Machine Learning for Privacy Threat Classification: A Systematic Review

Privacy threat classification plays a vital role in protecting sensitive information in today’s digital landscape, especially across domains such as the Internet of Things (IoT), smart devices, and cloud-based services. As traditional rule-based security mechanisms delay processing the limitations of under increasing data volume and threats, machine learning (ML) techniques have emerged as promising solutions. Despite significant advancements, these ML driven systems face challenges including data heterogeneity, sparsity, unlabeled data, class imbalance, evasion attacks, and concept drift which limit their performance in real-world applications. Existing surveys on this topic have largely focused on narrow domains such as IoT or federated learning, lacking limitations and omitting critical aspects such as how algorithms works, deployment scalability, and cross-domain applicability. In response, this systematic literature review aims to identify the available machine learning algorithms for privacy treat classification, limitations of the algorithms and evaluate current mitigation strategies. By addressing three core research questions, the study provides existing work, highlights unresolved challenges of ML algorithms, and proposes future research directions to mitigate these limitations