Optimized Deep Learning-Based Intrusion Detection System Using SMOTE and Genetic Algorithms

Abstract

With the increasing sophistication of network attacks, traditional intrusion detection systems (IDSs) face challenges in detecting new attack types, managing high-dimensional network traffic, and responding to adversarial threats. These systems often lack intelligent, efficient solutions. In this paper, we propose an intelligent intrusion detection system that integrates deep learning with optimization algorithms to address these challenges. Specifically, we introduce the synthetic minority over-sampling technique (SMOTE) to handle data imbalance and use genetic algorithms (GA) to optimize the hyperparameters of a deep neural network (DNN). Our results show that this combined approach significantly improves detection accuracy, enhances the detection capability for minority class samples, and increases computational efficiency. Additionally, the system demonstrates robust performance against adversarial attacks and adapts well to complex network environments. This research not only provides a novel method for network security but also lays a practical foundation for developing real-time, efficient IDS solutions.