AI-Based Identity Verification Through Behavioral Biometrics: Enhancing Remote Work Security

With the increased adaptability of remote work patterns, secure and efficient identity verification has also become a paramount concern. The traditional model of authentication involving the use of passwords or security tokens fails to provide any continuous verification of the user and is also prone to hacking attempts. Behavioral biometrics, especially keystroke dynamics and mouse movement patterns, offer an effective alternative as they allow for unobtrusive user authentication that is based on the individual user's behavior and is therefore continuous. This paper explores the application of Long Short-Term Memory (LSTM) networks, a sequence-based AI model, for studying and differentiating behavioral biometrics. We are using freely available data sets of keystroke and mouse dynamics to design and test an LSTM based system which is capable of making a distinction between users and imposters. We have shown that LSTM networks are significantly better in handling time series data such as state transition sequences than other statistical machinery traditional machine learning methods, such as Random Forests and Support Vector Machines, where they scored 89% accuracy between normal user operations and abuse activity. This aspect of the research addresses the reason why LSTM networks are appropriate for live remote identity verification systems, which is their ability to learn long strips of sequences and the behavioral flow. The originality of the study is to expand secure authentication solutions based on artificial intelligence systems that can be integrated to facilitate remote working in any sector for enhanced security.