Fortified IoT - Novel Intrusion Detection System Using Advanced Machine Learning and Zero Trust Architecture
Open Access DepositedDownloadable Content
secure device registration, real-time intrusion detection, and continuous behavioral monitoring. A novel method of device registration by employing Attribute-Based Access Control (ABAC) in the context of ZTA is proposed which is designed based on real dataset of IoT network. This mechanism dynamically evaluates devices based on network attributes (e.g., port numbers, packet size etc.) to guarantee that access to the network is restricted to verified and trusted devices, thus reducing spoofing and unauthorized registrations. A hybrid deep leaning approach using a combination of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks is employed for intrusion detection to classify benign and malicious traffic. The model is trained and tested on the real-world dataset Canadian Institute for Cybersecurity Internet of Things 2022 (CICIoT2022) which shows accurate and more precise learning of seven types of traffic classes. In addition to the deep learning model, a lightweight Light Gradient Boosting Machine (LightGBM) algorithm is also used to achieve faster inference with least resource overhead, thus, making the system to be adoptable in low-power IoT devices. ZTA concepts are further expanded to enable continuous device behavior monitoring. A policy engine examines attribute modifications and recognizes zero-day threats based on discovering an anomaly in real-time. The proposed models are evaluated using K-fold cross-validation, along with performance indicators including accuracy, precision, recall, and F1-score. Moreover, this paper provides empirical measurements on enhancements on detection accuracy, processing performance, and adaptive access control. Empirically, the CNN-LSTM model achieved an accuracy of 0.961, with corresponding precision, recall, and F1-score values of approximately 0.95 on the CICIoT2022 dataset. The inference time is achieved as 0.227 ms. The LightGBM model attained an accuracy of 98.39% but with far lower inference time (0.016 ms), suitable for real-time detection in low-powered condition. Furthermore, the paired t-test and Wilcoxon Signed-Rank Test for both models show superiority in comparison to existing state-of-the-art models presented in the literature. Additionally, the ZTA-based device registration successfully rejected 100% of unauthorized devices in the synthesized dataset, demonstrating the resilience of the ABAC-driven policy engine. These results highlight the effectiveness of the approach, as evidenced by achieving the tradeoff between detection accuracy, network overhead and proactive access control and provide solid basis for scalable and intelligent IoT security solutions.
Driven by the rapid expansion of the Internet of Things (IoT), modern digital infrastructures have adapted to enable connectivity of numerous devices in industrial, healthcare, residential, and commercial domains. This has, however, resulted in the emergence of intricate security problems as IoT networks are characterized by the decentralized nature, limited-resource environments and heterogeneous structures. In this study, a behavior-oriented, lightweight, and complete security framework is described that uses machine learning (ML) and Zero Trust Architecture (ZTA) to boost IoT security and trustworthiness. The proposed framework addresses three core aspects of IoT security
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.