Behavior-Driven Real-Time Risk Assessment for Secure Fusion of Social IoT and Digital Twins

The rapid growth of the Social Internet of Things (SIoT) and Digital Twin (DT) technologies in smart homes, healthcare, and industrial automation has created complex networks that demand robust security measures. Integrating these technologies into Social Internet of Digital Twin Things (SIDTT) ecosystems faces security challenges, primarily due to an open attack surface and inadequate real-time risk assessment frameworks. Current solutions often lack comprehensive integration of contextual information and user behavior analytics, leading to potential privacy issues and inefficient security measures. To address these gaps, this research proposes a novel Behavior-Driven Real-Time Risk Assessment framework. It introduces a Context-Aware Federated Access Control and Authorization Framework (CA-FA-CA) that leverages Federated Learning with Long Short-Term Memory (LSTM) networks for user behavior analysis and Isolation Forest for anomaly detection. Additionally, an adaptive risk-based access control that comes under AI (Artificial Intelligence) driven Cybersecurity known as Real-Time Deep Reinforced Risk Score Assessment Framework, which combines Reinforced Generative Adversarial Networks (RRS-GAN) with Deep Q-Network (DQN) and Adaptive Multi-Factor Authentication (AMF), is introduced for dynamically adjusting risk scoring and authentication measures. This comprehensive approach ensures enhanced security and user experience by continuously adapting to real-time data and evolving threats, giving an Accuracy of 98.2% and a throughput of about 1023 TPS, while overcoming the limitations of existing methods in SIDTT ecosystems.