Skip to main content

Deterrence via Intermittent Auditing

Researchers: Anupam Datta, Virgil Gligor

Research Area: Next Generation Secure and Available Networks


Another class of useful emergent properties for large-scale multi-agent systems, which can be modeled by behavioral game theory, is that of agent-misbehavior deterrence via strategies that efficiently spot-check agent correct behavior; i.e., strategies that utilize verification/auditing resources efficiently. The emergent property of spot-check protocols is the honest agent behavior in networks where honesty cannot be enforced by security mechanisms. The modeling of these spot-check protocols via leader-follower Stakelberg games (i.e., games where agents observe the spot-checking strategy before taking an honest/dishonest action), has already yielded promising results in real-life security protocols; e.g., the LAX airport, the Federal Air Marshal service, and TSA, in the US. Our research will model deterrence via spot check auditing in large-scale agent systems where exponentially many strategies make explicit adoption of randomized strategies impractical.