Key points are not available for this paper at this time.
The rapid expansion of Internet of Things (IoT) technologies has revolutionized connectivity but has also exposed users to increased cyber threats. This study aims to enhance Intrusion Detection Systems (IDSs) for IoT networks by employing advanced machine learning (ML) techniques and addressing the challenge of dataset selection, moving beyond outdated standards such as KDD99 and NSL-KDD99. We focus on evaluating the effectiveness of the Aegean Wi-Fi Intrusion Dataset (AWID) within the IEEE 802.11 standard for IoT security. Through classification algorithms, we assess AWID's effectiveness by measuring accuracy, detection rate and false positives using the WEKA tool. We introduce a robust ML framework tailored for wireless intrusion detection, conducting evaluations across various scenarios, including nominal and numeric classes and employ feature selection techniques to enhance model performance. Our experiments demonstrate the efficacy of ML-based intrusion detection, with the boosted decision tree (DT) excelling in overlapping feature selection methods. Remarkably, logistic regression achieves a 98.90% accuracy rate in the initial two evaluation phases. This research contributes significantly to intrusion detection by providing a comprehensive framework for identifying attacks in IoT contexts through ML techniques. Unlike previous studies, our approach utilizes the AWID dataset, which is more aligned with the current threat landscape, addressing the limitations of outdated datasets. Additionally, we bridge the gap between research and practical implementation by conducting our experiments using the WEKA tool. In summary, this study offers a novel and practical solution to the pressing challenge of intrusion detection in IoT networks, promoting safer and more resilient IoT deployments.
Thabit et al. (Sat,) studied this question.