Machine Learning Based Techniques for Cyberattacks Detection in the Internet of Things Infrastructure

The emergence of the concept of the Internet of Things has revolutionized many economic sectors and areas of human activity. At the same time, the spread of the Internet of Things has led to the emergence of cyber security risks in those areas of human activity for which cyber security problems were not relevant before. Most of the security problems in IoT infrastructure arise from a lack of basic security controls. In particular, open ports, security issues with the protocols used in the IoT infrastructure, outdated applications and components of IoT devices, lack of automatic firmware updates for smart devices (or no update releases at all), insecure update mechanisms, insecure settings (by default), weak passwords, vulnerabilities in software and web applications, direct network connection to the Internet, insecure authentication methods. Exploitation of vulnerabilities in routers, storage systems, access control and other IoT devices contributes to the spread of malicious software in the IoT infrastructure and compromising IoT devices. The low level of security of IoT devices leads to the fact that a large number of such devices can be compromised with a high degree of probability and used as a means to carry out various attacks both inside and outside the IoT infrastructure. Attacks on IoT infrastructure result in device hacking, data theft, financial loss, instability, or even physical damage to devices. In turn, given the specific nature of these hacked IoT devices, damage to them can lead to injury to people working or dependent on these devices. At the same time, the owners of hacked IoT devices indirectly become accomplices in cyber-crimes. The article provides an overview of known methods for detecting cyber-attacks on the infrastructure of the Internet of things based on machine learning methods. Despite the large number of such approaches, the problem of detecting zero-day cyber-attacks in the IoT infrastructure is still unresolved. This leads to the need to find new approaches that can solve this problem.