NITRO-IDS: Network IoT Traffic Robust Optimized Intrusion Detection System Using Deep Learning
Year: 2026
Type: conference-paper
Abstract: In this paper, we propose NITRO-IDS (Network IoT Traffic Robust Optimized Intrusion Detection System), a deep learning–based framework designed to address the security challenges of modern Internet of Things (IoT) environments. The rapid proliferation of IoT devices has significantly expanded the attack surface of network infrastructures, exposing resource-constrained systems to several cyber threats. Conventional intrusion detection systems often struggle to maintain reliable performance due to the highly imbalanced and noisy nature of IoT traffic datasets. To overcome these limitations, NITRO-IDS integrates optimized deep learning and advanced data balancing techniques to enable robust multi-class traffic classification in complex IoT scenarios. The proposed system integrates strategic feature engineering with a Convolutional Neural Network (CNN) to efficiently capture temporal and statistical patterns in network traffic flows. The model performs six-class traffic classification, distinguishing Normal traffic, Benign anomalous activity, and four major attack categories: Theft, Reconnaissance, DoS, and DDoS. To mitigate class imbalance, two data augmentation strategies are investigated: Synthetic Minority Over-sampling Technique (SMOTE) and Conditional Tabular Generative Adversarial Networks (CTGAN). Unlike traditional oversampling methods, CTGAN learns the underlying distribution of minority classes, generating more realistic synthetic traffic samples. Experimental results demonstrate that the CTGAN-enhanced NITRO-IDS significantly outperforms the SMOTE-based configuration. The proposed framework achieves 98.5% accuracy, compared to 93.67% accuracy obtained using SMOTE. These findings confirm that adversarial synthetic data generation substantially improves detection robustness and generalization capability in highly imbalanced IoT traffic scenarios. The proposed system provides a scalable and effective solution for next-generation IoT security architectures.
Institution: Canadian International College
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7167059989