Book

This book presents an intelligent provenance-aware framework for anomaly detection, verification, and adaptive deterrence in social networks. It introduces an integrated approach that combines unsupervised deep learning, correlation-based adaptive ensemble weighting, data provenance verification, and risk-based deterrence. The proposed ensemble integrates four advanced detection models—DOMINANT, GAAE, ARGA, and EfficientGAN—using a correlation-based dynamic weighting strategy to improve the identification of diverse anomalous patterns without requiring labelled data for weight optimization. The book further explores how data provenance can strengthen anomaly detection through the tracking of entities, activities, agents, and relationships. It incorporates SHA-256 cryptographic hashing, provenance completeness, consistency, and graph-contextual features to improve data integrity, traceability, and verification. A provenance-aware scoring mechanism combines anomaly scores with provenance-derived evidence, while an adaptive deterrence engine classifies observations into four risk levels: Normal, Suspicious, High Risk, and Critical/Repeated. These levels are associated with graduated responses, including monitoring, warning, content restriction, and quarantine.