Abstract:

The Internet of Things (IoT) is a brand-new paradigm that unifies the Internet with actual physical things from several domains, including home automation, business processes, environmental monitoring, and human health. It increases the prevalence of Internet-connected gadgets in our daily lives, bringing with it concerns linked to security issues in addition to many positive effects. Due to the vast diversity of IoT devices, limited computational resources, and protocols and standards, secured communication is a common difficulty. Even with certain security precautions, IoT networks are extremely susceptible to a variety of threats due to their enormous attack surface. Designing protection measures is therefore required for identifying attackers. However, due to the IoT's unique features including resource-constrained devices, distinct protocol stacks, and standards, applying typical IDS approaches to it is challenging. A number of issues with traditional IDS, such as the high false alarm rate and low detection accuracy, are brought out, just as they are in literature. Because of the computational limitations and inherent resources of IoT systems, it cannot be secured directly by using traditional security techniques. ML techniques integrated with IDS enable real-time detection of both unknown and known attacks on IoT devices. In this article, a thorough analysis of traditional Deep Learning (DL) and Machine Learning (ML) methodologies as well as cutting-edge technologies for intrusion detection in the Internet of Things is done. Our goal is to discover emerging trends, open issues, and promising areas for future research. In this review, various attack detection approaches are clearly discussed along with their advantages and disadvantages.

Description:

Internet of Things (IoT), Intrusion Detection System (IDS), Machine Learning (ML), Deep Learning (DL).