By Mike Strasser, U.S. Military Academy Public AffairsDecember 12, 2012
WEST POINT, N.Y. (Dec. 12, 2012) -- The U.S. Military Academy Network Science Center hosted a briefing Dec. 5, about an algorithm which could help shape future operations against terrorist organizations.
The team comprised of Maj. Devon Callahan and Maj. Paulo Shakarian, instructors in the Department of Electrical Engineering and Computer Science; Lt. Col. Anthony Johnson, Department of Mathematical Sciences professor; and Class of 2013 Cadet Jeffrey Nielsen, is developing this algorithm with the premise that attacking the head of a terror cell is not effective if its leadership can be replenished quickly.
Callahan, a Signal Corps officer, said U.S. forces learned this directly in 2006 after eliminating former Al-Qaida leader, Abu Musab al-Zarqawi, who was replaced two weeks later.
"The problem is that cutting off the head of an angry beast is not necessarily the right way to target these terrorist networks," Callahan said. "We feel the more traditional methods of targeting leadership are successful if the network is more centralized, or star-like."
Actual terrorist networks aren't shaped like a star network and are more complex, Callahan said.
"Our intuition is rather than attacking the network as it is, we should drive the network to a more star-like shape prior to attacking the leadership," Callahan said.
Identifying and removing an organization's "middle management" before targeting leadership would first centralize power and make the whole organization vulnerable.
The research team tested the theory and its mathematical model against the data sets of several real-world terror cells and other real-world social networks.
They also tested their approach to targeting terrorist organizations against other network analysis techniques and Callahan said their method clearly outperforms against the others in terms of preparing the network and making it more vulnerable to a leadership strike.
"The key idea here, which we believe is different from the traditional methods used, is that we've decided to shape the network based on a characteristic of the entire network rather than a characteristic of a single node," Callahan said.
To make this happen -- creating a more centralized network -- the research team used methods from mathematical sociology to derive a way to determine how centralized a network is. The researchers introduced their fragile measurement and used a greedy heuristic to speed an otherwise slow process.
"The idea behind it is to repeatedly select those nodes that provide the greatest increase in the network's fragility," Callahan said. "We feel it runs pretty fast and gives us an accurate solution."
The algorithm itself is called "GREEDY_FRAGILE," and was implemented in Python, a popular programming language. Their work is being funded by the Army Research Office and Callahan will present their team's paper at the 2012 Academy of Science and Engineering Social Informatics Conference in Washington, D.C., Dec. 14.
"This work is a new line of research for the West Point Network Science Center. Though it is only our first step into this new study, I believe we've obtained some rather exciting results," said Shakarian, the primary project investigator.
Shakarian hopes that this new research not only gains traction in the Army, but also opens up new opportunities for cadet education.
"Projects like these are important because it gives cadets like Jeff Nielsen an experience that they just can't get in the classroom," he said. "It allows them to explore problems from different perspectives (which is) an important trait for a future Army officer."