A secure federated learning privacy method for industrial IoT edge networks

The rapid growth of the Internet of Things (IoT) in industrial operations has driven the adoption of the Industrial Internet of Things (IIoT), necessitating intelligent networks of edge devices to efficiently generate, analyze, and utilize data from sensors. However, secure transmission of data within edge networks presents significant challenges, including privacy concerns and difficulties in secure data sharing. Existing methods addressing these issues often impose high computational overhead, negatively impacting efficiency. To address these limitations, a novel method, Federated Learning with Enhanced Privacy for Industrial IoT Edge Networks (FLEPNS), is proposed to adopt the edge network system and enhance privacy preservation while optimizing training efficiency. This approach incorporates the Paillier algorithm to implement an information masking mechanism and a shared token system, ensuring secure and obfuscated multi-device data sharing. FLEPNS achieves robust privacy protection without compromising model training accuracy or imposing substantial computational overhead. Additionally, a masking algorithm (SET) is introduced to counter adversarial attacks and ensure data integrity during sensor deployment and transmission between edge servers and devices. Experimental evaluations demonstrate that FLEPNS outperforms compared techniques for accuracy, showing a value of 62% for PAFLM and 70% for FLEPNS. For efficiency of privacy preservation, the FLEPNS has a higher value of 77% compared to 74% 495 datasets per time from about 5 devices. Further evaluation reveals computational overhead and bandwidth usage by PALFM of 4.122MBps, in contrast to 3.1MBps for FLEPNS), showing significant advantage over compared techniques. These results highlight the distinct performance and practical benefits of FLEPNS in industrial edge network applications.