Exploring Machine Learning Intrusion Detection: Addressing Security and Privacy Challenges in IoT - A Comprehensive Review

Abstract

With billions of IoT devices in operation globally, vast amounts of data are generated, posing significant security challenges throughout the data lifecycle. Machine learning (ML) offers a promising approach to safeguarding IoT systems by swiftly detecting anomalies and enforcing real-time security and privacy (S&P) measures. This systematic literature review investigates ML-based intrusion detection in IoT, examining academic journals from 2011 to 2021 through the IEEE and ProQuest databases. Utilizing the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework, we identify key insights and challenges. Our review reveals that while ML-based Intrusion Detection Systems (IDS) exhibit superior performance in detecting emerging attack trends, they also introduce complexities such as increased computational demands, susceptibility to adversarial attacks, scalability issues, and trade-offs between accuracy and false positives. Furthermore, deep learning methods outperform traditional ML techniques in anomaly detection. Addressing the evolving nature of attacks remains a continuous endeavor, underscoring the ongoing development of IDS.