List of Contributors xix
Foreword xxvii
About the Editors xxxi
Preface xxxv
Acknowledgments xxxvii
1 LLMs Are Not Few-shot Threat Hunters 1
Glenn A. Fink, Luiz M. Pereira, and Christian W. Stauffer
1.1 Overview 1
1.1.1 AI Is Not Magic 1
1.1.2 Inherent Difficulty of Human Tasks in Cybersecurity and Threat Hunting 3
1.2 Large Language Models 4
1.2.1 Background 4
1.2.2 Transformers 4
1.2.3 Pretraining and Fine-tuning 9
1.2.4 General Limitations 9
1.3 Threat Hunters 12
1.3.1 Introduction to Threat Hunting 12
1.3.2 The Dimensions of Threat Hunting 13
1.3.3 The Approaches to Threat Hunting 15
1.3.4 The Process of Threat Hunting 16
1.3.5 Challenges to Modern Threat Hunting 17
1.4 Capabilities and Limitations of LLMs in Cybersecurity 18
1.4.1 General Limitations of LLMs for Cybersecurity 18
1.4.2 General Capabilities of LLMs Useful for Cybersecurity 20
1.4.3 Applications of LLMs in Cybersecurity 22
1.5 Conclusion: Reimagining LLMs as Assistant Threat Hunter 24
References 27
2 LLMs on Support of Privacy and Security of Mobile Apps: State-of-the-art and Research Directions 29
Tran Thanh Lam Nguyen, Barbara Carminati, and Elena Ferrari
2.1 Introduction 29
2.2 Background on LLMs 32
2.2.1 Large Language Models 32
2.2.2 FSL and RAG 39
2.3 Mobile Apps: Main Security and Privacy Threats 43
2.4 LLM-based Solutions: State-of-the-art 47
2.4.1 Vulnerabilities Detection 48
2.4.2 Bug Detection and Reproduction 50
2.4.3 Malware Detection 52
2.5 An LLMs-based Approach for Mitigating Image Metadata Leakage Risks 53
2.6 Research Challenges 57
2.7 Conclusion 60
Acknowledgment 61
References 61
3 Machine Learning-based Intrusion Detection Systems: Capabilities, Methodologies, and Open Research Challenges 67
Chaoyu Zhang, Ning Wang, Y. Thomas Hou, and Wenjing Lou
3.1 Introduction 67
3.2 Basic Concepts and ML for Intrusion Detection 69
3.2.1 Fundamental Concepts 69
3.2.2 ml Algorithms for Intrusion Detection 70
3.2.3 Taxonomy of IDSs 72
3.2.4 Evaluation Metrics and Datasets 73
3.3 Capability I: Zero-day Attack Detection with ml 75
3.3.1 Understanding Zero-day Attacks and Their Impact 75
3.3.2 General Workflow of ML-IDS for Identifying Zero-day Attacks 75
3.3.3 Anomaly Detection Mechanisms 76
3.3.4 Open Research Challenges 77
3.4 Capability II: Intrusion Explainability Through XAI 79
3.4.1 Enhancing Transparency and Trust in Intrusion Detection 79
3.4.2 General Workflow of XAI 80
3.4.3 XAI Methods for IDS Transparency Enhancement 80
3.4.4 Open Research Challenges 83
3.5 Capability III: Intrusion Detection in Encrypted Traffic 84
3.5.1 Challenges in Intrusion Detection for Encrypted Traffic 84
3.5.2 Workflow of ML-IDS for Encrypted Traffic 84
3.5.3 ML-based Solutions for Encrypted Traffic Analysis 84
3.5.4 Open Research Challenges 87
3.6 Capability IV: Context-aware Threat Detection and Reasoning with GNNs 88
3.6.1 Introduction to GNNs in IDS 88
3.6.2 Workflow of GNNs for Intrusion Detection 88
3.6.3 Provenance-based Intrusion Detection by GNNs 89
3.6.4 Open Research Challenges 92
3.7 Capability V: LLMs for Intrusion Detection and Understanding 93
3.7.1 The Role of LLMs in Cybersecurity 93
3.7.2 Leveraging LLMs for Intrusion Detection 94
3.7.3 A Review of LLM-based IDS 94
3.7.4 Open Research Challenges 97
3.8 Summary 97
References 98
4 Generative AI for Advanced Cyber Defense 109
