Machine Learning and Game Theory for Counter-Terrorism

Over View

This project tackles the issue of efficient destabilization of terrorist networks. As a main approach to counter-terrorism, High-Value Targeting (HVT) or the targeted elimination of key leaders in terrorist organizations, often has low efficiency in reducing the long-term threat of the targeted organization. This is generally attributed to the fact that the eliminated members are often quickly replaced. In this work, we leveraged the fact that elements of a terrorist organization are rational agents, and their desire to remain in the network can be weakened through strategic targeting. Accordingly, we adopted the framework of network formation games to develop a behavioral model of the agents in a terrorist network. We further proposed a framework for inference of individual agents’ payoff functions and ties from the unstructured text of intelligence reports, and developed an iterative framework based on Reinforcement Learning for strategic decision-support in HVT operations. We also demonstrated the performance of our proposed HVT framework on the simulated targeting of the Al-Qaeda network in 2001.

Current Team Members: