Machine learning solutions for mobile internet of things security: A literature review and research agenda

Abstract

In recent years, the advancements in wireless technologies and sensor networks have promoted the Mobile Internet of Things (MIoT) paradigm. However, the unique characteristics of MIoT networks expose them to significant security vulnerabilities and threats, necessitating robust cybersecurity measures, including effective attack detection and mitigation techniques. Among these strategies, Artificial Intelligence (AI), and particularly Machine Learning- (ML) based approaches, emerge as a pivotal method for bolstering MIoT security. In this paper, we present a comprehensive literature survey regarding the utilization of ML for enhancing security in MIoT. Through an exhaustive review of existing research articles, we analyze the diverse array of ML-based approaches employed to safeguard MIoT ecosystems and provide a holistic understanding of the current landscape, elucidating the strengths and limitations of prevailing methodologies. We propose a structured taxonomy to categorize recent works in this domain, by distinguishing approaches based on Shallow Supervised Learning (SSL), Shallow Unsupervised Learning (SUL), Deep Learning (DL), and Reinforcement Learning (RL). By delineating existing challenges and potential future directions for cybersecurity in MIoT, we aim to stimulate discourse and inspire novel approaches towards more resilient and secure MIoT ecosystems.