Moqsadur Rahman, Aaron Sanchez, Krish Piryani, Siddhartha Das, Sai Munikoti, Luis de la Torre Quintana, Monowar Hasan, Joseph Aguayo, Monika Akbar, Shahriar Hossain, and Mahantesh Halappanavar
4.1 Introduction 109
4.2 Motivation and Related Work 111
4.2.1 AI-supported Vulnerability Management 112
4.3 Foundations for Cyber Defense 114
4.3.1 Mapping Vulnerabilities, Weaknesses, and Attack Patterns Using LLMs 115
4.4 Retrieval-augmented Generation 117
4.5 KG and Querying 118
4.5.1 Graph Schema 119
4.5.2 Neo4j KG Implementation 122
4.5.3 Cypher Queries 123
4.6 Evaluation and Results 126
4.6.1 RAG-based Response Generation 127
4.6.2 CWE Predictions Using RAG 131
4.6.3 CWE Predictions Using GPT4-o 136
4.7 Conclusion 142
References 142
5 Enhancing Threat Detection and Response with Generative AI and Blockchain 147
Driss El Majdoubi, Souad Sadki, Zakia El Uahhabi, and Mohamed Essaidi
5.1 Introduction 147
5.2 Cybersecurity Current Issues: Background 148
5.3 Blockchain Technology for Cybersecurity 150
5.3.1 Blockchain Benefits for Cybersecurity 150
5.3.2 Existing Blockchain-based Cybersecurity Solutions 153
5.4 Combining Generative AI and Blockchain for Cybersecurity 156
5.4.1 Integration of Generative AI and Blockchain 160
5.4.2 Understanding Capabilities and Risks 160
5.4.3 Practical Benefits for Cybersecurity 161
5.4.4 Limitations and Open Research Issues 161
5.5 Conclusion 162
References 163
6 Privacy-preserving Collaborative Machine Learning 169
Runhua Xu and James Joshi
6.1 Introduction 169
6.1.1 Objectives and Structure 171
6.2 Collaborative Learning Overview 172
6.2.1 Definition and Characteristics 172
6.2.2 Related Terminologies 174
6.2.3 Collaborative Decentralized Learning and Collaborative Distributed Learning 175
6.3 Collaborative Learning Paradigms and Privacy Risks 177
6.3.1 Key Collaborative Approaches 177
6.3.2 Privacy Risks in Collaborative Learning 182
6.3.3 Privacy Inference Attacks in Collaborative Learning 183
6.4 Privacy-preserving Technologies 187
6.4.1 The Need for Privacy Preservation 187
6.4.2 Privacy-preserving Technologies 188
6.5 Conclusion 195
References 196
7 Security and Privacy in Federated Learning 203
Zhuosheng Zhang and Shucheng Yu
7.1 Introduction 203
7.1.1 Federated Learning 203
7.1.2 Privacy Threats in FL 205
7.1.3 Security Issues in FL 207
7.1.4 Characterize FL 211
7.2 Privacy-preserving FL 215
7.2.1 Secure Multiparty Computation 215
7.2.2 Trust Execution Environments 216
7.2.3 Secure Aggregation 217
7.2.4 Differential Privacy 218
7.3 Enhance Security in FL 219
7.3.1 Data-poisoning Attack and Nonadaptive Model-poisoning Attack 220
7.3.2 Model-poisoning Attack 222
7.4 Secure Privacy-preserving FL 225
7.4.1 Enhancing Security in FL with DP 225
7.4.2 Verifiability in Private FL 226
7.4.3 Security in Private FL 227
7.5 Conclusion 228
References 229
8 Machine Learning Attacks on Signal Characteristics in Wireless Networks 235
Yan Wang, Cong Shi, Yingying Chen, and Zijie Tang
8.1 Introduction 235
8.2 Threat Model and Targeted Models 239
8.2.1 Backdoor Attack Scenarios 239
8.2.2 Attackers' Capability 240
8.2.3 Attackers' Objective 240
8.2.4 Targeted ML Models 241
8.3 Attack Formulation and Challenges 241
8.3.1 Backdoor Attack Formulation 241
8.3.2 Challenges 244
8.4 Poison-label Backdoor Attack 246
8.4.1 Stealthy Trigger Designs 246
8.4.2 Backdoor Trigger Optimization 249
8.5 Clean-label Backdoor Trigger Design 252
8.5.1 Clean-label Backdoor Trigger Optimization 253
