Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×
Censorship Communications Crime Facebook Social Networks The Internet Twitter Politics

Spotting And Culling Terrorist Groups On Social Media: Pipe Dream, or Possibility? (nytimes.com) 69

An anonymous reader writes: Can Twitter Spot Terrorists and Put Them In Jail? Hany Farid, the chairman of the Computer Science department at Dartmouth University, thinks so. He told the New York Times that there's "no fundamental technology or engineering limitation" to spotting terrorists on the Intertubes. In other words, he's figured out how to tell the difference between bragging terrorists and kids who are just joking about being "da bomb." Can artificial intelligence make these distinctions? Or will it generate a ton of false positives? Or is Prof. Farid just trolling for more grant money to make Dartmouth the premier department for spying on social media?
This discussion has been archived. No new comments can be posted.

Spotting And Culling Terrorist Groups On Social Media: Pipe Dream, or Possibility?

Comments Filter:
  • If Google and Facebook can target me with ads for male enhancement, breast implants and Elmo . . .
    • Caitlyn is that you?
      • by tiberus ( 258517 )
        Why do you ask? Looking for a date?
        • by KGIII ( 973947 )

          Kinda? She's kinda hot. I already have a girlfriend. It's not gay if you have sex with a transgendered woman. I have this on good authority.

    • by laird ( 2705 ) <lairdp@gma i l .com> on Tuesday December 08, 2015 @11:47AM (#51081257) Journal

      Ads can afford a low 'hit' rate, because the cost of being wrong is very low. A good ad gets a 3% response rate, meaning 97% of the time it's a "false positive", but it's still profitable because ads are dirt cheap (0.5 cents would be high) so if you make a few dollars on the 'hits' you can easily cover the misses. In counter-terrorism, each false positive requires detectives to work the lead, making them extremely expensive to pursue. That's why every data mining approach to counter-terrorism has failed so far - the cost of pursuing the false positives gives data mining leads negative value, because they pull resources away from more productive approaches.

      But the government's non-technical management loves the idea, and keeps allocating money to it, and unscrupulous researchers will keep taking the money.

  • by Anonymous Coward on Tuesday December 08, 2015 @10:04AM (#51080281)

    Or is Prof. Farid just trolling for more grant money

    If you want an infinite pool of grant money in electrical engineering in the UK, you go for something with clear defence applications. I expect similar applies for IT.

    What I find embarrassing in all this is that there is really nothing stopping thousands of people being shot every day by lone wolves except that people are generally not that shitty. And when there is propaganda to drive people to do horrific things that they would not normally do, it doesn't come from the DEEP DARK OMG WEB (you have to really want it in the first place to look hard for it there, by its very nature), but from regular media pounding the TV/radio/web sites with news about previous attacks and the threat of the enemy, marginalising and factionalising and dividing and conquering (while arms are sold to both sides, and politicians take great advantage).

    Here's the thing: terrorism is not a big threat in Western nations. In fact, world violence is at a relative low. What is at the highest of highs, however, is the ability to quickly set the narrative for news, getting people to panic about all the appropriate things, then turning their attention to some new event to stop them reflecting too much.

    • by fermion ( 181285 )
      It seems to me that when so white kid goes out and kills a bunch of kids or shoots up a movie theater, we look at their online activity and see that they were clearly deranged and should have under surveillance. That and he was from the Carolinas which are a hot bed of radicalizing terrorists. The problem is that separating those terrorists from the vast majority that are venting is a non trivial problem. Sure, we could require target every white male from North and South Carolina but that would not be a
  • At least at a certain level, with Anonymous taking out thousands of pro-Islamic State Twitter accounts with Operation Tango Down [battleswarmblog.com]. Now that's just one service, and nothing prevents them from signing on again. But you can slow momentum and make it harder for supporters of terrorism to broadcast their views to supporters without reprisals, and also limit or prevent coordinated action.

    Best of all, it's possible to do it merely for Terms of Service violation, without government action.

    Of course, to actually defeat terrorists, you have to kill them faster than terror organizations can create new terrorists, and to dry up their financial support (of which the Islamic State has plenty in "moderate" Sunni states...)

    • by Anonymous Coward on Tuesday December 08, 2015 @10:37AM (#51080533)

      Isn't that the game?

      If this man could identify terrorist speech (i.e. speech from terrorists), then you'd use his software and arrest the terrorists and there would be no Paris and no San Bernadino. He has a computer, he has an Internet connection, there is nothing stopping him writing his software.... yet he hasn't.

      Instead he (and you) define terrorists speech as different, e.g. "supporting or condoning groups you deem as terrorist", or speech from people you believe are supporter of terrorism.

      But that's basically saying "I can make a filter to select X" and "I define X as what my filter chooses". So he doesn't plan to filter terrorist speech, he plans on selected a group of texts in a filter and DEFINING THAT as the terrorists speech he wants to stop.

      Language is a funny thing, and language processing (e.g. for translation) is go awful, so of course he can't deliver on his promises, but he can define his promise to be what he delivers!

    • by Rei ( 128717 ) on Tuesday December 08, 2015 @10:42AM (#51080579) Homepage

      Unfortunately "violations for terms of service" is just as vulnerable to false positives. The other day I went to check out a video on Youtube and found out that it had been pulled for violation of the terms of service. Someone clearly glanced at it and saw screenshots all too familiar: a row of Daesh (ISIS) thugs in all black, their prisoners in orange jump suits before them forced to kneel, guns pointed to their head for an execution. Youtube has taken up a policy of banning such videos.

      Except, that's not what the video was. The video was not from Daesh but Levant Front, an anti-Daesh coalition. The people in black were Levant Front members and the people in the orange jumpsuits Daesh solders that they had captured. And rather than execute them, they all holstered their guns, took off their face masks, and walked away, while an imam showed up and gave a sermon to the Daesh prisoners about forgiveness and being fair and just in life. In short, it was precisely the opposite of the Daesh videos that Youtube is supposed to ban.

