EBIDS: efficient BERT-based intrusion detection system in the network and application layers of IoT

The rapid development and expansion of Internet of Things (IoT) systems have brought about increasingly complex security challenges due to the multi-layered structure of IoT, where attackers can exploit vulnerabilities at various layers. Traditional intrusion detection systems (IDS) often fail to detect novel attacks efficiently, especially in resource-constrained IoT environments. To address these security issues, an effective and efficient intrusion detection system (IDS) is crucial. This paper proposes a novel anomaly-based IDS leveraging deep learning techniques, specifically focusing on the BERT (Bidirectional Encoder Representations from Transformers) algorithm. BERT’s architecture enables it to perform less costly evaluations and detections compared to other modern algorithms, making it particularly suitable for resource-constrained IoT environments. Our proposed framework, EBIDS, harnesses the capabilities of BERT to enhance intrusion detection at both the network and application layers of IoT systems. The framework aims to improve detection accuracy, reduce computational overhead, and provide real-time intrusion detection. Experimental results on the Edge-IIoT and Canadian Institute of Cybersecurity Denial of Service 2017 (CICDos 2017) datasets demonstrate that EBIDS achieves faster and more accurate intrusion detection compared to existing methods.