Enhancing Host-Based Intrusion Detection: A NEAT-Optimized Ensemble Learning Approach

It is of utmost importance to security that intrusion detection systems are flexible and instantly respond since Cyber threats are getting so complex. This paper proposes a Hybrid Ensemble Intrusion Detection System that integrates deep learning and NeuroEvolution of Augmenting Topologies (NEAT) to enhance cybersecurity against complex threats. The system combines multiple classifiers—Support Vector Machines, Decision Trees, and Logistic Regression—with a dynamically adapting neural network optimized by NEAT to accurately distinguish normal from malicious activities. By leveraging diverse datasets, such as system logs and user activity patterns, it achieves robust threat detection with minimal computational effort. Pre-processing techniques, including standardization and Principal Component Analysis, ensure computational efficiency and optimal feature selection. This hybrid approach surpasses traditional signature-based and rule-based systems, offering scalability and adaptability to evolving cyber threats. Extensive experimental evaluations demonstrate superior accuracy and reliability, with reduced false positives and higher detection rates, making it a significant advancement over conventional intrusion detection methods and a highly effective, scalable solution for modern cybersecurity challenges.