Basmah Hayder Mahmod
Department of Computer Engineering, Urmia University, Urmia, Iran
Jamshid Bagherzadeh Mohasefi
Department of Computer Engineering, Urmia University, Urmia, Iran
Industrial Internet of Things (IIoT) is a major technological breakthrough that makes it possible for new kinds of automation, optimization, and cooperation between physical and virtual entities. However, complex interplay between digital and physical industrial components of these systems has increased their susceptibility to advanced cyberattacks. Although Intrusion Detection Systems (IDSs) can reliably identify attacks from normal behavior, it is necessary to have a system to categorize network traffic into various classes due to the increasing intricacy and subtlety of these traffics. The idea of federated learning is to solve data privacy issues by leveraging collaborative learning across a wide range of devices. In this paper, we have implemented homogenous and heterogenous FL-based methods to analyze and compare Edge-IIoT dataset to categorize network traffic into different categories. First, we implemented centralized deep learning methods on the specified dataset. Then, we applied homogeneous FL based methods and compared it with the centralized methods using various assessment methods. Then, we considered heterogenous FL models using combined data heterogeneity and structural heterogeneity. The experiments were conducted using 5 categories and 9 categories of traffic in the context of the IIoT dataset. The results show competitive results for both homogenous and heterogenous methods.