Generative AI for Communications Systems

Diep N. Nguyen is the Head of UTS 5G/6G Lab with the Faculty of Engineering and Information Technology at the University of Technology Sydney (UTS), Sydney, NSW, Australia. Nam H. Chu is a Faculty Member with the Department of Telecommunications Engineering at the University of Transport and Communications, Hanoi, Vietnam. He is also with the University of Technology Sydney (UTS), Australia, and the Crown Institute of Higher Education, Australia. Dinh Thai Hoang is a Faculty Member at the University of Technology Sydney (UTS), Australia. Octavia A. Dobre is a Professor and Canada Research Chair Tier-1 at Memorial University, Canada. Dusit Niyato is a President’s Chair Professor in Computer Science and Engineering in the College of Computing and Data Science at Nanyang Technological University, Singapore. Petar Popovski is currently a Professor with Aalborg University in Denmark where he heads the Section on Connectivity. He is also a Visiting Excellence Chair with the University of Bremen, Germany.

• List of Contributors xiiiPreface xxiiiAcronyms xxix1 Future AI-empowered Communications Systems 1Nguyen Van Huynh, Thien Huynh-The, and Quoc-Viet Pham1.1 Fundamental Background of Future Communications Systems 11.1.1 Overview of Future Communications Systems 11.1.2 Key Challenges and Research Trends 71.2 AI-powered Communication Enablers 101.2.1 Deep Learning-based Approaches 101.2.2 Reinforcement Learning-based Approaches 171.2.3 Federated/Distributed Learning-based Approaches 241.2.4 Existing Challenges 281.2.5 Potential of Generative AI 281.3 Conclusion 30References 302 Generative AI Background and Its Potentials for Future Communications Systems 39Asmaa Abdallah, Abdulkadir Celik, and Ahmed M. Eltawil2.1 Introduction 392.2 A Taxonomy of Generative Models 402.2.1 Explicit Density Models 412.2.2 Implicit Density Models 412.2.3 Ways GenAI Complements Discriminative AI 422.3 Prominent Generative Models 422.3.1 Generative Adversarial Networks 422.3.2 Variational Autoencoders 442.3.3 Flow-based Generative Models 472.3.4 Diffusion-based Generative Models 492.3.5 The Trilemma of GMs 512.3.6 Generative Autoregressive Models 522.3.7 Generative Transformers and LLMs 542.3.8 Strategies to Address LLM Limitations 592.4 GenAI Applications to Canonical Problems in Communications Systems 632.4.1 Physical Layer Design 632.4.2 Network Resource Management 642.4.3 Network Traffic Analytics 652.4.4 Cross-layer Network Security 652.4.5 Localization and Positioning 662.5 Future Communication Frontiers for GMs 662.5.1 Semantic Communications 662.5.2 Integrated Sensing and Communications 672.5.3 Digital Twins 682.5.4 AI-generated Content for 6G Networks 692.5.5 MEC and EAI 692.5.6 Adversarial Machine Learning and Trustworthy AI 702.6 Regulation and Policy 712.7 Summary 71References 723 Key Study Cases of Generative AI Applications to Communications Systems 79Mehdi Letafati, Samad Ali, and Matti Latva-aho3.1 Overview on the Roles of Generative AI in Communication Systems 793.1.1 Use Cases of Generative Adversarial Networks in Communications 793.1.2 Use cases of VAEs in Communications 813.1.3 Use-cases of Diffusion Models in Communications 823.2 Case Study: Diffusion Models in Wireless Communications 833.2.1 Working Mechanism of Diffusion Models 833.2.2 Case Study: Diffusion Models Applications for Data Reconstruction Enhancement in Communication Systems 893.3 Future Implications and Potential Impacts on Communication Systems 983.4 Chapter Summary 99References 994 Generative AI at PHY Layer: Native AI or Trainable Radios 105Eren Balevi4.1 Wireless Communications Empowered with Generative Models 1054.1.1 Motivations of GenAI at the PHY 1054.1.2 Applications of GenAI at the PHY 1064.2 Channel Modeling 1104.2.1 Generative Channel Modeling 1114.2.2 Site-specific Generative Models 1154.3 Generative Channel Estimation 1164.3.1 Narrowband Channel Estimation with Reduced Pilots 1204.3.2 Wideband Channel Estimation with Reduced Pilots 1224.4 Channel Compression 1234.5 Beamforming 1254.6 Summary 129References 1295 Generative AI at the MAC Layer 133Kemal Davaslioglu, Ender Ayanoglu, and Yalin E. Sagduyu5.1 Introduction 1335.2 Generative Models 1375.2.1 Variational Autoencoders 1395.2.2 Generative Adversarial Networks 1405.2.3 Diffusion Models 1415.3 Spectrum Awareness Applications 1425.3.1 Data Augmentation and Synthetic Data Generation 1435.3.2 Signal Classification Applications – UAV