Enhancing Detection of Low-Rate DoS Attacks Using CatBoost-RFE and Multilayer Perceptron

Low-rate Denial of Service (LDoS) attacks represent a significant challenge in network security due to their ability to mimic legitimate traffic patterns, making detection complex and posing threats to network integrity and availability.Unlike traditional Denial of Service attacks that are aggressive, and continuous, LDoS attacks use low-rate, intermittent strategies, increasing their chances of evading conventional security measures.This study proposes an innovative detection model integrating CatBoost-Recursive Feature Elimination (RFE) and Multilayer Perceptron.CatBoost-RFE, a feature selection technique, identifies crucial features that distinguish normal traffic from LDoS attack patterns, reducing noise and enhancing classification accuracy.Multilayer Perceptron, a deep learning classifier, then processes the selected features to accurately differentiate between normal and malicious traffic.The experimental results demonstrate that the model achieves 99.5% accuracy with a False Positive Rate (FPR) of just 0.0003, significantly outperforming existing detection methods.This evaluation was conducted using the CICIDS2017 dataset, a widely recognized benchmark for intrusion detection research.The dataset contains a diverse set of real-world traffic patterns, including both normal and LDoS attack traffic, ensuring a robust and reliable assessment of the proposed model.This low FPR is critical for minimizing the misclassification of legitimate traffic, thereby maintaining network performance.The findings provide a practical solution for improving LDoS attack detection and contribute to stronger network defense mechanisms.