SecDiff: Diffusion-Aided Secure Deep Joint Source-Channel Coding Against Adversarial Attacks

TL;DR

SecDiff tackles the vulnerability of deep JSCC to active wireless attacks by integrating a diffusion-based decoder with pseudoinverse-guided sampling, subcarrier masking, and an EM-driven blind channel refinement. The method casts jammed-subcarrier recovery as masked inpainting and treats pilot spoofing as a blind inverse problem solved via alternating signal and operator updates. Empirical results on OFDM channels with adversarial jamming and spoofing show SecDiff achieves superior reconstruction quality and perceptual metrics at lower latency than diffusion baselines. The work demonstrates a practical, attack-resilient approach to secure semantic communication with potential real-time deployment benefits.

Abstract

Deep joint source-channel coding (JSCC) has emerged as a promising paradigm for semantic communication, delivering significant performance gains over conventional separate coding schemes. However, existing JSCC frameworks remain vulnerable to physical-layer adversarial threats, such as pilot spoofing and subcarrier jamming, compromising semantic fidelity. In this paper, we propose SecDiff, a plug-and-play, diffusion-aided decoding framework that significantly enhances the security and robustness of deep JSCC under adversarial wireless environments. Different from prior diffusion-guided JSCC methods that suffer from high inference latency, SecDiff employs pseudoinverse-guided sampling and adaptive guidance weighting, enabling flexible step-size control and efficient semantic reconstruction. To counter jamming attacks, we introduce a power-based subcarrier masking strategy and recast recovery as a masked inpainting problem, solved via diffusion guidance. For pilot spoofing, we formulate channel estimation as a blind inverse problem and develop an expectation-minimization (EM)-driven reconstruction algorithm, guided jointly by reconstruction loss and a channel operator. Notably, our method alternates between pilot recovery and channel estimation, enabling joint refinement of both variables throughout the diffusion process. Extensive experiments over orthogonal frequency-division multiplexing (OFDM) channels under adversarial conditions show that SecDiff outperforms existing secure and generative JSCC baselines by achieving a favorable trade-off between reconstruction quality and computational cost. This balance makes SecDiff a promising step toward practical, low-latency, and attack-resilient semantic communications.

Figures (6)

• Figure 1: An illustration of SecDiff for secure JSCC over adversarial OFDM channels. Part A: images are transmitted over an OFDM channel using a deep JSCC, but suffer from quality degradation under jamming or pilot spoofing attacks. Part B the proposed SecDiff workflow, where a diffusion-based decoder, guided by pseudoinverse sampling.
• Figure 2: An overview of SecDiff for secure JSCC over adversarial OFDM channels. Part A: adversarial OFDM channels suffer from pilot spoofing and subcarrier jamming. Part B SecDiff detects jammed subcarriers, restores them via inpainting, and performs hybrid diffusion decoding. Part C SecDiff adopts an EM-style refinement of the channel operator for robust signal reconstruction.
• Figure 3: Quantitative evaluation of image quality under different J/R levels with fixed attack ratio $\rho=0.2$ and $\text{SNR} = 10~\mathrm{dB}$, using five IQA metrics.
• Figure 4: Qualitative comparison of reconstructed images under adversarial jamming with $\rho=0.2$ and $\text{J/R}=8~\mathrm{dB}$.
• Figure 5: Quantitative evaluation of image quality under different $\text{SNR}$ levels with pilot spoofing attacks, using six metrics.