Efficient Fragile Privacy-Preserving Audio Watermarking Using Homomorphic Encryption

Audio has become increasingly important in modern social communication and mobile Internet. In cloud computing, practical cloud-based applications should achieve high computation efficiency and secure data protection simultaneously. In this paper, we study the application of an efficient encrypted audio fragile watermarking using homomorphic encryption and the batching technique SIMD in cloud computing. We firstly implement the algorithm of Haar wavelet transform in the encrypted domain (BS-HWT) using the batching technique SIMD with the fully homomorphic encryption scheme CKKS. By performing BS-HWT on the encrypted audio, we transform the encrypted audio signal into the encrypted frequency-domain coefficients. The encrypted fragile watermark is then embedded into the encrypted discrete wavelet transform domain. Our experimental results show that the proposed watermarking scheme is highly efficient and sensitive to common audio attacks. We also present the proposed scheme has a good ability of tamper localization.