BEWSAT: Blockchain-Enabled Watermarking for Secure Authentication and Tamper Localization in Industrial Visual Inspection

Document Type

Conference Proceeding

Publication Date

8-2025

Department

Department of Applied Computing

Abstract

Thanks to fast advancements in computer vision and industrial Internet of Things (IIoT) technologies, vision-based inspection has been widely used in industrial manufacturing system. Meanwhile, collecting, storing, and sharing inspection data among participants also brings security concerns. An adversary can modify inspection data to make falsified quality reports or launch denial of service (DoS) attacks to make centralized inspection services be unavailable to legitimate users. Therefore, ensuring authenticity, integrity and availability of visual inspection data is crucial for operational safety, quality assurance, and regulatory compliance. In this paper, we present BEWSAT, a blockchain-enabled watermarking framework for secure authentication and tamper localization in industrial visual inspection systems. The watermarking method is based on a generative adversarial network (GAN), where the GAN's encoder consists of two components: feature refinement module (FRM) and attention-aware module (AAM). First, the FRM effectively captures and integrates both shallow and deep features of the image, enabling multilayer refinement of the watermark. To reduce distortion caused by watermark embedding, the AAM is employed, generating an attention mask derived from global image features. This mask emphasizes less prominent and textured areas, allowing for more robust watermark embedding, while downplaying other features to optimize the watermarking process. Additionally, a blockchain-based security fabric provides tamper-proof storage and verification of watermarking keys, which guarantees integrity and availability as storing and sharing inspection data. Experimental results demonstrate that BEWSAT is robust to various hybrid signal processing and geometric attacks. Moreover, numerical results show that BEWSAT incurs limited latency during industrial visual inspection data publication and verification. These findings highlight a significant improvement over the baseline, emphasizing the framework's superiority.

Publication Title

Proceedings of SPIE the International Society for Optical Engineering

ISBN

[9781510694231]

Link to Full Text

Share

COinS