Physical Layer Security for AI-Generated Content: Power and Elements Allocation for Active RIS
Junhao Duan, Ying Zhang, Jinyuan Gu, Lei Zhang, Wei Duan
Abstract
Artificial intelligence generated content (AIGC) as an AI technology with generative and creative capabilities, by means of data training and algorithm generating, it achieves independent generation and creation of various forms of content and data. However, on one hand, the massive information and data upload/download required for AIGC will pose a huge challenge to the limited spectrum efficiency. On the other hand, data security is a primary issue for future developments and applications. To address such issues, in this work, we investigate a novel active reconfigurable intelligent surface (RIS)-assisted physical layer security (PLS) for AIGC system, to simultaneously improve the spectrum efficiency and secrecy performance. In the proposed scheme, the AIGC can be amplified and reflected with a higher channel gain by integrating an additional active reflected power amplifier circuit on the RIS element to overcome the multiplicative fading effect due to the passive RIS reflected link. To maximize the secrecy capacity, an efficient alternative optimization algorithm is designed to obtain the optimal solution for RIS element allocation and power allocation, which is based on an equivalent expression for the secrecy capacity derived from the law of large numbers. Simulation results demonstrate that the proposed active RIS scheme can effectively overcome the multiplicative fading effect provided by the passive RIS scheme, and the secrecy capacity outperforms that of the random allocation and existing RIS-assisted schemes.