AI-DRIVEN ANOMALY DETECTION FOR DATA LOSS PREVENTION AND SECURITY ASSURANCE IN ELECTRONIC HEALTH RECORDS

Abstract

The study titled AI-Driven Anomaly Detection for Data Loss Prevention and Security Assurance in Electronic Health Records explored how artificial intelligence enhances the protection, monitoring, and assurance mechanisms within modern healthcare information systems. The research followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework to ensure transparency, replicability, and methodological rigor. A total of 96 peer-reviewed studies published between 2014 and 2024 were systematically analysed, encompassing empirical, experimental, and theoretical investigations drawn from major academic databases such as Scopus, IEEE Xplore, PubMed, and ScienceDirect. The analysis revealed that machine-learning and deep-learning algorithms—particularly autoencoders, recurrent neural networks, and ensemble hybrid models—significantly improved the precision and recall of anomaly detection in electronic health records (EHRs) when compared to traditional rule-based and signature-based systems. More than half of the reviewed studies reported detection accuracies exceeding 90%, confirming AI’s ability to identify subtle irregularities in user access, data transmission, and system behaviour. Furthermore, findings demonstrated that AI integration led to measurable improvements in data loss prevention (DLP), with reported reductions in unauthorized data transfers ranging between 40% and 65%. Quantitative linkages were also established between AI-driven detection accuracy and assurance outcomes, including higher compliance audit success rates, shortened incident dwell times, and increased containment efficiency. Contextual and behavioural analytics were identified as critical contributors to model performance, enabling systems to distinguish legitimate clinical variability from potential security threats. However, the review also identified methodological limitations, such as the absence of standardized benchmark datasets, limited real-world validation, and insufficient interpretability in deep learning models, which continue to constrain generalizability and adoption. Overall, the findings underscored that AI-driven anomaly detection offers a robust, adaptive, and evidence-based mechanism for safeguarding patient data, ensuring regulatory compliance, and reinforcing institutional trust in healthcare’s digital infrastructure. By transforming static, rule-based monitoring into dynamic, learning-oriented assurance systems, artificial intelligence demonstrated the potential to redefine how healthcare organizations achieve sustained data security and operational resilience in an increasingly interconnected clinical environment.