Continuous behavioral authentication using mouse dynamics based on Artificial Intelligence

Abstract

[Abstract]: Traditional authentication mechanisms are increasingly vulnerable to advanced attacks, highlighting the need for dynamic and continuous user verification. This work presents a continuous authentication platform based on mouse dynamics and artificial intelligence models. The proposed infrastructure integrates real-time event collection, feature extraction, and behavioral modeling within a scalable, containerized architecture. Experiments were conducted with real users in uncontrolled environments, evaluating CatBoost, Random Forest, Neural Networks, Decision Trees, and SVMs under different training volumes and optimized hyperparameters. Results show that CatBoost and SVM achieve the highest accuracy, precision, and recall, with performance improvements up to 20% of the data volume before reaching saturation. The findings confirm the feasibility of mouse-dynamics-based behavioral biometrics as a reliable complement to traditional authentication methods, advancing continuous verification research and opening new directions or multimodal and large-scale deployments.