world recently. It put out a call for advice on how to harvest information from Facebook, Twitter and other social networking sites. The bureau wants to troll these sites for signs of a national security threat.
AUDIE CORNISH, HOST:
That means scanning public posts, past and present, to learn about breaking news or to find patterns by so-called bad actors.
Here to talk more about this kind of intelligence work is Sean Gourley. He's worked with defense agencies in the past and now heads the high-tech intelligence firm called Quid. Hello there, Sean.
SEAN GOURLEY: Hi. How are you doing?
CORNISH: Good. So to start, on a technological level, what exactly does the FBI and the Justice Department want to do here? And how would such a program work?
GOURLEY: Well, what they're really trying to do is to try and gather what's called open-source intelligence. Now, open-source intelligence has been around since the time of the Second World War when you would look at the price of oranges in Germany, and try and understand if the fluctuation in the price of oranges would mean that a bridge had been bombed.
And now, it's kind of taken that into kind of the world that we live in today and saying, well, you know, does a stream of tweets mentioning, you know, the word bomber explosion mean that there's been an activity that's happened that we should be aware of?
Now, what they're trying to do is take this information and turn it into intelligence that they can use to guide emergence of threats or bombings or terrorist activities or, indeed, any other kind of thing that the FBI would want to do with it.
CORNISH: Now, let's get some examples here. What kind of security profile could they create on a person or a group based on the information that the average social network user already makes available?
GOURLEY: So, I think, you know, the first thing that they'll probably want to do with this kind of stuff is to see and get intelligence oversight to what's going on. So, if there's an attack that's sort of just being carried out in northern Afghanistan that they weren't aware of, there might be reports of that on the social media channels that they're watching. And they start to recombine this mosaic back together to say, yeah, we can be pretty sure that something happened here and here's what we think it is.
CORNISH: Now, in their call for information, they definitely mentioned that they'd want to be able to search terms like lockdown or bomb or white powder. And that seems pretty easy. You could do that on Twitter right now. But there were some other things that seemed a little bit more complicated, like one quote here, "to predict future events taken by bad actors." Now, what does that mean?
GOURLEY: So, the idea there is what we (unintelligible) again to statistical modeling. Then we can look at the statistical profiles of their social media postings or their geolocations from foursquare check-ins or something along those lines. And what they can do with that is then say here's a kind of profile of somebody that we'd potentially be interested in, even if we don't know that they're already a bad actor.
And so, you start to get into the kind of predictive modeling space. Now, I should, you know, kind of say that this stuff is all very experimental at the moment and by no means does it exist today.
CORNISH: The FBI says that this - said specifically in a statement that this would not focus on specific people or protected groups. But, you know, I have to wonder in an era where there's - where the privacy policies for Google or Facebook seemed like they're changing all the time, how do you really protect yourself as a citizen?
GOURLEY: Look, the world that we live in now is really one where we've sort of become accustomed to sharing little snippets of our lives. Now, we're sort of OK with this as long as it's just serving us ads and we can buy a pair of trainers. But I think we'll start to find in the coming years that the information trails that we leave behind will also start to define us.
CORNISH: So, where you spend your money, how you spend your time could sort of paint a picture of what you like to do, what kind of person you would be, how likely you are to attack someone else? I'm not sure which are saying.
GOURLEY: I mean, so at the moment, that information is being collected and is being used to serve the ads that you see on Facebook. You know, Facebook will serve you up an ad based on the history of things that you have done. What it really hasn't been applied to in any great sort of systematic way is looking to see if you're a terrorist. Now, it's still a moot point as to whether that can be done. Because, indeed, serving up a book recommendation is a lot more simple than serving up a kind of a warrant for an arrest.
CORNISH: Right. And even some of those companies don't do that very well when you think about it.
(SOUNDBITE OF LAUGHTER)
GOURLEY: Right. Now, we've all seen the kind of people like you might like this book and we've seen that go wrong. And that's OK when it goes wrong with a book recommendation. It may not be OK if it goes wrong and you end up in jail.
CORNISH: Sean Gourley, thanks so much for talking with us.
GOURLEY: Thank you.
CORNISH: Sean Gourley, he's head of the high-tech intelligence firm Quid.
(SOUNDBITE OF MUSIC)
BLOCK: This is NPR News. Transcript provided by NPR, Copyright NPR.