Two-step ensemble approach for intrusion detection and identification in IoT and fog computing environments

• Published in: Computers & electrical engineering Vol. 98; p. 107694
• Main Authors: de Souza, Cristiano Antonio, Westphall, Carlos Becker, Machado, Renato Bobsin
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier Ltd 01.03.2022; Elsevier BV
• Subjects: Artificial neural networks; Cloud computing; Ensemble learning; Fog computing; Internet of Things; Intrusion detection; Intrusion detection systems; Intrusion identification; Machine learning; Fog computing; Internet of things; Intrusion identification; Ensemble learning; Intrusion detection
• ISSN: 0045-7906; 1879-0755
• DOI: 10.1016/j.compeleceng.2022.107694

Due to Internet of Things devices resource limitations, security often does not receive enough attention. Intrusion detection approaches are important for identifying attacks and taking appropriate countermeasures for each specific threat. This work presents a two-step approach for intrusion detection and identification. The first step performs a traffic analysis with an Extra Tree binary classifier. Events detected as intrusive are analyzed in the second stage by an ensemble approach consisting of Extra Tree, Random Forest, and Deep Neural Network. An extensive evaluation was performed with the Bot-IoT, IoTID20, NSL-KDD, and CICIDS2018 intrusion datasets. The experiments demonstrated that the proposed approach could achieve similar or superior performance to other machine learning techniques and state-of-the-art approaches in all databases, demonstrating the robustness of the proposed approach. [Display omitted]