Leveraging osprey optimization algorithm with deep ensemble learning for cybersecurity in CPS environment
A cyber-physical system (CPS) incorporates many interconnected physical processes, networking units, and computing resources, along with monitoring the application and process of the computing system. Interconnection of the cyber and physical world initiates threatening security problems, particular...
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-10-01
|
| Series: | Ain Shams Engineering Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2090447925003533 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A cyber-physical system (CPS) incorporates many interconnected physical processes, networking units, and computing resources, along with monitoring the application and process of the computing system. Interconnection of the cyber and physical world initiates threatening security problems, particularly with the increasing sophistication of transmission networks. Analyzing and detecting cyber-physical attacks in complex CPS systems remains a threat. Researchers have turned to machine learning (ML) for cyber-physical security evaluation. Recent enhancements in deep learning (DL) and artificial intelligence (AI) enable the development of robust intrusion detection systems (IDS) for CPS platforms. A metaheuristic algorithm is employed for feature selection (FS) to mitigate the curse of dimensionality. The study introduces an Osprey Optimization Algorithm with Deep Ensemble Learning for Cybersecurity (OOADEL-CS) in the CPS platform. The proposed OOADEL-CS method identifies and classifies intrusions in the CPS platform. In the OOADEL-CS method, the linear scaling normalization (LSN) approach is utilized for uniform data scaling. For FS, the OOADEL-CS technique employs the OOA to select a subset of features. To detect the intrusions effectually, the OOADEL-CS technique utilizes an ensemble of three models, namely bidirectional long short-term memory (BiLSTM), autoencoder (AE), and multi-layer perceptron (MLP). A modified bacterial foraging optimization algorithm (MBFOA) approach improves the detection classifier rate. The simulation analysis of the OOADEL-CS approach is conducted on a benchmark dataset. The experimental validation of the OOADEL-CS approach portrayed a superior accuracy value of 99.47 % over existing methods. |
|---|---|
| ISSN: | 2090-4479 |