NEAT-ID: A Novel Method for Enhancing Threat Detection Process DDoS in Cybersecurity

Abstract

Cyberattacks, especially Distributed Denial-of-Service (DDoS) attacks, are highly dangerous to online infrastructure, as they use network resources and cause disruption of services. It is also hard to detect such attacks in real-time because the traditional rule-based intrusion detection system (IDS) and single machine-based learning models fail to contend with threat variations. In this paper, NEAT-ID (Neuro- Symbolic Ensemble of Anomaly-based Threat Detection) is described, which is a hybrid framework that combines both network and biometric signals to enhance the accuracy and interpretability of the detection. NEAT-ID is based on a wavelet-transformed feature extractor of temporal network patterns, a Transformer encoder with attention on biometric feature integration, a rulefit model of symbolic reasoning, a stacked ensemble of five classifiers (TabNet, LightGBM, Histogram-based GB, Naive Bayes, Logistic Regression), and an XGBoost meta-learner to provide the final prediction. The framework was tested on the CIC-dDoS2019 dataset, with NEAT-ID scoring 96% accuracy, 97% F1-score, and 0.9949 ROC-AUC, which is better than baseline IDS models and shows robust, interpretable, and high- performance intrusion detection.