Hybrid GenAI and Deep Learning Model for Digital Twin–Based Multimedia Server Security in CIoT Systems

CIoT multimedia servers are vulnerable to dynamic attacks, such as flooding and protocol abuse, due to high-volume streaming traffic and strict real-time requirements. To address this issue, the proposed framework integrates a digital twin, a deep learning detector, and a lightweight GenAI controller for multimedia server protection. The framework uses metaheuristic-based feature selection and hyperparameter optimization to improve attack detection performance, achieving an accuracy of 0.997199 and an F1-score of 0.9972, while the GenAI module generates role-aware security responses.