Homomorphic Encryption-Based Joint Source-Channel Coding for Semantic Communications
Yifan Yuan, Bizhu Wang, Rui Meng, Shujun Han, Mengying Sun, Xiaodong Xu · 2025
Semantic Communication, a key technology for 6 G, prioritizes the transmission of semantic-level over bit-level precision, enabling efficient and reliable communication, especially in channel-capacity-limited scenarios like satellite communication. However, eavesdropping remains a significant concern, particularly when transmitting sensitive data wirelessly. To the best of our knowledge, almost all existing techniques for securing semantic information transmission perform semantic encoding before encryption, which deviates from traditional communication protocols. To address this, we propose a Homomorphic Encryption-based Joint Source-Channel Coding (HE-JSCC) scheme. HE-JSCC utilizes the Cheon-Kim-Kim-Song (CKKS) Fully Homomorphic Encryption (FHE) scheme to enable direct semantic encoding and decoding on ciphertext. Additionally, we refine the original DeepJSCC model to enhance compatibility with FHE by substituting certain layers with operations restricted to addition and multiplication, ensuring compliance with fully homomorphic computations. Simulation results show that HEJSCC achieves comparable semantic similarity to unencrypted transmission while also evaluating the additional transmission delay and memory usage introduced by the HE-JSCC, highlighting its limitations and providing directions for future improvements.