Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Weixuan Chen, Qianqian Yang

TL;DR

This work targets secure semantic communication over challenging MIMO fading wiretap channels by introducing a DRL-enabled multi-level jamming scheme that combines semantic-layer jamming (text-encoded) and physical-layer jamming (Gaussian noise). A DDPG agent jointly designs three precoders to balance high PSNR for the legitimate user with degraded performance at the eavesdropper, trained via a five-stage alternating strategy that integrates SemCom and reinforcement learning. Experimental results on CIFAR-10 and related text benchmarks show that the proposed approach achieves security comparable to encryption-based and encoded-jammer baselines while improving the legitimate receiver’s PSNR by up to about 0.6 dB across various SNRs and compression ratios. The method offers a practical secure SemCom solution without secret-key exchange and with negligible inference-time overhead, enabling robust privacy protection in shared-channel scenarios.

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.

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

TL;DR

This work targets secure semantic communication over challenging MIMO fading wiretap channels by introducing a DRL-enabled multi-level jamming scheme that combines semantic-layer jamming (text-encoded) and physical-layer jamming (Gaussian noise). A DDPG agent jointly designs three precoders to balance high PSNR for the legitimate user with degraded performance at the eavesdropper, trained via a five-stage alternating strategy that integrates SemCom and reinforcement learning. Experimental results on CIFAR-10 and related text benchmarks show that the proposed approach achieves security comparable to encryption-based and encoded-jammer baselines while improving the legitimate receiver’s PSNR by up to about 0.6 dB across various SNRs and compression ratios. The method offers a practical secure SemCom solution without secret-key exchange and with negligible inference-time overhead, enabling robust privacy protection in shared-channel scenarios.

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.

Paper Structure

This paper contains 21 sections, 20 equations, 3 figures.

Table of Contents

  1. Introduction
  2. System Model
  3. Proposed Method
  4. Transmitter
  5. Receivers at Bob and Eve
  6. The DDPG Agent
  7. Training Strategy
  8. Stage 1
  9. Stage 2
  10. Stage 3
  11. Stage 4
  12. Stage 5
  13. Performance Evaluation
  14. Experimental Settings
  15. Datasets
  16. ...and 6 more sections

Figures (3)

  • Figure 1: The framework of our proposed secure SemCom system.
  • Figure 2: Comparison of PSNR performance for both users under different channel SNRs using different approaches. The compression ratio is fixed at 1/96, with $N_m$ and $N_n$ both set to 4. We set $\rm SNR_{leg} = \rm SNR_{eve}$.
  • Figure 3: Comparison of PSNR performance for both users under different compression ratios using different approaches. We set $\rm SNR_{leg} = \rm SNR_{eve} = 10dB$, with $N_m$ and $N_n$ both set to 4.