1st Edition
Adversarial Reasoning Computational Approaches to Reading the Opponent’s Mind
The rising tide of threats, from financial cybercrime to asymmetric military conflicts, demands greater sophistication in tools and techniques of law enforcement, commercial and domestic security professionals, and terrorism prevention. Concentrating on computational solutions to determine or anticipate an adversary's intent, Adversarial Reasoning: Computational Approaches to Reading the Opponent's Mind discusses the technologies for opponent strategy prediction, plan recognition, deception discovery and planning, and strategy formulation that not only applies to security issues but also to game industry and business transactions.
Addressing a broad range of practical problems, including military planning and command, military and foreign intelligence, antiterrorism, network security, as well as simulation and training systems, this reference presents an overview of each problem and then explores various approaches and applications to understand the minds and negate the actions of your opponents. The techniques discussed originate from a variety of disciplines such as stochastic processes, artificial intelligence planning, cognitive modeling, robotics and agent theory, robust control, game theory, and machine learning, among others. The beginning chapters outline the key concepts related to discovering the opponent's intent and plans while the later chapters journey into mathematical methods for counterdeception. The final chapters employ a range of techniques, including reinforcement learning within a stochastic dynamic games context to devise strategies that combat opponents.
By answering specific questions on how to create practical applications that require elements of adversarial reasoning while also exploring theoretical developments, Adversarial Reasoning: Computational Approaches to Reading the Opponent's Mind is beneficial for practitioners as well as researchers.
Intent Inferencing
Representing and Reasoning under Uncertainty
Adversary Intent Inferencing Model (AII)
Future Work
References
HUMAN FACTORS IN OPPONENT INTENT
Intent Recognition in Human Opponents
A Cognitive Approach to Modeling Opponents
Knowledge-Based Abduction
Creating a Model of Opponent's Intent
Knowledge-Based Intention Projection
KIP Architecture
Heuristics for Reducing the Number of Opponent Goals
Evidence-Based Goal Selection
Experimental Results
References
EXTRAPOLATION OF THE OPPONENT'S PAST BEHAVIORS
Standing on the Shoulders of Giants
Ant-Like Agents with Humanistic Behavior
Exploring Possible Worlds
Playing Together with Other Approaches
Experimental Experience
Looking Ahead
References
PLAN RECOGNITION
An Example of Plans and Plan Recognition
PHATT System Basics
Algorithmic Complexity and Scalability
Handling Partial Observability
Limitations of the PHATT Algorithm
Lessons Learned: Computer Network Security
References
DETECTING DECEPTION
Why Deception Works
Detecting Deception
Implementing ACH-CD
Applying Automated ACH-CD to D-Day
Application without Automation: The Battle of Midway
Future Applications of ACH-CD
References
DECEPTION AS A SEMANTIC ATTACK
Semantic Attacks in Relation to Other Topics in This Book
Changing the Behavior of Humans
Perception Management
Semantic Attacks and Information Warfare
Deception Detection
Semantic Attacks and Intelligence and Security Informatics
Current Countermeasures for Semantic Attacks
New Countermeasures for Semantic Attacks
Information Trajectory Modeling
Linguistic Countermeasures to Semantic Attacks
News Verification: An Instance of Multiple
Source Semantic Attacks
Process Query Systems for Information Trajectory Countermeasures
Tracking Hypotheses in the Financial Fraud Domain
References
APPLICATION AND VALUE OF DECEPTION
Information, Computation, and Deception
Games of Deception
The Rational Side of Deception
References
ROBUSTNESS AGAINST DECEPTION
In Search of Anti-Deception
Modeling the Game
Deception-Rejection Machines
A Seemingly Simple Game
Analysis of the Fully-Observable Case
Analysis of Partially-Observed Case
Pruning
Comparison of the Risk-Averse and Deception-Robust Approaches
Implementation
References
THE ROLE OF IMPERFECT INFORMATION
Classical Game-Tree Search
Game-Tree Search in Imperfect Information Games
Case Study: Texas Hold'em
Case Study: Kriegspiel Chess
Summary
References
HANDLING PARTIAL AND CORRUPTED INFORMATION
The Deterministic Discrete-Time Linear-Quadratic
Game Formulation
The Discrete-Time LQG Game: Formulation and Previous Work
LQG Game with Partial Information:
An Indirect Approach
Properties of the Blocking Filter
Effects of Partial Information
Properties of the Equilibrium Cost: The Saddle Interval Application of Results to an Adversarial Environment
References
STRATEGIES IN LARGE-SCALE PROBLEMS
Game-Solving Approaches in Practical Problems
Overview of LG
Game Construction
Game Solving
Accuracy of the Solution
Scale of the Problems
Appendix: State Transition Systems
References
LEARNING IN STRATEGIZE
Our Wargame
Reducing the State-Action Space Dimension through Symmetries and Heuristics
Review of Reinforcement Learning Methods
Results
Future Possibilities for Abstraction and Approximation in Complex Domains
References
LEARNING FROM AND ABOUT THE OPPONENT
Fictitious Play
Reinforcement Learning
Extensive Game
Learning in Extensive Games
Multi-Agent Learning Automaton (MLA)
MQ-Learning
MLA Experiment
MQ-Learning Experiment
Toward Practical Applications
References
Index
Biography
Alexander Kott, William M. McEneaney
