Enhancing Intrusion Detection in Wireless Sensor Networks Using a GSWO-CatBoost Approach.

Area of Science:

• Computer Science
• Network Security
• Artificial Intelligence

Background:

• Intrusion detection systems (IDSs) are crucial for wireless sensor network (WSN) security.
• Effective feature selection (FS) is vital for IDS performance but often faces limitations.
• Conventional methods can struggle with premature convergence and balancing exploration/exploitation.

Purpose of the Study:

• To introduce a novel feature selection and classification approach for WSN intrusion detection.
• To enhance the efficacy and efficiency of IDSs using advanced optimization and machine learning.
• To overcome the limitations of existing methods, particularly premature convergence in optimization algorithms.

Main Methods:

• Developed the Genetic Sacrificial Whale Optimization (GSWO) algorithm, combining genetic algorithms (GA) and whale optimization algorithms (WOA).
• Incorporated a three-population division strategy and conditional inherited choice (CIC) to improve WOA's global search and avoid premature convergence.
• Utilized the CatBoost model for classification, enhanced with GSWO for hyperparameter fine-tuning via quantization.

Main Results:

• The GSWO-CatBoost approach demonstrated superior performance across multiple benchmark datasets (WSN-DS, WSNBFSF, NSL-KDD, CICIDS2017).
• Achieved high accuracy rates: 99.65% (WSN-DS), 99.99% (WSNBFSF), 99.76% (NSL-KDD), and 99.74% (CICIDS2017).
• Exhibited significantly faster inference times, nearly 100x quicker than deep learning methods, indicating real-time applicability.

Conclusions:

• The proposed GSWO-CatBoost method offers a highly accurate and efficient solution for intrusion detection in WSNs.
• The novel GSWO algorithm effectively balances exploration and exploitation, enhancing global search capabilities.
• The approach proves versatile and accurate across diverse datasets, highlighting its practical value for network security.