The study of networks has grown substantially in the last decade or so. Mathematicians, sociologists, and scientists from various disciplines have found new ways in which network theory can be applied to an existing situation. An interesting article by Patrick Radden Keefe of the NY Times examines the role that network theory may play in uncovering terrorists. Much of the techniques and terminology he uses has been touched upon by Professor Kleinberg towards the beginning of the semester when we were discussing network theory. To reiterate, in its basic form, network theory represents discrete objects (in this case, people) and the relationships shared between them in the form of nodes and edges.

In wake of the September 11th attacks, network theory has been more extensively applied in an attempt to uncover elusive terror cells. One social network consultant found that all 19 hijackers were connected to one another by only a few links, with the majority of them converging on the leader, Mohamed Atta. There are several government programs utilizing network theory as well. In 2002, the controversial Total Information Awareness program as launched in an effort to search through large volumes of data for hidden connections. Another project called Able Danger by the US Army aimed to create a graph of Al Qaeda by “identifying linkages and patterns in large volumes of data”.

The US government has a treasure trove of information, courtesy of the NSA’s (National Security Agency) communication interception abilities. With somewhere around 650 million communications a day, the question of who to listen to first becomes an issue. One technique used is to begin with a suspect and progress outward in a “spider-web”, or in other words, a breadth-first search. The problem with this approach is that, a typical person is connected by two degrees of separation to thousands of people, and to hundreds of thousands of people by three degrees. Also, the National Counterterrorism Center lists about 325,000 names of suspected terrorists. Hence, even this technique is prone to providing the NSA with an overload of information, as well as many false positives.

Before wiretapping or any sort of comprehensive surveillance is attempted, the NSA looks through metadata, which are records of who is contacting whom via phone or email. This simplification of data allows NSA analysts to search for network hubs, or nodes with exponentially more links than average. Algorithms have been developed to determine the “role structure” in a network, establishing any logistical and hierarchical relationships. The strength of weak ties are also taken into account, in that the most valuable information may be exchanged between nodes whose respective networks have no other common connections. Another approach suggested by Duncan Watts of Columbia University is using affiliation networks. These are networks consisting of people joined together by a common activity. Certain signature behaviors could be searched for in order to track terrorists. One example given is to look for Muslim militants who play paintball. This technique - looking for certain activities or behaviors - is similar to those used to catch credit card fraud. It is not certain of how well this will work in finding terror networks, since there is not as many reliable patterns and behaviors for terror networks as there are in say, credit card fraud.

Sources:

http://www.nytimes.com/2006/03/12/magazine/312wwln_essay.html