An Intelligent Distributed-Data Processing Method with Privacy Protection for Industrial Internet of Things

As the rapid development of the industrial Internet of Things (IIoT) progresses, some data in the IIoT start to present the following characteristics: huge volume, high dimensions, distributed storage across multiple devices, and restricted data sharing due to privacy protection concerns. Such data presents a significant challenge to existing data processing methods. To this end, this work proposes an intelligent distributed-data processing method with privacy protection for IIoT (I2DPM). In this method, a federated feature integrator is first designed to capture the global feature subset of the distributed data under privacy protection. Based on the captured feature subset, a many-objective feature selection model is constructed by including the feature number, feature cost, cross-entropy, accuracy, and recall as the five objectives, where these five objectives represent the key factors influencing the feature selection performance. Then, an feedback-assisted information clustering many-objective evolutionary algorithm (MaOEA-IFC) is developed to solve the constructed model and thus obtain the optimal feature subsets, which fully utilizes the ideas of feedforward and feedback control. Finally, MaOEA-IFC is first compared with five state-of-the-art methods on two benchmark test suites to validate its ability to obtain reliable experimental results, and then our method is tested on eight datasets. Extensive results demonstrate that MaOEA-IFC is highly competitive, and our method can obtain the feature subsets with good comprehensive performance on the premise of protecting data privacy. In summary, this work provides a method for processing the data with the above characteristics in IIoT.