Intrusion Detection with Mouse Movements and Self-Supervised Learning

Abstract

Adding valid safety measures to security layers can be achieved by behavioural biometrics. In this period in which big data solutions are improved and marketable, it can be a logical choice to identify users with big data consisting of their behaviours in addition to other security layers. For this reason, a self-supervised model has been proposed with the Balabit Dataset. This self-supervised model is created with autoencoders and it is demonstrated that the model performance outperforms the previously proposed self-supervised methods. Generally, the model performance was evaluated under the Area Under Curve and Equal Error Rate (EER) evaluations. Comprehensive experiments show that our model’s performance is comparable with the models based on supervised methods. Keywords—Balabit Dataset, intrusion detection, mouse dynamics, self-supervised learning