Classification 1475.3.3 Anomaly Detection in RF Spectrum 1485.4 RF Spectrum Security Applications 1495.4.1 Emitter Identification 1495.4.2 Wireless Spoofing 1515.4.3 Enhanced Jamming Attacks 1525.5 Scheduling Applications 1535.5.1 Traffic Prediction and Pattern Generation 1535.5.2 Adaptive Scheduling Algorithms 1545.5.3 Interference Patterns 1545.5.4 Fairness and QoS 1545.5.5 Millimeter-wave Networks 1555.6 Open Problems and Future Research Directions 1555.6.1 Reconfigurable Intelligent Surface (RIS)-assisted Networks 1565.6.2 Spectrum Sharing in the Presence of Interference 1575.6.3 Integrated Sensing and Communications (ISAC) 1595.6.4 Link Scheduling in Large Networks 1605.6.5 Enhancing Wireless MAC-layer Security 1605.7 Concluding Remarks 162References 1626 Generative AI at Network Layer 169Athanasios Karapantelakis, Pegah Alizadeh, Abdulrahman Alabbasi, Kaushik Dey, and Alexandros Nikou6.1 Introduction 1696.2 Network Layer in Mobile Networks 1726.2.1 Radio Access Network 1726.2.2 Core Network 1756.3 Generative AI in the Network Layer 1776.3.1 Introduction 1776.3.2 Advantages of GenAI Models 1776.3.3 Short-term Applications (GenAI for Network Layer) 1806.3.4 Long-term Applications (Network Layer for GenAI) 1846.4 Challenges and Opportunities for GenAI in the Network Layer 1856.4.1 Challenges 1856.4.2 Research Opportunities 1866.5 Summary 187References 1877 Generative AI at Application Layer: Mobile AI-generated Content 191Paria Mohammadzadeh Hesar, Amirhossein Mohammadi, and Hina Tabassum7.1 Introduction to AIGC 1917.1.1 General Overview 1917.1.2 AIGC in the Application Layer 1927.1.3 AIGC Product Lifecycle 1947.2 Collaborative Network Infrastructure for Enabling GenAI Services 1967.2.1 Enabling AIGC – Challenges 1967.2.2 Infrastructure Components and Capabilities 1987.2.3 Collaborative Edge-cloud Infrastructure 2017.3 Network Resource Efficient GenAI Methods 2037.3.1 Model Optimization Techniques 2037.3.2 Service Optimization Methods 2067.4 Security and Privacy at Application Layer 2087.4.1 Security Threat Models and Privacy Risks 2097.4.2 Ethical Considerations in AIGC services 2117.4.3 Enabling Secure AIGC-as-a-Service 2127.5 Use Cases of Mobile AIGC 2137.5.1 AI-generated Content in Social Media 2137.5.2 Immersive Streaming (AR/VR) 2157.5.3 Personalized AI Services 2197.6 Conclusion and Research Directions 2227.7 Summary 223References 2248 Applications of GenAI on Wireless and Cybersecurity 239Brian Kim, Yalin E. Sagduyu, Tugba Erpek, Yi Shi, and Sennur Ulukus8.1 Introduction to GenAI in Wireless and Cybersecurity 2398.2 Adversarial Machine Learning in Wireless Communications 2428.2.1 Different Types of Attacks Against GenAI-driven Wireless Applications 2438.2.2 Defense Against Adversarial Attacks for GenAI-driven Wireless Applications 2458.3 GenAI for Wireless Security and Cybersecurity 2458.3.1 GenAI for Wireless Security 2458.3.2 GenAI for Cybersecurity 2488.3.3 GenAI-driven Attacks Against Wireless and Cybersecurity Applications 2498.4 Ethical Issues Related to GenAI for Wireless Communications and Cybersecurity 2518.5 Summary 252References 2539 Challenges and Opportunities for Generative AI in Wireless Communications and Networking 261Songyang Zhang and Zhi Ding9.1 Introduction 2619.2 Challenges of Applying Generative AI in Wireless Communications 2629.2.1 Efficiency and Robustness 2639.2.2 Cost and Complexity 2679.2.3 Standardization, Regulation, and Policy 2699.3 Adopting Generative AI in NextG Communications: Case Studies 2709.3.1 Integration of Generative AI and Physical Communications Models 2709.3.2 Trustworthy Generative AI for Distributed Wireless Communications 2779.4 Summary 281References 28110 Future Research Directions 285Nam H. Chu, Diep N. Nguyen, Dinh Thai Hoang, Octavia A. Dobre, Dusit Niyato, and Petar Popovski10.1 Introduction 28510.2 Emerging Foundational Research Frontiers 28610.2.1 Dedicated GenAI Models for Communication Systems 28710.2.2 Fusion of GenAI and Emerging Technologies 28810.3 Enhancing Generative AI Models for Wireless Communication Systems 29010.3.1 Model Optimization and Generalization 29010.3.2 Energy Efficiency 29110.3.3 Generative AI for Spectrum Management 29210.3.4 AI-driven Network Management and Orchestration 29310.3.5 Security and Privacy Concerns 29510.4 Practical Case Studies 29610.4.1 AI-powered Network Optimization by T-Mobile 29610.4.2 DeepSig’s Generative AI for Wireless Communications 29710.5 Conclusion 297References 298Index 305