Creating a cyber security strategy for active distribution systems is challenging due to the integration of distributed renewable energy source. This essay presents a methodology for adaptive hierarchical cyberattack localisation and detection for distributed active distribution systems utilising electrical waveform analysis. The foundation for cyber attack detection is a sequential deep learning model, which enables the detection of even the tiniest cyberattacks. The two-stage approach first estimates the cyber-attack sub-region before localising the specified cyber-attack within it. For the "coarse" localization of hierarchical cyber-attacks, we propose a modified spectral clustering-based method of network partitioning. Second, it is recommended to use a normalised impact score based on waveform statistical metrics to further pinpoint the location of a cyber attack by defining various waveform features.Finally, a detailed quantitative evaluation using two case studies shows that the proposed framework produces good estimation results when compared to established and cutting-edge approaches.

Keywords : SVM, Random Forest, Gradient Boosting, Logistic Regression, Cyber Attack Detection.