AI-Based Physical Layer Key Generation in Wireless Communications: Current Advances, Open Challenges, and Future Directions

Physical layer secret key generation (PHYSKG) is a promising technology to provide lightweight and information-theoretically secure key sharing. It has been widely studied in wireless communications and the Internet of Things (IoT). Yet, this technology is nontrivial for large scale wireless networks due to high overhead, insufficient security, and poor adaptability. Recently, artificial intelligence (AI)-based PHYSKG has shown greater performance and higher security with lower overhead than traditional schemes. Although many existing works have been devoted to exploiting AI-based PHYSKG, a scientific review is still missing, but highly anticipated in the literature. In this article, we perform a serious review on AI-based PHYSKG in wireless communications. After a comprehensive introduction to the processing procedure of AI-based PHYSKG, we propose a set of criteria for justifying its performance with regard to key generation efficiency, reliability, security, and generality. Based on these criteria, we review and analyze the state-of-art, focusing on randomness extraction and quantification. On the basis of the review, we further indicate open issues, current challenges, and future research direction in this field.