Full Description
This book is intended to give researchers and practitioners in the cross-cutting fields of artificial intelligence, machine learning (AI/ML) and cyber security up-to-date and in-depth knowledge of recent techniques for improving the vulnerabilities of AI/ML systems against attacks from malicious adversaries. The ten chapters in this book, written by eminent researchers in AI/ML and cyber-security, span diverse, yet inter-related topics including game playing AI and game theory as defenses against attacks on AI/ML systems, methods for effectively addressing vulnerabilities of AI/ML operating in large, distributed environments like Internet of Things (IoT) with diverse data modalities, and, techniques to enable AI/ML systems to intelligently interact with humans that could be malicious adversaries and/or benign teammates. Readers of this book will be equipped with definitive information on recent developments suitable for countering adversarial threats in AI/ML systems towards making them operate in a safe, reliable and seamless manner.
Contents
Part I: Game-Playing AI and Game Theory-based Techniques for Cyber Defenses.- 1. Rethinking Intelligent Behavior as Competitive Games for Handling Adversarial Challenges to Machine Learning.- 2. Security of Distributed Machine Learning:A Game-Theoretic Approach to Design Secure DSVM.- 3. Be Careful When Learning Against Adversaries: Imitative Attacker Deception in Stackelberg Security Games.- Part II: Data Modalities and Distributed Architectures for Countering Adversarial Cyber Attacks.- 4. Adversarial Machine Learning in Text: A Case Study of Phishing Email Detection with RCNN model.- 5. Overview of GANs for Image Synthesis and Detection Methods.- 6. Robust Machine Learning using Diversity and Blockchain.- Part III: Human Machine Interactions and Roles in Automated Cyber Defenses.- 7. Automating the Investigation of Sophisticated Cyber Threats with Cognitive Agents.- 8. Integrating Human Reasoning and Machine Learning to Classify Cyber Attacks.- 9. Homology as an Adversarial Attack Indicator.- Cyber-(in)security, revisited: Proactive Cyber-defenses, Interdependence and Autonomous Human Machine Teams (A-HMTs).
