A watermarking-based authentication and recovery scheme for encrypted audio

To protect the audio stored on third-party servers, a novel authentication and recovery watermarking scheme for encrypted audio is proposed. We define the relative energy (RE) feature and analyze the characteristic of the feature. At the embedding stage, the host audio is encrypted firstly. Then the encrypted audio is divided into frames. We embed watermark bits into each frame by quantifying the RE feature. At the decoding stage, the receivers locate the attacked frames based on the watermark extraction and substitute the attacked frames using 0 amplitude signals, which are scattered over different segments after anti-scrambling transformation and do not influence the expressed meaning of watermarked signal. Experimental evaluation results demonstrate that the scheme improves the security of audio signals stored on third-party servers.