Data-centric context-aware intrusion detection and mitigation framework for cloud security

The rapid growth of cloud computing and data centers has significantly improved capacity and resource management, but has also led to more security issues because of increased network traffic. Existing IDS struggle to accurately detect both known and evolving cyber threats in dynamic cloud data center environments. The research presents a hybrid intrusion detection and mitigation framework designed for cloud-based data centers. The framework utilizes the LITNET-20 dataset contains real-world network flow data, to evaluate its performance. This framework contains two complementary detection phases namely, hybrid supervised learning model, which combines categorical feature representation and attention-based sequence modeling, enabling high detection accuracy with reduced computational latency. The second phase uses hybrid unsupervised detection methods, combining reconstruction-based anomaly detection with temporal pattern learning to identify unknown and developing cyber threats. The system incorporates a context-aware mitigation layer for automatically connecting detected threats to their corresponding response strategies when network damage occurs. The system uses an adaptive mechanism for reducing intrusions while ensuring stable cloud data center performance. The proposed cloud-based intrusion detection and mitigation system can detect known attacks with high accuracy while efficiently detecting unknown attacks with a low number of false alarms for the cloud data centers.