Security is a critical concern around the world and the defender has limited security resources that preclude full security coverage of important potential targets at all times. Thus, allocations of limited security resources must be intelligent, taking into account differences in priorities of targets requiring security coverage, the response of adversaries to the security posture based on knowledge gained from surveillance, and potential uncertainty over the types, capabilities, knowledge, and priorities of adversaries faced. By casting these problems as a Stackelberg game and solving these games to yield randomised security policies, we have pioneered the first-generation of operational game-theory based systems in many security settings, including that for Los Angeles International Airport (LAX), United States Federal Air Marshals Service, the United States Coast Guard, and wildlife conservation organisations.
We have designed algorithms for solving large-scale security games taking into consideration the uncertain information about our adversaries, heterogeneous targets, and human behaviour. In the Stackelberg game framework, the defender commits to a mixed (randomised) strategy of patrols, which is known to the attacker. This is a reasonable approximation of the practice because the attacker is expected to conduct surveillance to learn the mixed strategies that the defender carries out, and responds with a devised strategy of an attack on a target. The optimisation objective is to find the optimal mixed strategy for the defender.
The technology has been applied to many settings, including those at the airport, air marshal, coast guard, and wildlife conservation. It has received international reputation. Under the hightened level of terror threats and growing cyber threats in recent years, this technology promises to contribute to a safer Singapore.