Optimizing Feature Selection in Intrusion Detection Systems Using a Genetic Algorithm with Stochastic Universal Sampling

• Published in: International journal of advanced computer science & applications Vol. 16; no. 1
• Main Authors: Akula, RadhaRani, Kumar, GS Naveen
• Format: Journal Article
• Language: English
• Published: West Yorkshire Science and Information (SAI) Organization Limited 2025
• Subjects: Accuracy; Anomalies; Computer science; Datasets; Denial of service attacks; Efficiency; Feature selection; Genetic algorithms; Intrusion detection systems; Literature reviews; Machine learning; Malware; Multiple objective analysis; Optimization; Performance evaluation; Regression analysis; Sampling
• ISSN: 2158-107X; 2156-5570; 2156-5570
• DOI: 10.14569/IJACSA.2025.01601102

The current study presents a hybrid framework integrating the Genetic optimization algorithm with Stochastic Universal Sampling (GA-SUS) for feature selection and Deep Q-Networks (DQN) for fine-tuning an ensemble of classifiers to enhance network intrusion detection. The proposed method enhances genetic algorithms with stochastic universal sampling (GA-SUS) combined with recursive feature elimination (RFE). An ensemble of machine learning methods which includes gradient boosting and XG boost as base learners and subsequently logistic regression as meta learner is developed. A deep Q-networks (DQN) is used to optimize the base algorithms XG boost and gradient boost. The suggested method attains an accuracy of 97.60% on the popular NSL-KDD dataset and proficiently detects several attack types, such as probe attacks and Denial of Service (DoS), while tackling the issue of class imbalance. The multi-objective optimization approach is evident in anomaly detection and enhances model generalization by diminishing susceptibility to fluctuations in training data. Nonetheless, the model's efficacy regarding infrequent attack types, such as User to Root (U2R), remains inadequate due to their sparse representation in the dataset.