8.6 Evaluation 255
8.6.1 Victim ML Model 255
8.6.2 Experimental Methodology 255
8.6.3 RF Backdoor Attack Performance 257
8.6.4 Resistance to Backdoor Defense 259
8.7 Related Work 261
8.8 Conclusion 262
References 263
9 Secure by Design 267
Mehdi Mirakhorli and Kevin E. Greene
9.1 Introduction 267
9.1.1 Definitions and Contexts 268
9.1.2 Core Principles of "Secure by Design" 269
9.1.3 Principle of Compartmentalization and Isolation 273
9.2 A Methodological Approach to Secure by Design 275
9.2.1 Assumption of Breach 275
9.2.2 Misuse and Abuse Cases to Drive Secure by Design 276
9.2.3 Secure by Design Through Architectural Tactics 277
9.2.4 Shifting Software Assurance from Coding Bugs to Design Flaws 282
9.3 AI in Secure by Design: Opportunities and Challenges 283
9.4 Conclusion and Future Directions 284
References 284
10 DDoS Detection in IoT Environments: Deep Packet Inspection and Real-world Applications 289
Nikola Gavric, Guru Bhandari, and Andrii Shalaginov
10.1 Introduction 289
10.2 DDoS Detection Techniques in Research 294
10.2.1 Network-based Intrusion Detection Systems 295
10.2.2 Host-based Intrusion Detection Systems 300
10.3 Limitations of Research Approaches 303
10.4 Industry Practices for DDoS Detection 305
10.5 Challenges in DDoS Detection 309
10.6 Future Directions 311
10.7 Conclusion 313
References 314
11 Data Science for Cybersecurity: A Case Study Focused on DDoS Attacks 317
Michele Nogueira, Ligia F. Borges, and Anderson B. Neira
11.1 Introduction 317
11.2 Background 319
11.2.1 Cybersecurity 320
11.2.2 Data Science 326
11.3 State of the Art 333
11.3.1 Data Acquisition 334
11.3.2 Data Preparation 335
11.3.3 Feature Preprocessing 336
11.3.4 Data Visualization 337
11.3.5 Data Analysis 338
11.3.6 ml in Cybersecurity 339
11.4 Challenges and Opportunities 340
11.5 Conclusion 341
Acknowledgments 342
References 342
12 AI Implications for Cybersecurity Education and Future Explorations 347
Elizabeth Hawthorne, Mihaela Sabin, and Melissa Dark
12.1 Introduction 347
12.2 Postsecondary Cybersecurity Education: Historical Perspective and Current Initiatives 348
12.2.1 ACM Computing Curricula 348
12.2.2 National Centers for Academic Excellence in Cybersecurity 356
12.2.3 ABET Criteria 359
12.3 Cybersecurity Policy in Secondary Education 361
12.3.1 US High School Landscape 362
12.4 Conclusion 367
12.5 Future Explorations 368
References 368
13 Ethical AI in Cybersecurity: Quantum-resistant Architectures and Decentralized Optimization Strategies 371
Andreou Andreas, Mavromoustakis X. Constandinos, Houbing Song, and Jordi Mongay Batalla
13.1 Introduction 371
13.1.1 Motivation 372
13.1.2 Contribution 373
13.1.3 Novelty 373
13.2 Literature Review 373
13.3 Overview and Ethical Considerations in AI-centric Cybersecurity 374
13.4 AML and Privacy Risks in AI Systems 378
13.5 Forensic and Formal Methods for AI Security 380
13.5.1 Auditing Tools for Security and Privacy 383
13.5.2 Transparency, Interpretability, and Trust 383
13.5.3 Building Secure and Trustworthy AI Systems 384
13.6 Generative AI and Quantum-resistant Architectures in Cybersecurity 385
13.6.1 Opportunities and Risks 385
13.6.2 Threats and Countermeasures 386
13.6.3 Strategies for Resilience 387
13.7 Future Directions and Ethical Considerations 387
13.8 Conclusion 390
References 391
14 Security Threats and Defenses in AI-enabled Object Tracking Systems 397
