Home > Cool Tech > Could this attack prediction algorithm help stop…

Could this attack prediction algorithm help stop ISIS before it strikes next?

The disturbing power of terrorist attacks lies in their fundamental unpredictability; terrible events of human cruelty which could be unleashed on anyone at any time.

But a new Big Data project created by a University of Miami physicist named Dr. Neil Johnson suggests they don’t have to stay unpredictable. Working with a team of researchers, Johnson has created an algorithm designed to scan social media posts by the Islamic State, in a variety of languages, and use them to predict where the group will strike next.

Related: Algorithm can predict potential seizures 20 minutes before they strike

“Lots of people have and are looking at social media,” Johnson tells Digital Trends. “But the focus is always on individuals — who is the bad guy — and maybe stretches to who is connected to who on, for example, Twitter.” Johnson’s algorithm casts its net a bit wider. Focusing on the Russia-based social network Vkontakte, it analyzed the correlation between emergent self-organized groups online and real-world terror attacks — resulting in a “statistical model aimed at identifying behavioral patterns among online supporters of ISIS [for helping] predict the onset of major violent events.”

Johnson discovered that radicalized people rarely stay isolated, but rather join larger online groups which can then be used to predict where the Islamic State may attack. “Sudden escalation in the number of ISIS-supporting ad hoc web groups (“aggregates”) preceded the onset of violence in a way that would not have been detected by looking at social media references to ISIS alone,” the study notes.

While it won’t uncover instances in which groups of just two or three individuals are working together, the algorithm is suited to finding potential terror activity when terrorists become more organized. Once this happens, the groundswell of online activity means the potential attack could be stamped out before it becomes a large-scale, possibly deadly, incident.

“I hope it does open up a bigger collaboration between security officials and academics,” Johnson tells us about his work. “After all, none of what we did involved hacking any accounts; it was all open source information that any of us can get — and yet it yielded new information and a new way of thinking about pro-ISIS online activity.”