[Paper Review] Generative AI for Secure Physical Layer Communications: A Survey

Changyuan Zhao, Hongyang Du+7 more|arXiv (Cornell University)|Feb 21, 2024

Cellular Automata and Applications4 citations

TL;DR

This survey explores the application of Generative AI (GAI) in enhancing physical layer security for wireless communications, leveraging models like GANs, VAEs, and diffusion models to improve confidentiality, authentication, and resilience. It demonstrates GAI’s ability to adapt to dynamic channel conditions and generate synthetic data for robust threat detection and secure transmission.

ABSTRACT

Generative Artificial Intelligence (GAI) stands at the forefront of AI innovation, demonstrating rapid advancement and unparalleled proficiency in generating diverse content. Beyond content creation, GAI has significant analytical abilities to learn complex data distribution, offering numerous opportunities to resolve security issues. In the realm of security from physical layer perspectives, traditional AI approaches frequently struggle, primarily due to their limited capacity to dynamically adjust to the evolving physical attributes of transmission channels and the complexity of contemporary cyber threats. This adaptability and analytical depth are precisely where GAI excels. Therefore, in this paper, we offer an extensive survey on the various applications of GAI in enhancing security within the physical layer of communication networks. We first emphasize the importance of advanced GAI models in this area, including Generative Adversarial Networks (GANs), Autoencoders (AEs), Variational Autoencoders (VAEs), and Diffusion Models (DMs). We delve into the roles of GAI in addressing challenges of physical layer security, focusing on communication confidentiality, authentication, availability, resilience, and integrity. Furthermore, we also present future research directions focusing model improvements, multi-scenario deployment, resource-efficient optimization, and secure semantic communication, highlighting the multifaceted potential of GAI to address emerging challenges in secure physical layer communications and sensing.

Motivation & Objective

• Address the limitations of traditional AI in adapting to dynamic physical layer channel conditions and evolving cyber threats.
• Explore the potential of advanced GAI models—such as GANs, VAEs, and diffusion models—to model complex signal distributions and enhance security.
• Investigate GAI’s role in improving core security properties: confidentiality, authentication, availability, integrity, and resilience.
• Identify future research directions, including lightweight GAI models, multi-scenario deployment, and integration with semantic communications.
• Enable secure, efficient, and adaptive physical layer communication in 6G and IoT networks through GAI-driven solutions.

Proposed method

• Utilize Generative Adversarial Networks (GANs) to generate realistic channel state information and adversarial examples for intrusion detection.
• Apply Variational Autoencoders (VAEs) to learn latent representations of legitimate signals and detect anomalies via reconstruction error.
• Employ diffusion models to synthesize high-fidelity, diverse wireless signal samples for data augmentation and threat simulation.
• Integrate GAI with physical layer security mechanisms such as beamforming and artificial noise generation to enhance secrecy capacity.
• Propose distributed and federated learning frameworks to train GAI models on edge devices while preserving data privacy and reducing computational load.
• Introduce Mixture-of-Experts (MoE) architectures to dynamically route tasks to specialized GAI sub-models for scenario-specific security enforcement.

Figure 2: The overall architecture of the AE-Based [ 75 ] and VAE-Based secure transceiver [ 29 ] . Part A demonstrates a wiretap system model with AWGN. Part B illustrates the whole framework of AE-Based secure transceiver, which is trained by two loss functions: the mean-squared error between tran

Experimental results

Research questions

• RQ1 How can GAI models like GANs and VAEs improve the detection of physical layer anomalies and cyber-attacks in dynamic wireless environments?
• RQ2 In what ways can generative models enhance communication confidentiality and reduce information leakage in wireless transmissions?
• RQ3 How can GAI be optimized for deployment in resource-constrained environments such as IoT and edge devices?
• RQ4 What role can GAI play in enabling secure and efficient semantic communications by focusing on task-level performance rather than bit-level accuracy?
• RQ5 How can GAI be integrated with emerging technologies like XL-MIMO and near-field beam focusing to strengthen physical layer security in 6G networks?

Key findings

• GAI models such as GANs and VAEs demonstrate strong potential in modeling complex, non-stationary wireless channel characteristics for improved anomaly detection.
• Diffusion models can generate high-fidelity, diverse wireless signal samples, enabling robust data augmentation for training secure physical layer systems.
• Federated and lightweight GAI training approaches significantly reduce computational and communication overhead, enabling deployment on edge and IoT devices.
• The integration of GAI with Mixture-of-Experts (MoE) enables dynamic, scenario-aware security enforcement across heterogeneous wireless environments.
• GAI-aided semantic communications shift performance metrics from bit error rate (BER) to task fulfillment quality, enabling efficient, secure transmission in resource-constrained networks.
• GAI-enhanced beamforming and artificial noise generation techniques improve secrecy capacity and reduce eavesdropping risks in XL-MIMO-enabled 6G systems.

Figure 3: Proposed CGAN training architecture in [ 34 ] . In Part A , the conditional information is the previous magnitudes of the CSI elements associated with time. The output of the discriminator is the probability value, representing the likelihood from zero to one based on its perception of whe