Techniques Demystifying Ai Network Forensics in Cyberspace Comprised of Diffusion Models and Hybrid Anomaly Detection Based on Transformers
Ponuswamy Suresh, V Sheeja Kumari, Aswathy RH · 2025
The progressive tendencies and patterns behind cybercrimes hinder network forensic methods. The volume and complexity of perpetrated network attacks is way beyond the global average. To solve these issues, this thesis proposes a framework for AI enhanced network forensic investigations using diffusion models with a hybrid anomaly detection using transformers. Because models can capture complicated patterns, diffusion models are used to generate synthetic network traffic patterns that aid in the identification of subtle anomalies that other conventional models would not detect. Transformed networks having attention components and the ability to process sequential information further increase precision on the still questionable area of real-time anomaly detection. Further practical validation of the proposed framework proves its effectiveness in a range of other domains. Comparison with state-of-the-art techniques, novel approaches, and most importantly, known and complex cyber threats are covered. With the adaptation of transformer structures, the broad scope of vector attack emergence can be controlled. The accuracy and reliability of cyber forensic investigations are greatly improved as well. This new innovation in AI application for network forensics makes it powerful for cybercrime investigation and mitigation