Trustworthy interaction model: continuous authentication using time–frequency joint analysis of mouse biometrics

The rapid development of information technology attracts more attention to information security. Static authentication (SA) technology is widely used in account protection. However, once this shield is broken, it will lead to a serious information security crisis. Therefore, we propose a trustworthy interaction model that continuously authenticates users’ human–computer interaction behaviour, which is a supplement to SA. Specifically, the Hilbert–Huang transform is used to extract time-frequency domain features of user mouse behaviour. Then, the users’ unique mouse behaviour patterns are modelled by LSBT to quantify the deviation between current mouse behaviour and true patterns. Finally, the dynamic trust model is deployed to continuously monitor the current user’s identity credibility score in human-computer interaction. Notably, a 30+ month dataset of the mouse behaviours of 32 participants is collected from a real website to prove the effectiveness of TIM. Two real-world scenarios, comprising 1,344 attacks, were simulated to evaluate the trustworthy interaction model performance. All 1,344 attacks were successfully detected, with an average time of 1.63 min to lock imposters out. The TIM is easy to deploy to continuously authenticate users and can accurately and quickly detect imposters.