This survey outlines the methodology for constructing a Network Intrusion Detection System (NIDS) for Industrial Internet of Things (IIoT) scenarios utilizing machine learning algorithms. The process is delineated into three primary steps. The first step involves the acquisition of datasets for training, testing, and validating the model. Various publicly available datasets are identified, highlighting their limitations and the specific challenges in the IIoT sector, such as the lack of representation for devices like HMIs, SCADA systems, and PLCs. The second step addresses the challenge of feature extraction from these datasets, noting the absence of standardization and reliance on individual experimental approaches by developers and researchers. The final step focuses on the application of machine learning algorithms, demonstrating that complex models incorporating multiple algorithms achieve higher accuracy compared to simpler classical models. This finding suggests the potential for further optimization and innovation in NIDS design.