Expertise as a signature: continuous implicit authentication based on behavior

Expertise as a signature: continuous implicit authentication based on behavior

연구 분야: Analysis

학회: Cluster Computing

Today, with the increasing digitalization of processes and transactions, enterprises face new challenges regarding the security of sensitive data. Cyberattacks are becoming more sophisticated, necessitating advanced protection measures to prevent security breaches and mitigate potential risks. In this context, user authentication plays a crucial role in safeguarding confidential information and preventing malicious intrusions. The main objective of this work is to propose a system that can identify genuine experts from impostors who impersonate identities and attempt to infiltrate the system. The proposed system guarantees implicit and continuous secure authentication based on three deep learning algorithms: Multi-Layer Perceptron (MLP), Long Short-Term Memory and Recurrent Neural Networks (RNN). By utilizing behavioral characteristics specific to our experts, we constructed a new dataset comprising 1,652 experts, which includes features such as keystroke dynamics, semantic, syntactic, and lexical structures of business rules, as well as their inconsistencies. To select the most relevant and informative features, we combined several feature extraction methods, including Analysis of Variance (ANOVA), Kruskal-Wallis, and Recursive Feature Elimination. Experimental results demonstrate that our proposed system was able to detect genuine experts from impostors with a precision of 99.2%.