A Trusted Architecture Based on Semantic Communication for Matching Data and Models
Hongchao Jiang, Haotai Liang, Mengran Shi, Chen Dong, Xiaodong Xu · 2023
In this paper, a new layered architecture named Trusted Sub-layer (TSL) is proposed, which features a matching mechanism for protecting the privacy security of data and models effectively based on the different protection levels required. Specifically, TSL is designed to perform three processes to map adapted codec models for specific data transmission tasks, including the Confirmation of Data and Semantic Codec Models, the Trusted Matching (TM) and the Trusted Handshake Protocol for Semantic Transceivers. Moreover, the process of TM is experimentally simulated based on different TM algorithms. And the effects of the number and distribution of privacy features on matching error rate are analyzed. The results show that the Transformer-based TM algorithm achieves the privacy protection effects at 96.5% when the number of privacy features is 6. Finally, the applicability of the algorithm on general nodes of terminal devices is discussed.