Cyber Times International Journal of Technology & Management

Intrusion Detection Systems (IDS) have become indispensable components of contemporary network security architectures. However, traditional IDS approaches relying on signature-based or single-model anomaly detection mechanisms continue to exhibit significant limitations in terms of high false positive rates, poor generalization to zero-day attacks, and inadequate performance under class-imbalanced conditions. This paper proposes a novel Hybrid Artificial Intelligence based Intrusion Detection System (HAI-IDS) that integrates four complementary computational intelligence paradigms: Genetic Algorithm (GA) for optimal feature selection, Random Forest (RF) for preliminary ensemble classification, Deep Neural Network (DNN) for multi-class attack prediction, and Support Vector Machine (SVM) for anomaly boundary verification. The proposed architecture is evaluated on two benchmark datasets — NSL-KDD and CICIDS2017 — after rigorous preprocessing that includes normalization, feature engineering, and Synthetic Minority Oversampling Technique (SMOTE) for class imbalance correction. Experimental results demonstrate that HAI-IDS achieves an overall accuracy of 98.7%, a detection rate of 98.6%, a precision of 98.2%, a recall of 98.5%, and a false positive rate of only 0.9% on NSL-KDD — outperforming traditional IDS, pure SVM-IDS, standalone Random Forest IDS, and deep learning-only architectures across all evaluation metrics. These results establish HAI-IDS as a competitive and practically deployable solution for real-world network intrusion detection.