Recurrent Highway Network-Based Model for Mouse Dynamic Authentication

Identity authentication serves as an effective means to protect systems from intrusion. Among various methods, mouse dynamic behavior authentication stands out for its simplicity, efficiency, and ease of deployment as it requires no additional verification devices. Existing mouse dynamic behavior authentication methods typically either construct feature engineering based on machine learning algorithms or directly input raw data into deep learning algorithms. However, these two construction strategies exhibit significant performance variations across different users. Therefore, this paper proposes a mouse dynamic behavior authentication model based on Recurrent Highway Network (RHN). Within a given feature space, this model achieves mouse dynamic behavior authentication based on deep learning. Experimental comparisons on publicly available datasets demonstrate that the proposed model outperforms other benchmarks in static identity authentication tests across different users, effectively realizing user identity authentication through mouse dynamic behavior.