Mengjie Jia, Yanyan Li, and Jiawei Yuan
14.1 Introduction 397
14.2 Related Works 398
14.2.1 UAV Object Tracking 398
14.2.2 Adversarial Tracking Attacks 399
14.2.3 Robustness Enhancement Against Attacks 400
14.3 Methods 401
14.3.1 Model Architecture 403
14.3.2 Decision Loss 403
14.3.3 Feature Loss 404
14.3.4 l 2 Norm loss 405
14.4 Evaluation 405
14.4.1 Experiment Setup 405
14.4.2 Evaluation Metrics 405
14.4.3 Results 406
14.4.4 Tracking Examples 409
14.5 Conclusion 413
Acknowledgment 413
References 413
15 AI for Android Malware Detection and Classification 419
Safayat Bin Hakim, Muhammad Adil, Kamal Acharya, and Houbing Herbert Song
15.1 Introduction 419
15.1.1 Security Threats in Android Applications 420
15.1.2 Challenges in Android Malware Detection 422
15.1.3 Current Approaches and Limitations 423
15.2 Design of the Proposed Framework 424
15.2.1 Core Components and Architecture 424
15.2.2 Feature Extraction with Attention Mechanism 425
15.2.3 Feature Extraction with Attention Mechanism 425
15.2.4 Dimensionality Reduction and Optimization 427
15.2.5 Classification Using SVMs 427
15.3 Implementation and Dataset Overview 428
15.3.1 Dataset Insights 428
15.3.2 Preprocessing Strategies 429
15.3.3 Handling Class Imbalance 429
15.3.4 Adversarial Training and Evaluation 429
15.4 Results and Insights 431
15.4.1 Experimental Setup 431
15.4.2 Performance Analysis 435
15.4.3 Performance Insights with Visualization 436
15.4.4 Benchmarking Against Existing Methods 438
15.4.5 Key Insights 439
15.5 Feature Importance Analysis 439
15.5.1 Top Feature Importance 439
15.5.2 Feature Impact Analysis Using SHAP Values 441
15.5.3 Global Feature Impact Distribution 442
15.6 Comparative Analysis and Advancements over Existing Methods 442
15.6.1 Feature Space Optimization 444
15.6.2 Advances in Adversarial Robustness 445
15.6.3 Performance Improvements 445
15.6.4 Summary of Key Advancements 445
15.7 Discussion 446
15.7.1 Limitations and Future Work 446
15.8 Conclusion 447
References 447
16 Cyber-AI Supply Chain Vulnerabilities 451
Joanna C. S. Santos
16.1 Introduction 451
16.2 AI/ML Supply Chain Attacks via Untrusted Model Deserialization 452
16.2.1 Model Deserialization 453
16.2.2 AI/ML Attack Scenarios 457
16.3 The State-of-the-art of the AI/ML Supply Chain 458
16.3.1 Commonly Used Serialization Formats 458
16.3.2 Deliberately Malicious Models Published on Hugging Face 460
16.3.3 Developers' Perception on Safetensors 462
16.4 Conclusion 466
16.4.1 Implications for Research 466
16.4.2 Implications for Practitioners 467
References 467
17 AI-powered Physical Layer Security in Industrial Wireless Networks 471
Hong Wen, Qi Wang, and Zhibo Pang
17.1 Introduction 471
17.2 Radio Frequency Fingerprint Identification 474
17.2.1 System Model 474
17.2.2 Cross-device RFFI 476
17.2.3 Experimental Investigation 480
17.3 CSI-based PLA 481
17.3.1 System Model 482
17.3.2 Transfer Learning-based PLA 484
17.3.3 Data Augmentation 488
17.3.4 Experimental Investigation 490
17.4 PLK Distribution 493
17.4.1 System Model 493
17.4.2 AI-powered Quantization 495
17.5 Physical Layer Security Enhanced ZT Security Framework 498
17.5.1 ZT Requirements in IIoT 499
17.5.2 PLS Enhanced ZT Security Framework 500
References 502
18 The Security of Reinforcement Learning Systems in Electric Grid Domain 505
Suman Rath, Zain ul Abdeen, Olivera Kotevska, Viktor Reshniak, and Vivek Kumar Singh
