Secure Communication Guarantees for Diverse Extended-Reality Applications: A Unified Statistical Security Model
Yuquan Xiao,Qinghe Du,Wenchi Cheng,Nan Lu
DOI: https://doi.org/10.1109/JSTSP.2023.3304117
IF: 7.695
2023-01-01
IEEE Journal of Selected Topics in Signal Processing
Abstract:Privacy and security assurance over wireless transmissions is one of critical issues for future extended reality (XR) communication systems expected to be supported by the sixth generation of mobile communications networks (6G). In light of the strong anti-eavesdropping capability, physical layer security (PLS) techniques have been recognized as a competitive candidate to provide secure transmissions for XR services. However, existing key performance evaluation metrics, such as the security capacity and the outage security capacity, cannot well capture diverse features of quality-of-security (QoSec) requirements raised by XR services. To overcome the problem, in this article we propose a unified statistical security model that can characterize fine-grained security requirements for various XR applications. Specifically, the eavesdropping process at the eavesdropper is modeled by a queuing system. The arrival process represents the legitimate user's streaming data correctly-captured by the eavesdropper. The departure process represents data that are outdated and moved out of the queue, fitting the essential time-sensitive nature of XR applications. Yet storing the overheard data in the queue is not equivalent to data recovery, the eavesdropper has to accumulate a sufficient of amount correctly-captured data in the queue to successfully decipher some data each time. Under this framework, leveraging the effective bandwidth theory in statistically queuing analyses, we develop the concept of statistical security capacity, which is used to evaluate the legitimate user's throughput with the constrained information level leaked to the eavesdropper. The statistical security model is featured with a parameter called QoSec exponent, quantitatively indicating the fine-grained level of security requirement. Following this model, we formulate the nonconvex statistical-security-capacity maximization problems with the internal and external eavesdroppers, respectively, associated with the cases with and without eavesdropper's CSI known at the legitimate transmitter. Solving the two problems, we derive the corresponding optimal resource schemes over the time-varying fading channels. Simulation results demonstrate our proposal as an effective model for security requirements, and our scheme can significantly improve security-constrained throughput in XR communications compared to the baseline schemes.