Preserving Privacy for Blockchain-driven Image Watermarking using Fully Homomorphic Encryption

Blockchain-driven image watermarking enables traceable and transparent watermark embedding for undeniable authenticity or proof of ownership. Nonetheless, blockchain public ledger is not suitable for privacy concern applications such as for personal health records or digital image transactions. This paper investigates the feasibility and impact of using Fully Homomorphic Encryption (FHE) to preserve user privacy during watermark embedding while retaining the watermarking traceability using blockchain smart contracts. We evaluate the FHE feasibility using SVD-based watermarking by breaking down the SVD operations into fully and partial SVD operations. We consider three performance metrics: the noise budget, storage requirement, and computation cost. For quality evaluation, we analyze the effect of fractional encoding on the quality of the watermarked image and the embedded watermark. The results show that only the partial singular value addition (SVA) operation has a feasible performance by having low computation cost and low noise expense. It incurs a negligible degradation of the watermark quality, either on the watermarked image or in the extracted watermark data.