UM researchers mine data to uncover terrorist threats

U.S. counterterrorism efforts monitor and sort vast databases of information for clues on potential plots. Now a team of University of Maryland researchers have used data-mining techniques employed by online giants like Google and to aid in the fight against terror.

In the same way corporate America uses algorithms to predict what consumers are most likely to buy or what ads they might click, the researchers analyzed two decades of data on Pakistani terrorist group Lashkar-e-Taiba. They were looking for patterns around attacks, including the 2008 shootings and bombings in Mumbai that some compare to the Sept. 11 attacks in the U.S.

The study found that the most effective means of thwarting such violence include stirring dissension within the terrorist group, promoting government crackdowns on its activity and thwarting recruitment efforts — not arresting or attacking the group's leaders.

The research has implications for current counterterrorism efforts, its authors say, and could give insight into other terror networks and how to best prevent future attacks.

"What it does is to bring our attention to possible avenues of inquiry that we might not have identified," said Marvin G. Weinbaum, a scholar in residence at the Middle East Institute in Washington, said of the algorithm-based research.

The uproar earlier this month in Egypt, Libya and other countries over a film insulting the Prophet Muhammad highlight the continuing presence of anti-American sentiment in the Middle East, potentially fueling extremists. U.S. officials said they consider terrorists responsible for at least one of those attacks, on a U.S. consulate in Benghazi, Libya. Recent U.S. counterterrorism efforts have included a drone strike that killed al-Qaida's No. 2 leader in Yemen on Sept. 10.

V.S. Subrahmanian, lead author of the study and director of the university's Laboratory for Computational Cultural Dynamics in College Park, began collecting data on Lashkar-e-Taiba in 2007. He and his colleagues had previously published papers on other terrorist groups better known in the U.S., such as Hamas and Hezbollah, but delved into Lashkar-e-Taiba because the group had been less studied and was beginning to pose a global threat.

The research was presented at a Sept. 10 symposium on the terrorist group in Washington and was published as a book, "Computational Analysis of Terrorist Groups: Lashkar-e-Taiba."

The group dates to the late 1980s and is focused on gaining Pakistani control of the disputed Kashmir region of India and on spreading its version of Islam. Unlike other Muslims, members of the group give equal importance to violent jihad, or holy war against enemies of Islam, and to dahwah, the preaching of Islam.

It wasn't until November 2008 that the group emerged as a threat to the U.S., and to Israel and other countries, in the coordinated attacks across Mumbai. Two Chicago men were among those arrested and charged in connection with the strike; one was acquitted, while the other was convicted. Six Americans and about two dozen other foreign nationals were among the estimated 163 people who died in the attacks, many of which targeted tourism centers.

"When [the Mumbai attacks] occurred, we knew our instincts to study this group had been right," Subrahmanian said. "We also knew it was important to do the job thoroughly."

The information they studied included some 770 variables, including arrests or raids on the group, its involvement in Pakistani elections, its interactions with the Pakistani military, and the social services it may or may not be offering to the community.

Subrahmanian's models translated the updates into numerical data, parsing them for patterns that appear, both around times of violence and times of idleness.

Among the rules it found, for example, was that if Lashkar-e-Taiba operatives were arrested and placed on trial, there was a high likelihood of deadly clashes between other group members and local security forces. Lashkar-e-Taiba members, unlike some Islamic extremists, don't believe in suicide missions but consider it an honor to die in the service of their cause, said Aaron Mannes, co-author of the research with Subrahmanian.

The model allows analysts to take hundreds of rules and patterns that would otherwise be "a little hairy, intellectually" and boil them down into counterterrorism strategies and tactics, Mannes said.

"Some of them, you're kind of like, 'I could have guessed that,'" said Mannes, a Pikesville native. "Then there's cases where you know, "OK, that's new; that's interesting; that's telling me something about what leads the group to make its decisions."

Data mining has increasingly played a role in U.S. intelligence-gathering and counterterrorism efforts, said John Parachini, director of the RAND Corp.'s Intelligence Policy Center. In the Cold War era, the focus was on monitoring changes in missile fields or large armies, but since the Sept. 11 attacks, in particular, there has been more monitoring of individuals' behavior.

While applying the technology to a group like Lashkar-e-Taiba could help shed new insights on its behavior, some said a human touch is still needed to make reliable predictions on future actions. Counterterrorism efforts also focus on police-style methods of interviewing, collecting clues and following promising leads to learn of planned attacks, experts said.

Using the past to predict future behavior can be particularly tricky when it comes to terrorist groups, which often look for new tactics to surprise their enemies, said Daniel Markey, senior fellow for India, Pakistan and South Asia at the Council on Foreign Relations.

"It's an imprecise science," Parachini said of behavior prediction. More open-source data like the kinds Subrahmanian used is available than ever, but "the question is, can you exploit it in ways that give you insight on people's behavior. That's the hard part."

Still, many recognized the research is worth exploring to gain as many counterterrorism leads as possible.

"These are complex problems," said Stephen Tankel, an assistant professor at American University who focuses on jihadist violence in South Asia. "There is merit in looking at them from different angles."

Subrahmanian and Mannes are not working directly with any counterterrorism agencies, but said they hope their research helps inform future strategies and can be applied to other terrorist groups.

Lashkar-e-Taiba, though, remains a particular danger, given the tension it can cause between Pakistan and India, two nations with nuclear capabilities. The group is thought to have the support of Pakistan's intelligence agencies and military. It also is working with the Taliban in Afghanistan, the researchers said.

"They are still active," Mannes said. "They have had a lower profile since 2008, but they are by no means out of business."