PORTLAND, Ore. Proxy servers can often shield terror groups from detection on the Internet. Now, increased use of social networks may provide a way to track them down.
Though still in its formative stages, algorithms could identify communities within social networks to target terrorist cells as easily as customers are tracked by advertising campaigns.
|By graphing the relationships (black lines) among people (red dots) using sophisticated algorithms, graph theory could trace terrorists back to their lair based on their online interactions.|
"I'm a researcher of graph theory, and that's basically a theory of dots and lines," said professor Anthony Bonato at Ryerson University (Toronto). "In online social networks, the dots are people and lines are friendship relationships between them.
"We view these networks like Twitter, Facebook and LinkedIn as complex, massive networks with hundreds of millions of nodes and edges, for which we are developing rigorous mathematical models about how these things evolve and function."
Using smaller anonymous databases from LiveJournal and Flickr, Bonato has been able to prove the concept and obtain initial funding to start building a software tool for monitoring social networks.
"We're hoping to develop ways of pinpointing users on the network by just using the network's graph structure," said Bonato. "Say you are looking for an individual or a group from among hundreds of millions of pages. Hopefully we can narrow that down to just a few hundred."
Bonata's model could be used to trace Web posts on social networks back to their source by understanding the structure of how the information propagates across the Internet. The model would then work back from current conditions to prior causes. In the process, the technique could uncover entire networks of users of particular products, or even to detect and track terrorist cells.
Bonato will present his research results in a paper, "A Survey of Properties and Models of Online Social Networks" at the International Conference on Mathematical and Computational Models.