The rapid increase in Internet of Things (IoT) devices has significantly transformed modern technological environments. However, this growth has also introduced numerous security vulnerabilities and various attacks. Traditional Machine Learning (ML) and Deep Learning (DL) have been intensively utilized in Intrusion Detection Systems (IDSs) implementation to detect and classify cyberattacks in an IoT environment. ML relies on a feature engineering approach for selecting optimal features, and DL requires high computational resources to achieve adequate classification accuracy. Therefore, this study proposes an IDS model implemented based on lightweight Bidirectional Encoder Representations from Transformers (BERT) models integrated with a meta‐classifier. This is accomplished by employing three lightweight models named DistilBERT, MobileBERT, and TinyBERT as base models, with their outputs combined by a Random Forest (RF) -based meta-classifier to achieve robust detection and classification accuracy while using a lightweight architecture. The proposed model was evaluated on three diverse IoT datasets: ToNIoT, Edge-IIoTset, and X-IIoTID to demonstrate its effectiveness in detecting and classifying cyberattacks in IoT environments. In addition, the results were compared with existing benchmark IDS models. The experiments confirm the robustness of integrating lightweight BERT models with a meta-classifier for the IoT IDS field.
AlGarni et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: