VFL-HMF: Enhancing IIoT Security with Federated Learning and Homomorphic Matrix Factorization

Abstract

Matrix factorization is a key technique in recommendation systems, offering dimensionality reduction and collaborative filtering benefits. However, in the context of the Industrial Internet of Things (IIoT), implementing matrix factorization raises critical challenges related to data privacy, security, and computational efficiency. To address these concerns, this study introduces Verifiable Federated Learning with Homomorphic Matrix Factorization (VFL-HMF), a novel model designed to secure sensitive data while enabling collaborative learning. The proposed VFL-HMF model integrates homomorphic encryption and federated learning to ensure data privacy and integrity during computation. By leveraging the VGG-16 Convolutional Neural Network (CNN) architecture, the model extracts detailed features from casting dataset obtained from an industry, achieving high accuracy and robust performance. Experimental results demonstrate that VFLHMF achieves a remarkable accuracy of 93%, surpassing existing approaches, while reducing complexity to 𝑶􁈺𝟏􁈻.This work bridges the gap between privacy-preserving computation and effective collaborative learning in IIoT environments. The VFL-HMF model not only protects sensitive information but also guarantees the verifiability of results, making it a critical solution for secure and efficient data processing. These findings highlight the potential of this approach to revolutionize IIoT applications, paving the way for further advancements in secure federated learning.