IRS-Enhanced Cross-Layer Semantic Security

Task-oriented SemCom can alleviate network congestion and enhance communication efficiency, which is a promising technology for the sixth-generation (6G) communications. Several works demonstrate the efficiency of SemCom from mathematical theory. As one of emantic communication paradigms, artificial intelligence (AI)-driven SemCom can characterize the task-oriented information over semantic symbols by leveraging the formidable knowledge extraction capabilities of machine learning (ML), thus significantly reducing the irrelevant information and improving the transmission efficiency. However, the open nature of wireless channel poses the challenge of semantic eavesdropping, leading to serious task exposure and privacy leakage. Different from the traditional secure communication that relies on bit-level security at the physical layer, task-oriented secure SemCom should be concerned with task-level privacy preservation. Hence, it is necessary for SemCom networks to rethink secrecy protection from the task-centered semantic perspective.