Abstract: In this paper, a technique of watermark using in audio authentication and recovery is presented. The arithmetic is based of audio characteristic to create watermark, and make this watermark embedded in the audio itself. It can not only check and locate tamper to the audio, but also similarly recover. The main idea of this arithmetic is to divide audio into blocks, and make each block DCT transformed. And the main DCT coefficients are quantized and encrypted to be the watermark. It is then be embedded to LSB bit of the audio itself. This idea can guarantee the recovery quality. The watermark is based of main DCT coefficients and its position, therefore, it can recover with blind source.