In the wake of the tragedy of Sept. 11, 2001, it became clear to experts worldwide that international terrorist organizations - Al-Qaeda in particular - rely heavily on cybercommunications for their planning. This prompted a group of Rensselaer researchers in the departments of computer science and decision sciences and engineering systems to start developing techniques for modeling the evolution of social groups in Web chat rooms, newsgroups, and bulletin boards, with the specific goal of detecting potentially harmful groups.

The research team - William Wallace, professor of decision sciences and engineering systems, and Mark Goldberg, Malik Magdon-Ismail, Mukkai Krishnamoorthy, and Bulent Yener, all faculty in the computer science department - is using an approach built upon statistical learning algorithms and graph theory to analyze online communications. By monitoring the frequency and persistence of exchanges between specific Internet users, the Rensselaer teamÕs algorithms will be able to pinpoint individuals and groups with non-regular communication patterns. Their algorithms will be able, for example, to identify users that employ multiple online identities in different communication forums, or groups whose members attempt to hide their communications within regular cyber-activity. Intelligence experts can then investigate further and interpret the messages of suspicious users.

The work is based on social network models and properties of Internet usage, such as the theory that "normal" communication between individuals on the Internet is random in both subject and frequency. However, groups "hiding" online converse in a more planned and purposeful manner. Their communications also are more persistent because they must constantly update each other with necessary information, and their communications recur in small, dense patterns.

"The ability to model how cybergroups evolve may be useful on many other levels-not just on the anti-terrorist front," Goldberg says. "It also will help system administrators to budget the correct amount of money for their online forums, simulate the spread of e-mail viruses, and develop appropriate measures to combat them, and even define membership policies to promote better online behavior."

The Rensselaer team has received three National Science Foundation grants totaling $647,000 to support the research.

Read the related NSF story here