Dynamic Trust Boundary Frame Work for Enterprise Data Leakage Prevention

Abstract

With the growing trend of cloud systems, remote work places, and enterprise network connections, enterprise data leakage is an emerging cybersecurity problem. Conventional Data Leakage Prevention (DLP) solutions are typically rigid and based on trust assumptions and pre-established security policies, which fall short when facing the changing landscape of insider threats and unusual user behavior. This study presents a Dynamic Trust Boundary Framework (DTBF) for enterprise data leakage prevention leveraging the CERT Insider Threat Dataset, to overcome these limitations. The proposed architecture combines data pre-processing, behavioral profiling, anomaly detection, computation of dynamic trust, risk estimation and adaptive access control. 25,000 activity records were analyzed during the preprocessing, which resulted in a total of 1520 duplicate records, 1100 missing records and 850 inconsistent records. The most influential security indicators are revealed to be login frequency (0.92), file access activity (0.88) and email behavior (0.81) by feature importance analysis. The anomaly detection module detected high risk events with scores of 95 and 98 respectively, and the dynamic trust scores ranged from 0.61 to 0.83 based on the user behaviour. The experimental results showed that the detection rate of DTBF was 98.5%, the classification accuracy was 98.2%, which is better than DLP, Zero Trust, Random Forest, SVM, and KNN. This research has proven that DTBF is an effective, adaptive, and scalable solution to prevent data leakages proactively in enterprise.