On the Applicability of Multi-Characteristics for the Continuous Authentication in IIoT Scenarios

By exploiting characteristics from channel state information (CSI) profiles and the behavioral habits of users during their routine work processes in IIoT systems, this paper proposes a passive multi-characteristics user authentication framework for continuous user authentication. We first characterize user physical layer identities using a well-known eXtreme Gradient Boosting (XGBoost) machine learning algorithm, and then depict user behavioral characteristics by formulating the authentication decision process as a Hidden Markov Model (HMM) to further confirm user identities. Extensive experiments are performed to show the authentication performance of the proposed user authentication framework in terms of ROC curves and accuracy in various IIoT scenarios. We also investigate the performance of the proposed framework for resisting impersonation attacks in the IIoT scenario.