Anomaly intrusion detection systems in IoT using deep learning techniques: a survey

• Main Authors: Alsoufi, Muaadh. A., Razak, Shukor, Md. Siraj, Maheyzah, Ali, Abdulalem, Nasser, Maged, Abdo, Salah
• Format: Conference or Workshop Item
• Published: 2021
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/100282/; http://dx.doi.org/10.1007/978-3-030-70713-2_60
• Institution: Universiti Teknologi Malaysia

Security has a major role to play in the utilization and operations of the internet of things (IoT). Several studies have explored anomaly intrusion detection and its utilization in a variety of applications. Building an effective anomaly intrusion detection system requires researchers and developers to comprehend the complex structure from noisy data, identify the dynamic anomaly patterns, and detect anomalies while lacking sufficient labels. Consequently, improving the performance of anomaly detection requires the use of advanced deep learning techniques instead of traditional shallow learning approaches. The large number of devices connected to IoT which massively generate a large amount of data require large computation as well. This study presents a survey on anomaly intrusion detection using deep learning approaches with emphasis on resource-constrained devices used in real-world problems in the realm of IoT. The findings from the reviewed studies showed that deep learning is superior to detect anomaly in terms of high detection accuracy and false alarm rate. However, it is highly recommended to conduct further studies using deep learning techniques for robust IDS.