      False positives are a real thing.

      • Human communications all take place within a context. Without context, you cannot tell whether a person is quoting someone else, mocking them,a greeing with them or anything else at all.
    • The much vaunted #OpParis has not been a success as advertised [warontherocks.com]. The level of false positives has been massive, with many of the so-called IS accounts and websites being pro-Kurdish, pro-Palestinian, or even simply targeted because they happen to be in Arabic. Think about it, does Anonymous really have the depth of knowledge on terror networks, or even enough competent Arabic speakers, to vet thousands of Twitter accounts and websites? Do they know enough about Arab cultures to understand the nuances of whe
    • " to actually defeat terrorists, you have to kill them faster than terror organizations can create new terrorists "

      That's never worked, didn't in Iraq, didn't in Afghanistan, won't here. Helping democracy grow is a long term process, especially after a population 's been bombed, profiteered, and overall, alienated.

      " and to dry up their financial support "

      First I think you need to look at why is it so easy for them to get so much financial support.

      • by laird ( 2705 )

        Terrorism isn't a military contest, it is a political contest, taken by people who are committed to a cause that is clearly losing (islamic fanatics, christian fanatics, white supremacists). If you kill terrorists, you kill many non-terrorists as well, and that just makes more people hate you, aiding their recruiting. Layer on top idiot politicians (Trump) that spew racist broadsides that promote terrorist recruiting even more, and the situation never gets better.

        The way to defeat terrorists is to remove th

  • by Nikkos ( 544004 ) on Tuesday December 08, 2015 @10:12AM (#51080339)

    The article below this one is full of 'Trump is an idiot' (and he is), but here in the next article we talk about using AI to cull posts.

    'Closing up the internet in some way' would be akin to spotting and censoring a group of people's comments, yes? Effectively limiting their internet use, yes?

    Potatoes, Potatos.

    • by Tomahawk ( 1343 )

      Interesting. However, Donald wants to close it up for everybody, whereas this is talking about identifying (but not actually blocking, yet that _could_ be a next step) certain types of speech behaviour. This is more about identification than obstruction.

    • by tiberus ( 258517 ) on Tuesday December 08, 2015 @10:23AM (#51080429)

      Uh, not just no.

      There's an entire universe between the excrement that Trump spews and using AI, heuristics, machine learning, etc. to identify terrorists or more likely terrorist like behavior. If nothing else this would give humans searching for terrorists a place to start. It would be a constant arms race but, it is certainly worth the try.

  • One one hand that certainly slow down *a bit* recruitment, amount unknown, but that also mean they go underground are are much more difficult to spy on. Much better they stay up, FBI / GIGN /Whoever spy them on, and can catch recruitment attempt or anything suspect.

  • How about working on something useful instead of pouring all this time and money into a solution for something that is more rare than getting hit by lightning while riding a dolphin?
    • In the light of the San Bernardino and Paris attacks - as well as the random lone wolf attacks in Israel, let alone the ongoing civil wars in Syria, Yemen, Sudan, Libya and Afghanistan, it seems likely that more people are being killed by Islamists than by lightning, let alone whilst riding a dolphin. Which is a sad thought.

      However you are correct that terrorism is still a rare phenomenon in the West, and there are better things to spend money on in terms of return
      • Yeah, various perspectives are informative

        How many civilians have been turned into "collateral damage" by US and allied forces in the middle east over past decades.

  • Answers (Score:4, Insightful)

    by ZorroXXX ( 610877 ) <hlovdalNO@SPAMgmail.com> on Tuesday December 08, 2015 @11:23AM (#51080963)

    Can Twitter Spot Terrorists and Put Them In Jail?

    No.

    Can artificial intelligence make these distinctions?

    No.

    Or will it generate a ton of false positives?

    Yes.

    Or is Prof. Farid just trolling for more grant money to make Dartmouth the premier department for spying on social media?

    Yes.

    As always, follow the money (or alternatively, he is incredible stupid and actually belives in it).

    There is always an easy solution to every human problem -- neat, plausible, and wrong. -- H. L. Mencken,

  • by hey! ( 33014 ) on Tuesday December 08, 2015 @11:29AM (#51081053) Homepage Journal

    ... if you don't mind mis-identifying non-terrorists as terrorists.

    It should be so obvious that it goes without saying, but the people who cobbled together things like the anti-terrorist watch lists after 9/11 didn't seem to grasp this: the wider you catch your net, the smaller proportion of what you haul out of the ocean is comprised of fish.

  • Farid

    Match detected. Roll the SWAT team.

  • by DriveDog ( 822962 ) on Tuesday December 08, 2015 @05:25PM (#51084217)
    Didn't BASIC come from Dartmouth? Haven't they caused enough damage already?
  • They create sock puppets and use them to drown out dissent. They brought back the opium market in Afghanistan after the Taliban had destroyed it. They destabilize foreign countries (see John Perkins' "Confessions of an Economic Hit Man"). Paying taxes is adding to evil. I haven't yet decided to fight the IRS though. One approach would be to make checks to that collections agency "payable to the US Treasury", not to that illegitimate organization. And I've been studying other methods of escaping from t
  • The problem with this is the "base rate." That's what decides your bias as to whether you will have more false positives or false negatives, assuming equal probability of both.

    If 1 person in 1,000,000 is a terrorist, and you have a 99.9% accuracy rate (for both false positive and negative), then that means roughly 1 innocent person in 1,000 will be flagged as a terrorist. That's 1,000 people per million...or, in the United States with its current population of 321 million people (as of July 2015, accordin

If all else fails, lower your standards.

Working...