This story is over 5 years old.


Mathematical Models Suggest New Ways of Targeting Pro-ISIS Groups Online

To break up large pro-ISIS groups, agencies should start small.
Image: Shutterstock

Non sequitur arguments in favor of responding to domestic terrorism by waging foreign wars at least serve to illuminate the deep problem of how to predict and prevent so-called lone wolf attacks, in which some random person decides to purchase a legal assault rifle and commit mass murder. Gathering meaningful information from would-be perpetrators in the absence of organizational collusion isn't that much better than searching for a needle in a haystack, even when those perpetrators can't stop talking to everyone about what a badass terrorist they think they are.


The internet, however, provides an edge. Researchers from the University of Miami led by computer scientist Stefan Wuchty have developed a mathematical model for identifying self-organized aggregate groups of extremists or "self-radicalized actors" on VKontakte, Europe's most popular social networking service. This model, they suggest, may be sufficient to predict forthcoming pro-ISIS attacks even when no actual dates have been set for those attacks. Their work is described this week in Science.

"Support for an extremist entity such as Islamic State (ISIS) somehow manages to survive globally online despite considerable external pressure and may ultimately inspire acts by individuals having no history of extremism, membership in a terrorist faction, or direct links to leadership," Wuchty and co. write in the paper. "Examining longitudinal records of online activity, we uncovered an ecology evolving on a daily time scale that drives online support, and we provide a mathematical theory that describes it."

As a starting place, the researchers note that individuals just casually namedropping ISIS isn't enough to predict violence. Their approach is rather more clever and involves tracing backwards from relevant hashtags (#isn, #khilafah, #fisyria, etc.) to aggregates (groups within VKontakte akin to Facebook Groups) made up of individual posters in which pro-ISIS posts or posts in favor of jihad in the name of ISIS were found.


"Groups are by definition always public domain," Wuchty told me. "Individual profiles are not, so there are some limitations about what we can get. What we found is that groups with a particular interest are linked to groups with similar interests. We find with our hashtag search a particular group and that group has potentially links to other groups with similar interests. So we're jumping around those links to the other groups and we follow links to other links. We connect groups until we don't find new groups anymore."

Every day, as more aggregates are added to the list of potential pro-ISIS groups, a subject expert is brought in to manually analyze new discoveries and to make the call as to whether or not they should be included in the final pro-ISIS list. "The experts are there in the beginning, so there is this non-automatic part where a person comes in to make a judgement call," Wuchty explained. "But what we found so far is that the number of groups we identify doesn't waver very much."

This is a labor-intensive process, but Wuchty and his group were able to find closures on a daily basis, eventually tallying 196 pro-ISIS aggregates involving 108,086 individual followers. Daily links ranged up to 134,857. Again, this was all on VKontakte, which was selected largely because, unlike Facebook, it does not have a policy of immediately shutting own pro-ISIS groups and because it's a known vehicle for spreading ISIS propaganda among Russian populations.


From the Science paper:

Our theoretical model generates various mathematically rigorous yet operationally relevant predictions. First, anti-ISIS agencies can thwart development of large aggregates that are potentially far more potent by breaking up smaller ones. Adding a simple cost into the model for shutting down an aggregate makes this strategy actually more effective than targeting the largest aggregates. Second, if anti-ISIS agencies are insufficiently active in countermeasures and hence the overall rate at which they fragment pro-ISIS clusters becomes too small—specifically, if the aggregate fragmentation rate—then pro-ISIS support will grow exponentially fast into one super-aggregate. Third, when fragmentation rates drop below a critical value, the system enters a regime in which any piece of pro-ISIS material can spread globally across the pro-ISIS support network through contagion.

Furthermore, the model indicates that "lone wolf" actors are likely to remain lone wolves only temporarily before finding and being absorbed into an aggregate. This is likely to be on the order of weeks.

Wuchty's data indicates that pro-ISIS aggregates naturally adapt and grow, even though each aggregate is an ad hoc group of followers who likely have never met, do not know each other, and do not live in the same city or country. This adaptability can take the form of group name changes, content visibility, and reincarnation, in which a group disappears and then reappears as a new group, albeit one with most of the same members as the old group.

"Our findings suggest that instead of having to analyze the online activities of many millions of individual potential actors worldwide, interested parties can shift their focus to aggregates, of which there will typically be only a few hundred," Wuchty and co. conclude. "Our approach, combining automated data-mining with subject-matter expert analysis and generative model-building drawn from the physical and mathematical sciences, goes beyond existing approaches to mining such online data."