Novel hybrid firefly algorithm: an application to enhance XGBoost tuning for intrusion detection classification

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Network intrusion detection systems (NIDS) face challenges with high false positive and false negative rates.
• Optimizing machine learning classifiers like XGBoost is crucial for improving NIDS performance.

Purpose of the Study:

• To present a novel, improved firefly algorithm (IFA) for optimizing XGBoost hyper-parameters.
• To address the challenge of high error rates in network intrusion detection.

Main Methods:

• The improved firefly algorithm was validated on CEC2013 benchmark instances against original firefly algorithm and other metaheuristics.
• The IFA was used to tune XGBoost hyper-parameters for network intrusion detection.
• The optimized XGBoost classifier was evaluated on the NSL-KDD and USNW-NB15 datasets.

Main Results:

• The improved firefly algorithm demonstrated superior performance in hyper-parameter optimization tasks.
• The optimized XGBoost classifier achieved improved classification accuracy and average precision on network intrusion detection datasets.
• Experimental results confirm the efficacy of the proposed metaheuristic approach.

Conclusions:

• The proposed improved firefly algorithm shows significant potential for machine learning hyper-parameter optimization.
• This method can enhance the accuracy and precision of network intrusion detection systems.
• The study contributes a robust solution for improving cybersecurity through advanced AI techniques.