A DRL-Empowered Multi-Level Jamming Approach for Secure Semantic Communication

2025-10-30

AI Summary

To address the security risk of semantic information leakage in semantic communications over MIMO fading wiretap channels, this paper proposes a deep reinforcement learning (DRL)-driven multi-level interference mechanism. The method innovatively integrates semantic-layer interference—encoding-agnostic text perturbation—with physical-layer interference—controllable Gaussian noise—within a two-tier precoding framework. A Deep Deterministic Policy Gradient (DDPG) algorithm is employed to dynamically optimize the precoding matrix, jointly enhancing legitimate user reception quality and degrading the eavesdropper’s semantic reconstruction capability. An alternating optimization training strategy enables joint maximization of semantic security and communication efficiency. Experimental results demonstrate that, under equivalent security guarantees, the proposed scheme improves the legitimate user’s peak signal-to-noise ratio (PSNR) by up to 0.6 dB compared to encryption-based (ESCS) and encoding-based jamming (EJ) baselines.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Reinforcement Learning
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Semantic communication (SemCom) aims to transmit only task-relevant information, thereby improving communication efficiency but also exposing semantic information to potential eavesdropping. In this paper, we propose a deep reinforcement learning (DRL)-empowered multi-level jamming approach to enhance the security of SemCom systems over MIMO fading wiretap channels. This approach combines semantic layer jamming, achieved by encoding task-irrelevant text, and physical layer jamming, achieved by encoding random Gaussian noise. These two-level jamming signals are superposed with task-relevant semantic information to protect the transmitted semantics from eavesdropping. A deep deterministic policy gradient (DDPG) algorithm is further introduced to dynamically design and optimize the precoding matrices for both taskrelevant semantic information and multi-level jamming signals, aiming to enhance the legitimate user's image reconstruction while degrading the eavesdropper's performance. To jointly train the SemCom model and the DDPG agent, we propose an alternating optimization strategy where the two modules are updated iteratively. Experimental results demonstrate that, compared with both the encryption-based (ESCS) and encoded jammer-based (EJ) benchmarks, our method achieves comparable security while improving the legitimate user's peak signalto-noise ratio (PSNR) by up to approximately 0.6 dB.

Problem

• Enhancing semantic communication security against eavesdropping over MIMO channels
• Optimizing multi-level jamming with semantic and physical layer signals
• Improving legitimate user performance while degrading eavesdropper reconstruction quality

Innovation

• Multi-level jamming combines semantic and physical layers
• DDPG algorithm optimizes precoding matrices dynamically
• Alternating optimization jointly trains SemCom and DRL agent