Model-Hopping Semantic Communication System for a Reliable and Secure Transmission

As an effective mechanism for safeguarding information against eavesdropping in semantic communication systems, existing semantic encryption methods face limitations: vulnerability to sophisticated attacks from active eavesdroppers and compromised recovery performance of legitimate receivers in multipath channels with strong effects. To address these challenges, we propose the Model-Hopping Semantic Communication System (MHSCS) by drawing inspiration from frequency hopping technique. Specifically, model-hopping coding is designed to implement an enhanced method of semantic coding by adaptive model selection, boosting transmission reliability over multipath channels. Considering that semantic coding is a natural protection layer, a double-layer semantic encryption method is proposed, which can improve the encryption effects in severe information leakage scenarios. The experimental results show MHSCS increases legitimate receivers’ image recovery performance by up to 18.60% under strong multipath effects compared to baselines’ optimal performance and prevents up to 81.87% of privacy leakage in severe leakage scenarios.