Cyber-Physical Fusion-Based Approach for Intelligent Attack Detection in Smart Power Grids

The integration of smart technologies in modern power grids has enhanced operational efficiency but has also introduced significant vulnerabilities to cyber-physical attacks. This paper presents a Cyber-Physical Fusion-Based Intelligent Attack Detection System leveraging advanced machine learning techniques to identify and mitigate potential threats in smart grids. The proposed approach combines multi-source data, including operational parameters such as year, actor information, industry codes, motive, and event-specific attributes, to build a robust predictive model. A dual-layer architecture integrates supervised learning classifiers including Random Forest, LightGBM, CatBoost, with a feature encoding mechanism that adapts to unseen data. The system analyzes real-time parameters like actor involvement, event subtypes, geographic indicators, and anomaly patterns to classify threats into four categories: Normal, Frequent Attack, Medium Severity Attack, and Critical Attack. By employing ensemble learning strategies, the detection accuracy is significantly enhanced, achieving over 94% classification performance across multiple models. Furthermore, the framework includes actionable recommendations tailored to the identified threat level, enabling faster mitigation and proactive risk management. The proposed methodology provides a scalable and adaptable solution for securing smart power grids, reducing downtime, and minimizing the economic impact of cyber-physical disruptions.

Keywords

Smart Power Grid, Cyber-Physical Security, Machine Learning, Intelligent Attack Detection, Ensemble Learning, Graph Neural Network, LightGBM, CatBoost, Random Forest, Predictive Maintenance, Anomaly Detection, Critical Infrastructure Protection.