MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202641088643 A) filed by Dayananda Sagar University on July 21, 2026, for Digital Twin Based Cyber Attack Prediction Framework.
Inventors include Suresh Kumar Natarajan; Praveen S.; Athul K; Dr. Sivasankaran Harish; Prof. Vivek Gupta; and Prof Ajith Kumar B.
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: Digital Twin Based Cyber Attack Prediction Framework ABSTRACT The increasing complexity of cyber-physical systems and Industry 4.0 infrastructures has made conventional cybersecurity mechanisms insufficient for predicting sophisticated cyber attacks before they occur. This paper proposes a Digital Twin Based Cyber Attack Prediction Framework that creates a real-time virtual replica of network infrastructure to continuously monitor system behavior, detect emerging threats, and forecast potential attack scenarios. Initially, network traffic, system logs, endpoint activities, and device telemetry are collected through distributed monitoring agents. The collected data undergo preprocessing using Isolation Forest for anomaly filtering, feature normalization, and missing value handling. Subsequently, Temporal Graph Neural Networks (TGNN) model dynamic relationships among network entities by capturing both spatial and temporal communication patterns. The extracted graph embeddings are processed using a Transformer-based Temporal Encoder to learn longterm attack sequences and evolving adversarial behaviors. A Deep Reinforcement Learning (DRL) agent based on Proximal Policy Optimization (PPO) continuously interacts with the digital twin environment to simulate multiple attack paths, evaluate security risks, and predict the most probable future attack vectors before they affect the physical infrastructure. To improve interpretability, SHAP (SHapley Additive Explanations) identifies the most influential features contributing to attack prediction, enabling security analysts to understand model decisions. Finally, predicted threats, confidence scores, and recommended mitigation strategies are securely recorded using a permissioned Hyperledger Fabric blockchain, ensuring tamperresistant forensic evidence and trustworthy incident management. The proposed framework significantly improves cyber attack prediction accuracy, reduces false alarm rates, and enables proactive cyber defense through continuous simulation, intelligent threat forecasting, and secure digital twin synchronization.
Disclaimer: Curated by HT Syndication.
