Pentagon wants software to monitor Facebook, Twitter to predict terrorism


DARPA is funding programs to detect recruiting and targeting.

The Pentagon wants computer programs that predict “cyber terrorism events” by detecting how criminal groups and hackers interact on the Internet, contracting databases indicate.

The military research arm wants scientists to build the tools to comb through networking sites -- such as Facebook and Twitter -- to analyze the group dynamics of online communities. The Defense Advanced Research Projects Agency will fund the development of algorithms that make sense of the chatter of over a million Internet users, and track how online groups evolve. The goal is to help strategists identify how communities are recruiting and collaborating, who they are targeting, and the shifting allegiances in these spaces.

DARPA is interested in software that can spot conflicts between groups and weak links that the Pentagon and law enforcement could exploit for “strategic military operations” and espionage prevention. “Little attention has been paid to how the groups compete with each other for members and influence on opinions of other teams and communities,” the solicitation for proposals reads.

The first phase of funding will support teams that can identify ways to measure group dynamics.  The second phase will create a demo program that can create a coherent narrative by drawing on the movements of more than 1,000 groups and a million-plus members generating more than 100,000 online posts each day.

Scientists that make it to the third phase of funding are expected to produce tools that market analysts, polling organizations and defense agencies can tap. The Defense Department plans to start accepting proposals on August 27. The solicitation closes on September 26.

Building the tools that can formulate coherent narratives about any online network is an ambitious task. As a case in point, the various offshoots of the hacker group Anonymous are fluid communities. They succumb to internal politics; they mobilize, dissolve and resurface in new forms.

Analytical tools that study group behavior have been unable to nail down this volatility. Such software has typically been most effective when looking at smaller, static clusters of close-knit individuals, the tender acknowledges. The Wild West that is the social media world could offer a different set of challenges.