Encrypted Semantic Communication Using Adversarial Training for Privacy Preserving

Semantic communication is implemented based on shared background knowledge, but the sharing mechanism risks privacy leakage. In this letter, we propose an encrypted semantic system (ESCS) for preserving, which combines universality and confidentiality. The reflected in fact that structures of all network modules proposed ESCS are public, training database shared, suitable large-scale deployment practical scenarios. Meanwhile, confidentiality achieved through key encryption. Based adversarial training, design encryption scheme to guarantee accuracy both unencrypted modes. Experiment results show with can perform well regardless whether information encrypted. It difficult attacker reconstruct original from eavesdropped message.