Audio watermarking is the technique of information insertion into an audio signal without altering the significant origin of the audio host. Audio watermarking is usually used to protect the copyright of a work in the form of songs, recording of state and other secrets to avoid irresponsible persons. Audio watermarking technique is passed through two stages including embedding, and extracting. The extracted audio signal must be resistant to several attacks. This research analyzes the third scheme of audio watermarking system on Arnold Transform with Discrete Wavelet Transform-Singular Value Decomposition (DWT-SVD) methods. The results show that the systems we designed are robust to some attack types, namely LPF 10 kHz, cropping, MP3 compression 256 kHz, and MP3 compression 512 kHz. Our system has BER=0.0012, PSNR=29.2082 and CC=0.9976 for compression and LPF attacks; while with cropping attack this paper shows BER=0, PSNR= ∞ and CC=1.