Machine learning architectures for early detection and classification of botnet attack in the internet of things

Abstract

Internet of Things (IoT) is an area that deals with pervasive systems, connected by communication standards such as Bluetooth, WiFi and 5G. To provide low-latency services, edge computing emerges as a paradigm that allows processing data closer to the sensing layer, where information is acquired from environment. Recent IoT applications require real-time solutions, with the security of data and devices guaranteed, focusing on correct data management and the development of energy efficient systems. Applying machine learning techniques to IoT devices is a possible solution that can help with requirements such as data governance (security, reliability and usability), balancing processing and decision making, such as early detection of attacks coming from the network, recognition of objects, or self-improvement of IoT systems. To assist with these requirements, edge computing has also been widely used for the purpose of creating real-time and energy-efficient systems. The goal of this work is to evaluate the use of machine learning in IoT and edge computing environments, focusing on the analysis of the security requirement of applications and devices. For this, two empirical experiments are proposed regarding the security requirement, conducting a study on botnet attacks on IoT devices and possible strategies to protect them against these types of threats.