An Effective Audio Watermarking Algorithm for Authentication With Perceptual Hashing and the Normalized Cuts

Audio watermarking technology is crucial for copyright protection and content authentication in cloud storage, but current methods still have limitations in watermark capacity, data integrity, and authenticity. To address this, this paper introduces a novel high-capacity audio watermarking method for authentication, which is based on perceptual hashing and the normalized cuts (Ncuts). The method first uses Discrete Wavelet Transform (DWT) to decompose the audio signal into multiple frequency bands. By calculating the short-term energy(STE) features of the low-frequency components, a perceptual hash value is generated for the audio. Then, the generated perceptual hash value of the audio is used as a watermark. The audio's approximate coefficients are segmented into multiple clusters through normalized cuts, and the watermark is embedded into the norms using Quantization Index Modulation (QIM). Experimental results show that the method ensures imperceptibility and strong robustness, supporting audio authentication, tampering detection, and localization. Compared to existing technologies, our method achieves a remarkable 1622.66 % improvement in watermark embedding capacity, and under common attacks, the average Bit Error Rate (BER) is reduced by at least 42.34 %.