A Network Intrusion Detection System using Deep Learning against MQTT Attacks in IoT
Fatemeh Mosaiyebzadeh, Luis Gustavo Araujo Rodriguez, Daniel Macêdo Batista, Roberto Hirata
Abstract
Cyber-attacks and threats are growing fast in the Internet of Things (IoT) infrastructure as applications in smart cities gain momentum. Usually, IoT devices communicate via machine-to-machine protocols such as Message Queuing Telemetry Transport (MQTT). Due to the heterogeneous structure in IoT and the absence of security by design methodologies, security mechanisms in environments with MQTT traffic are needed, and they can be deployed as Intrusion Detection Systems (IDS). This paper proposes a Deep Learning (DL) based Network IDS trained using a public dataset containing MQTT attacks. We assess the proposal using standard performance metrics such as accuracy, precision, recall, F1-score, and weighted average. When evaluating the performance of our DL-based Network IDS, it obtained, in average, 97.09% of accuracy and an F1-score equal to 98.33% in the detection of MQTT attacks. Another important contribution of our work is the sharing of the experiments on GitHub, which guarantees the reproducibility of the research.