Over the last few years, there have been a number of projects aiming to mine social media data for insights. Whilst it’s tempting to think of such projects as looking into the future, in reality they are merely trying to spot what is happening in the present faster than current methods.
Nowhere is this more pressing than in the hunt for terrorist activity, so it’s interesting that a recent paper, published in Science, attempts to examine whether it’s possible to detect a potential terrorist attack by monitoring the social web.
Mining for terrorists
The researchers developed an algorithm to mine social media to look for signs of terrorist activity so that we might be able to get a head start on preventing their terrible acts. The authors believe that when terrorists communicate online, they use some characteristics that are common enough to allow them to train an algorithm to look out for them in future.
For this particular exploration, the researchers examined the Russian social network VKontakte, which claims to have around 360 million users around the world. They managed to manually identify 196 groups on the network that were supportive of ISIS, with those groups containing over 108,000 members.
When they began to monitor these groups, it was evident that they were experiencing growth, with the biggest tending to form from a merger of smaller groups. Over a six month period, the researchers were able to collect a significant amount of data on the behavior of the groups, which they used to train their algorithm to predict potential terrorist behavior in more open groups.
Spotting groups
The study found that it is more effective to look out for groups of terrorists online than it is to look for individuals. What’s more, it’s better to go after smaller groups who are at a weaker stage before they combine into larger groups.
Interestingly, the algorithm suggests that the rate by which groups come together to form larger groups is a good indicator of a likely attack or event.
It remains to be seen what the researchers will do with the algorithm, although they do believe that it could be used by security agencies in the fight against terrorism.
“It would be possible to create automated machinery that then looks across the different online media sites, and detects the aggregates, detects their dynamics, checks it out, looks for the escalation, and therefore heightens alerts when there’s an escalation of aggregate creation,” they say.
Of course, any such action would require the cooperation of the social networks themselves, and the likes of Facebook do have a team that look for dubious activities on the network. The scale of the challenge however was underlined by the news from earlier this year that Twitter had zapped some 125,000 accounts with connections to ISIS.
It’s possible therefore, that this kind of tool could make such a hunt slightly less labor intensive. Time will tell just how useful it turns out to be.
