Smart-Grid Cyber-Attack and Disturbance Detection via Interpretable and Accurate Genetic-Fuzzy Data-Mining/ Machine-Learning Approach.
Saved in:
| Title: | Smart-Grid Cyber-Attack and Disturbance Detection via Interpretable and Accurate Genetic-Fuzzy Data-Mining/ Machine-Learning Approach. |
|---|---|
| Authors: | Gorzałczany, Marian B.1 (AUTHOR), Rudziński, Filip1 (AUTHOR) f.rudzinski@tu.kielce.pl |
| Source: | Energies (19961073). Jul2026, Vol. 19 Issue 14, p3388. 19p. |
| Subject Terms: | *Smart power grids, *Internet security, *Data mining, *Multi-objective optimization, *Fuzzy logic, *Feature selection, *Machine learning |
| Abstract: | This paper makes a contribution to smart-grid cybersecurity by proposing an application of our approach, previously published in this journal, to design—from smart-grid data—transparent, interpretable, and accurate systems for detection and classification of cyber-attacks, natural-event-related disturbances, and regular operation of the smart grid. The collection of 15 open-source simulated smart-grid data sets provided by Mississippi State University, USA, in collaboration with Oak Ridge National Laboratories, USA, is used in our experiments. To address this problem, we use our knowledge-based data-mining/machine-learning approach which ultimately generates a set of fuzzy rule-based classifiers—each with a different trade-off between its accuracy and the interpretability of its knowledge base (both are the subjects of maximization in a multi-objective optimization process using evolutionary algorithms). The paper also contributes an extensive cross-validation-based experiment showing that while our approach is not as accurate as alternative and inherently accuracy-oriented "black boxes," it surpasses them in terms of transparency and interpretability of the generated classification decisions, while ensuring acceptable accuracy of these decisions. Therefore, it can act as a second cybersecurity layer that uses transparent rules to help SCADA operators instantly verify and explain alerts triggered by accurate, fast "black-box" detectors. This paper also provides a contribution to a more general and important area of the automatic and effective selection of input variables (features) that are essential in a given decision-making problem (such as smart-grid contingency classification considered in this work). [ABSTRACT FROM AUTHOR] |
| Database: | Energy & Power Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
Be the first to leave a comment!
