A Review on Ai based Data Authentication by Monitoring Behavioural Pattern

Abstract

In this study, we do an experiment to examine the viability of a continually authenticating approach based on the monitoring of users' activities to confirm their identities using particular user profiles that are modeled using AI techniques. To carry out the experiment, a unique application was created to collect user data in a supervised situation in which certain tasks must be finished in advance. After anonymization, this dataset will be made accessible to the public. Furthermore, a publicly available dataset was utilized for benchmarking, enabling our methods to be verified in an unguided environment. These data were processed to identify several important properties that might be utilized for training three distinct AI methods: Multi-Layer Perceptrons, Support Vector Machines, and a deep learning network. These methods proved to be successful in both situations and were able to effectively authenticate users. To detect imposters when an authenticated session is hijacked in a real-world setting, a continuous authentication method was designed and tested utilizing weighted sliding windows, and a rejection test was finally carried out.