Pervasive growth of Internet of Things (IoT) is visible across the globe. The 2016 Dyn cyberattack exposed the critical fault-lines among smart networks. Security of Internet of Things (IoT) has become a critical concern. The danger exposed by infested Internet-connected things not only affects the security of IoT, but also threatens the complete Internet eco-system which can possibly exploit the vulnerable Things (smart devices) deployed as botnets. Mirai malware compromised the video surveillance devices and paralyzed Internet via distributed denial of service (DDoS) attacks. In the recent past, security attack vectors have evolved bothways, in terms of complexity and diversity. Hence, to identify and prevent or detect novel attacks, it is important to analyze techniques in IoT context. This survey classifies the IoT security threats and challenges for IoT networks by evaluating existing defense techniques. Our main focus is on Network Intrusion Detection Systems (NIDS); hence, this paper reviews existing NIDS implementation tools and datasets as well as free & open-source network sniffing software. Then, it surveys, analyzes and compares state-of-the-art NIDS proposals in the IoT context in terms of architecture, detection methodologies, validation strategies, treated threats and algorithm deployments. The review deals with both traditional and machine learning (ML) NIDS techniques and discusses future directions. In this survey, our focus is on IoT NIDS deployed via Machine Learning since learning algorithms have a good success rate in security and privacy. The survey provides a comprehensive review of NIDSs deploying different aspects of learning techniques for Internet of Things, unlike other top surveys targeting the traditional systems. We believe that, the paper will be useful for academia and industry research, first, to identify IoT threats and challenges, second, to implement their own NIDS and finally to propose new smart techniques in IoT context considering IoT limitations. Moreover, the survey will enable security individuals differentiate IoT NIDS from traditional ones.
Deep learning approaches, which are a part of methods we call today artificial intelligence (AI), have become indispensable for a wide range of applications requiring analysis and understanding of a large amount of data. They can produce promising results that outperform human capabilities in various decision tasks relating to visual content classification and understanding such as face detection,1 object detection and segmentation,2 image denoising,3 video-based tasks such as sports action recognition4 and saliency detection,5 and among others. The success of deep learning-based systems in these tasks have also paved the way for their applications to be developed for a variety of medical diagnosis tasks such as cancer detection6 and Alzhiemer’s disease detection7 on different imaging modalities just to name a few. Along with the usefulness of these tools, the trustfulness and reliability of such systems are also being questioned.
1 October 2018 in Journal of Information Security and Applications