Real-Time Network Threat Detection in Intelligent Power Monitoring Systems Using Multiobjective Horse Herd Optimization and Online Streaming Random Forest

Abstract

As the threats to critical infrastructure have grown more sophisticated, securing intelligent power monitoring systems have taken on new critical importance. The paper introduces a scenario of network security situation awareness based on Big Data particularly in the intelligent power monitoring setting. The suggested framework incorporates Multiobjective Horse Herd Optimization (MHHO) to allow optimal feature selection, and Online Streaming Random Forest (OSRF) to allow real-time threat detection so that network activities could adapt to the dynamic environment in an online scenario. Depending on the set of KDD Cup99 dataset, the system will capture and analyse security risks based on usage of huge size of heterogeneous data to detect some type of attacks like DoS, Probe, U2R and R2L. MHHO optimizes the detection channel to choose the most informative features and OSRF scales to high-velocity streaming data efficiently; they also guarantee real-time identification of changing patterns of attacks. Experimental analysis shows that the proposed approach is more accurate as 96.2 %, precise as 95.8%, has higher recall as 95.4 % and F1-score 95.6 % and lower error rates (MSE, RMSE) when compared to the previously tested intrusion detection techniques, which proves its effectiveness and flexibility in changing network conditions. The contribution of this work is a scalable, real time, and intelligent security solution able to assist in proactive decision making in power monitoring activities. Future work will concentrate on experimenting with deep learning-based hybrid detection models, graph-based threat correlation, and explainable AI approaches to the improve the level of detection and interpretability further.