African buffalo optimization with deep learning-based intrusion detection in cyber-physical systems.

Area of Science:

• Cyber-physical Systems Security
• Machine Learning Applications

Background:

• Cyber-physical systems (CPS) face increasing security challenges due to complex interconnections.
• Detecting intrusions in CPS is difficult, often limited by inadequate feature selection in machine learning models.
• Existing intrusion detection systems (IDS) struggle with imbalanced datasets where intrusions are rare.

Purpose of the Study:

• To propose an advanced African Buffalo Optimizer Algorithm with Deep Learning Intrusion Detection (ABOADL-IDS) model for robust CPS security.
• To enhance intrusion detection accuracy by optimizing feature selection and deep learning model hyperparameters.
• To address the challenge of identifying subtle cyber-physical attacks within complex network environments.

Main Methods:

• The ABOADL-IDS model employs data normalization followed by African Buffalo Optimizer (ABO) for effective feature selection.
• A stacked deep belief network (SDBN) is utilized for the core intrusion detection and identification process.
• Seagull Optimization (SGO) is implemented to fine-tune the hyperparameters of the SDBN model for improved performance.

Main Results:

• The ABOADL-IDS model demonstrated a high accuracy of 99.28% in detecting intrusions.
• Performance was validated using benchmark datasets NSLKDD2015 and CICIDS2015.
• The proposed model significantly outperformed existing methods in various performance metrics.

Conclusions:

• The ABOADL-IDS model offers a highly effective solution for intrusion detection in cyber-physical systems.
• Optimized feature selection and deep learning, enhanced by metaheuristic algorithms, are crucial for advanced CPS security.
• The research highlights the potential of combining novel optimization algorithms with deep learning for cybersecurity.