An Effective Detection of Ransomware Using Improved Long Short-Term Memory

Ransomware detection focuses on recognizing the unusual system activities which indicate an unauthorized data access and restriction. It continuously monitors the system behavior continuously to detect potential threats before significant damage occurs. By identifying anomalies and patterns, it effectively distinguishes ransomware attacks from normal operation. However, Ransomware attacks cause a serious threat because of far-reaching effects and its zero-day variants are even more hazardous. This research proposes Improved Long Short-Term Memory (LSTM) to detect the ransomware effectively and accurate detection of Ransomware. The proposed improved LSTM learns the sequential patterns in system behavior and detects subtle abnormalities over time. The enhanced model improves detection accuracy by learning past data and adapts to new attack variants including zero-day threats. By identifying anomalies, the proposed method enables timely detection and reduces the impact of ransomware attacks. To select the appropriate features, the Dung Beetle Optimization (DBO) is used to optimize the feature subset for better detection performance. The proposed improved LSTM obtains a better accuracy of 95% when compared to existing method like cost sensitive Pareto Ensemble classifier (CSPE-R).