Harnessing Metaheuristics for Superior Intrusion Detection: Deep Learning on Benchmark Datasets

Abstract

While intrusion detection systems (IDS) are crucial to safeguarding network environments from evolving cyber threats, traditional methods often struggle to optimize detection capabilities and balance computational efficiency. To enhance the performance of IDS, this study integrates metaheuristic optimization techniques with deep learning algorithms. Combining various metaheuristics, including Genetic Algorithms, Particle Swarm Optimization, and Harris Hawks Optimization, with deep learning models, such as Graph Neural Networks, Recurrent Neural Networks, and Convolutional Neural Networks, is a systematic approach. Using benchmark datasets CICIDS-2017 and KDD, our research shows that metaheuristic optimization improves model accuracy, reduces false positives, and improves overall robustness. Based on performance metrics, Capsule Networks with metaheuristic optimization achieve superior results, with accuracy rates of 89% and 87%, respectively, on the CICIDS-2017 and KDD datasets. In addition to advancing the effectiveness of intrusion detection, this study provides a robust framework for future explorations and practical applications in the field of cybersecurity.