Enhancing cybersecurity via attribute reduction with deep learning model for false data injection attack recognition
Abstract Cyberattacks have given rise to several phenomena and have raised concerns among users and power system operators. When they are built to bypass state estimation bad data recognition methods executed in the conventional grid system control room, False Data Injection Attacks (FDIA) pose a si...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-01-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-024-82566-6 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Cyberattacks have given rise to several phenomena and have raised concerns among users and power system operators. When they are built to bypass state estimation bad data recognition methods executed in the conventional grid system control room, False Data Injection Attacks (FDIA) pose a significant security threat to the operation of power systems. Therefore, real-time monitoring becomes inevitable with the quick implementation of renewables within the grid operator. The state estimation algorithm plays a major role in defining the grid’s operating scenarios. FDIA creates a significant risk to these estimation strategies by adding malicious information to the measurement obtained. Real-time recognition of these attack classes improves grid resiliency and ensures a secure grid operation. This study introduces a novel Attribute Reduction with a Deep Learning-based False Data Injection Attack Recognition (ARDL-FDIAR) technique. The primary goal of the ARDL-FDIAR technique is to improve security via the FDIA detection process. The ARDL-FDIAR technique uses Z-score normalization to scale the input data. The attribute reduction process gets invoked using the modified Lemrus optimization algorithm (MLOA) to choose optimal feature sets. Moreover, the FDIA detection process is performed by modelling an improved deep belief network (IDBN) model. Furthermore, the performance of the IDBN model is improved by the Cetacean Optimization Algorithm (COA)-based hyperparameter tuning process. A series of experiments were performed to ensure the enhancement of the ARDL-FDIAR technique. The results indicated the enhanced security performance of the ARDL-FDIAR technique compared to other DL approaches. |
|---|---|
| ISSN: | 2045-2322 |