Deep Learning Framework for DDoS Intrusion Detection in IoMT Networks: Combining CNN, GRU, and CatBoost Classifier

In recent years, there are numerous malicious attack on communication and commercial services. The Intrusion Detection System (IDS) predict the anamolies and threats happens on the network with high accuracy. The conventional IDS faces more challenges in feature extraction and failed to address the spatial and temporal dependencies which inherent in medical data. In this paper a hybrid IDS framework is developed where integrated with Convolution Neural Networks (CNN), Gated Recurrent Units (GRU) for feature extraction with CatBoost for classification. The KDDCup 1999 dataset is utilized for simulates the intrusion and normal network behavior in medical network environment. The CNN capture the spatial correlation and GRU process the sequential layer which analysis the dynamic and multi-dimensional intrusion patterns. These features are fed into the CatBoost Classifier and predict the categories into normal or malicious network behavior. The performance metrics of this proposed approach is evaluated and find the accuracy, precision, recall, and F1-Score (99%) which is higher than existing approach. Therefore, this approach will be scalable and reliable solution for enhancing the network security and avoid cyber threats.