Machine Learning-Based Security Strategies in the IIoT: A Systematic Review (#1061)

Authors

Flores Cruz, Cesar Alexander

Yauri Quispe, David

Rada Mota, Luis Carmelo

Marzal Martinez, Walter Rolando

Abstract

This Systematic Literature Review (SLR) aimed to analyze the applications of machine learning (ML) in the security of the Industrial Internet of Things (IIoT ), identifying current advances, challenges and gaps. To structure the research questions, the PICO methodology was first used, which allowed the review to be oriented towards risks, applied strategies, comparisons with traditional methods and improvements achieved. Subsequently, the PRISMA protocol was applied for the process of selection and refinement of studies, obtaining a total of 31 scientific articles from the Scopus and Web of Science. The results show that ML has significantly improved the detection of threats such as ransomware, zero-day attacks and APTs, outperforming traditional strategies in accuracy, adaptability and efficiency. Strategies such as neural networks, federated learning, hybrid models and edge architectures were identified. However, limitations such as poor validation in real environments, lack of interpretability and vulnerability to adversarial attacks persist. In conclusion, machine learning represents a key advance in the protection of IIoT infrastructures, although further applied research, development of explainable solutions and adoption of common standards are required to strengthen its effective implementation.