Retracted: AI-Driven Threat Intelligence Framework for Real-Time Cybersecurity using Federated Deep Learning and Cloud Orchestration

As cyber risks targeting distributed infrastructures are expanding rapidly, centralized detection methods are becoming problematic on all fronts: scalability, privacy, and latency. The paper introduces a new artificial intelligence-based threat intelligence framework that incorporates federated deep learning and cloud orchestration into the framework to achieve real-time, privacy-preserving cybersecurity. The structure makes use of a hybrid CNN-BiLSTM network plus edge-node attention controls accurate recognition of temporal-spatial attack patterns on small data volumes without releasing sensitive data. FedProx algorithm is used to synchronize the model updates to facilitate heterogeneous data distributions. Cloud orchestration with Kubernetes provides a highly available and elastic distribution of detected nodes, in a fault tolerant manner. On benchmark intrusion detection data sets NSL-KDD and CICIDS2017, the system is evaluated to be 97.8 accurate, outperforming existing centralized and federated models in both communication overhead and detection latency. The method shows that real-time cyber threat intelligence can be addressed with a scalable, secure and robust solution with potential protective implications to both enterprises and critical infrastructures.