Security Games: Key Algorithmic Principles, Deployed Applications and Research Challenges(*)
summary
Overview of concrete claims, methods and results: deployable Stackelberg security-game systems ARMOR, PROTECT, PAWS use mixed-integer programs, column/double-oracle and marginal flow formulations to avoid enumerating 10^41 defender strategies (often reducing to ~500), integrate MDPs for execution uncertainty, and learn boundedly rational adversary models (SUQR/SUQRR) from field and AMT data. Deployed patrols report a 350% increase in defender expected utility in ports and ~60% higher ferry-evader capture rate; a startup (ArmorWay) commercializes the tech.
AI-written summary · about
chapters (23)
- 0:00Intro
- 1:13Global Challenges for Security: Game Theory for Security Resource Optimization
- 1:55Example Model: Stackelberg Security Games
- 2:17Stackelberg Security Games Security Resource Optimization: Nor 100% Security
- 4:54Security Games: Research & Applications Game theory+Optimization Uncertainty+Learning+...
- 7:17Startup: ARMORWAY
- 7:44Outline: Security Games Research (2007-)
- 8:34ARMOR Airport Security: LAX [2007] Basic Stackelberg Security Game Model
- 10:24Basic Security Game Operation [2007] Using ARMOR as an Example
- 11:18Security Game MIP (2007)
- 14:03IRIS: Federal Air Marshals Service [2009] Scale Up Number of Defender Strategies
- 17:02IRIS: Incremental Strategy Generation Column Generation
- 18:18Road, Social Networks [2013] Scale-up: Double Oracle
- 23:26TRUSTS: Frequent adversary interaction games Patrols Against Fare Evaders
- 24:24TRUSTS: Patrols Against Fare Evaders Uncertainty in Defender Action Execution
- 25:33Protecting Forests, Fish, Rivers & Wildlife: Green Security Games
- 26:31Wildlife Protection Murchison Falls National Park, Uganda
- 27:42Green Security Games: Repeated Stackelberg Game
- 30:48Subjective Utility Quantal Response Models of Bounded Rationality
- 32:40PAWS: Protection Assistant for Wildlife Security Trials in Uganda and Malaysia: 20141
- 34:09Evaluating Deployed Security Systems Not Easy: Are Security Games Better at Optimizing Limited Resources
- 35:15Field Evaluation of Schedule Quality
- 37:27Field Test Against Adversaries: Mock Attackers Example from PROTECT
related talks
14mDivide and Conquer: A Compositional Approach to Game-Theoretic Security
24mAutomated game-theoretic security analysis of blockchain protocols
16mDesigning Holistic Security Programs
51mSecure Applications, By Design
56mWhen Unstoppable Frontier Agents Meet Immovable Attack Vectors
05mEnterprise security: A new hope
31mSpeed Panel: Next-Gen Game Security: How To Protect Your Players In Games.
51mThinking Differently About Security