Behavioral Biometrics Authentication in Critical Infrastructure Using Siamese Neural Networks

Behavioral Biometrics Authentication in Critical Infrastructure Using Siamese Neural Networks

연구 분야: Infrastructure

학회: International Conference on Human-Computer Interaction

Cybersecurity is a crucial issue in today’s critical infrastructure to ensure a secure connection between the administrator and the session. Detecting insiders is a difficult task for cybersecurity professionals, as insiders are hard to detect and identify and thus require advanced techniques to prevent their activities. These users may be current or former employees with access to the organization’s data. A methodology for authenticating users of critical infrastructure systems using deep learning networks is proposed in this paper. Behavioral biometric data or user behavioral characteristics are converted into an image and used in the proposed methodology for authentication. The keystroke data obtained from the login password is transformed into a more acceptable format for deep neural networks. Siamese neural networks can be used for image similarity detection to distinguish a real user from an insider. In the current investigation, numerical keystroke data has been transformed into graphical representations. The transformed data are then subjected to comparative analysis, leading to the determination of similarities between the biometric keystroke profiles. The experiments have shown there is a tendency for the accuracy of a Siamese neural network with a triplet loss function to decrease with increasing margin size. The results obtained are promising, showing that using a deep learning-based approach to analyze images derived from user keystroke data can improve intrusion detection accuracy and perform user authentication more efficiently.