An Optimized IDS Framework for Big Data Environments: Integrating Gravitational Search and SMOTE-IPF Data Balancing for High-Accuracy IDS

Abstract In today’s interconnected digital world, network security is essential for protecting all confidential data and ensuring its integrity within various systems. Intrusion Detection Systems (IDS) form a transitional security component of networks as they monitor malicious activities within networks. For many years now, researchers have spent considerable effort on enhancing IDS from exploring machine learning (ML) and deep learning techniques (DL) towards improved attack classification accuracy and reduced false positives. Although much progress has been made, there are a lot of challenges including high-dimensional data, class imbalance, and evolving attack patterns. Thereby, we present novel IDS framework based on Gravitational Search Algorithm (GSA) for feature selection, Synthetic Minority Over-sampling Technique with Iterative-Partitioning Filter (SMOTE-IPF) for data balancing, and ML models for attack classification. We assessed the framework on widely popular cybersecurity datasets NSL-KDD and UNSW-NB15. The experimental results demonstrated superior performance by the Random Forest classifier attaining 99.6% accuracy for the NSL-KDD dataset while attaining 94.4% accuracy for the UNSW-NB15 dataset. These findings also accentuate the effectiveness of the proposed model in enhancing IDS efficiency and lowering false positives; hence it forms a significant contribution to the ongoing research on cybersecurity.