A Survey on Machine Learning Approaches For IOT Device Security

Abstract

IoT is a system of connected devices which can be uniquely identified over a network and transfer huge amounts of data, communication, monitoring and remote control of automated devices through internet. IoT are generally designed as embedded systems and perform a critical role to support vertical applications by connecting heterogeneous devices. Since, IoT networks are usually resource constrained it makes them an easy target for cyber-attacks. Conventional authentication and authorization approaches based on static digital mechanism and computational complexity are not good enough. Isolated security designs for different layers and links while ignoring overall protection leads to cascading security risks as well as growing communication latency in IoT devices as well as traditional cryptographic approaches are not enough as the nodes have massive real time data, heterogeneity, extensive dynamic behavior of networks. Therefore, machine learning and deep learning provide embedded intelligence IoT uses enabling technologies such as Software Defined Systems (SDN) , cloud computing and fog computing.