18.1 Introduction 505
18.2 RL for Control 506
18.2.1 Overview of RL Algorithms 506
18.2.2 DQN Algorithm 510
18.3 Case Study: RL for Control in Cyber-physical Microgrids 513
18.4 Related Work: Grid Applications of RL 516
18.5 Open Challenges and Solutions 518
18.6 Conclusion 522
Acknowledgments 524
References 524
19 Geopolitical Dimensions of AI in Cybersecurity: The Emerging Battleground 533
Felix Staicu and Mihai Barloiu
19.1 Introduction 533
19.1.1 A Conceptual Framework 534
19.2 Foundations of AI in Geopolitics: From Military Origins to Emerging Strategic Trajectories 536
19.2.1 Historical Foundations: The Military and Intelligence Roots of Key Technologies 536
19.2.2 Early International Debates on AI Governance and Their Geopolitical Dimensions 537
19.2.3 The Two-way Influence Between AI and Geopolitics: Early Signals of Strategic Catalysts and Normative Vectors 538
19.3 The Contemporary Battleground: AI as a Strategic Variable 540
19.3.1 AI-infused IO: Precision, Persistence, and Policy Dilemmas 540
19.3.2 Fusion Technologies for Battlefield Control, Unmanned Vehicles, and AI Swarming 542
19.3.3 Regulatory Power as Soft Power: Competing Models for Global AI Norms 543
19.3.4 Global Rivalries: The US-China AI Race and the Fragmenting Digital Ecosystem 545
19.4 Beyond Today's Conflicts: Future Horizons in AI-driven Security 548
19.4.1 2050 Hypothesis-driven Scenarios in the International System 548
19.4.2 AI in the Nuclear Quartet 551
19.4.3 AI in Kinetic Conventional Military Capabilities 553
19.4.4 AI in Cybersecurity and Information Warfare 554
19.4.5 A Holistic View of AI's Impact on International Security 556
19.5 Conclusions and Recommendations 558
19.5.1 Integrative Insights 558
19.6 Conclusion 560
Acknowledgments 561
References 561
20 Robust AI Techniques to Support High-consequence Applications in the Cyber Age 567
Joel Brogan, Linsey Passarella, Mark Adam, Birdy Phathanapirom, Nathan Martindale, Jordan Stomps, Olivera Kotevska, Matthew Yohe, Ryan Tokola, Ryan Kerekes, and Scott Stewart
20.1 Introduction 567
20.2 Motivation 568
20.3 Explainability Measures for Deep Learning in High-consequence Scenarios 570
20.3.1 Gradient-based Methods 571
20.3.2 Perturbation-based Methods 572
20.3.3 Comparisons Between Explainability Methods 572
20.4 Improving Confidence and Robustness Measures for Deep Learning in Critical Decision-making Scenarios 573
20.4.1 Introduction 573
20.4.2 Dataset Description 574
20.4.3 Methodology 575
20.4.4 Attribution Algorithms 576
20.4.5 Confidence Measure Algorithms 576
20.4.6 Results and Analysis 581
20.4.7 Discussion and Future Work 581
20.5 Building Robust AI Through SME Knowledge Embeddings 583
20.5.1 Explicit Knowledge in Structured Formats 586
20.5.2 Fine-tuning and Evaluating Foundation Models 587
20.6 Flight-path Vocabularies for Foundation Model Training 588
20.6.1 Introduction 588
20.6.2 Dataset 589
20.6.3 Methodology 590
20.6.4 Results and Discussion 591
20.7 Promise and Peril of Foundation Models in High-consequence Scenarios 592
20.7.1 Adversarial Vulnerabilities of Foundation Models 593
20.7.2 Privacy Violation Vulnerabilities in Foundation Models 594
20.7.3 Alignment Hazards When Training Foundation Models 594
20.7.4 Performance Hazards When Inferring and Generating with Foundation Models 595
20.8 Discussion 596
Acknowledgments 596
References 596
Index 601
