Collusion attack is considered as one of the most popular and challenging attacks for audio signals, which violates the copyright seriously. Conventional methods cannot cope with the hybrid attack which consists of collusion attack and other attacks. In this paper, we propose a pre-adjustment process (PAP) based mechanism to tackle it by destroying the perceptual quality of the colluded signal, which removes the motivation of traitors to implement collusion attack. We transform the host audio signal into DCT domain and segment the DCT coefficients into blocks. Then the DCT coefficients are modified according to a pre-designed adjustment matrix (AM) to generate the PAP signal. When multiple PAP signals are averaged to generate the colluded signal, the energy of certain frequency bands in the colluded signal will be eliminated or reduced, which degrades the perceptual quality of the colluded signal greatly. The proposed method can withstand not only collusion attack but also hybrid attacks. It is also secure, as the secret keys used in PAP will not be passed to the receiver. By combining the proposed method with other leading-edge watermarking algorithms, its performance on copyright protection can be further improved. Theoretical analysis and experimental results show the superiority of the proposed mechanism.