DESIGN AND IMPLEMENTATION OF A SYSTEM FOR EARLY DETECTION AND MITIGATION OF RANSOMWARE ATTACKS USING EXPLAINABLE AI

Abstract

The exponential growth of big data and user-generated content has introduced critical security vulnerabilities, as real-time sentiment sensing is frequently obstructed by "noisy" data while digital ecosystems face increasingly sophisticated ransomware. This research proposes a hybrid framework centered on SentiAddaxNet, which integrates BERT and RoBERTa for advanced feature extraction from unstructured data. The architecture employs Bi-LSTM to capture sequential relationships and Swin Transformers for hierarchical context. Optimized by the Hybrid Addax Optimization Algorithm (HAOA) to ensure minimal error rates, the system features Explainable Artificial Intelligence (XAI) to provide transparent, human-interpretable justifications for detection decisions. By monitoring behavioral indicators such as abnormal I/O operations and encryption patterns, the framework mitigates financial risks and enhances strategic decision-making through verified insights, bridging the gap between automated detection and human trust.