A Comprehensive Taxonomy of Adversarial Attacks on Machine Learning in IoT Application
M. Praveena, S Madhumitha, J Menakadevi, V Lalith Akkash · 2023
This research is a comprehensive taxonomy of threats against Internet of Things (loT-)based Machine Learning (ML) systems. As the number of IoT applications continues to surge, protecting machine learning models from malicious attacks is more essential than ever. This study categorizes various attack techniques, ranging from white-box to black-box, evasion to poisoning, and beyond, to provide an overview of the evolving security environment within IoT systems. The proposed study will provide attack techniques, evaluation metrics, defense mechanisms, and case studies to shed additional light on securing Internet of Things (IoT) devices. This research contributes to understanding how to use machine learning safely in the Internet of Things by analyzing the prevalence and severity of attacks, the effectiveness of countermeasures, and the effect on system performance.