In response to the inefficiency of the existing advGAN method, which requires calculating the differences between the adversarial samples and the normal samples one by one when generating adversarial perturbations, this paper proposes a quantum adversarial sample generation algorithm (QASGA) based on adversarial quantum generative adversarial networks. First, the real samples are encoded into quantum states, then the generator G of QGAN is used to generate adversarial perturbations, which are superimposed with normal samples to obtain adversarial samples. At the same time, the SWAP-TEST method is used to calculate the similarity between all real samples and adversarial samples at once, thereby accelerating adversarial attacks. Experimental results show that the QASGA algorithm proposed in this paper can generate high-fidelity adversarial samples with a one-time similarity calculation on the IRIS dataset, and can reduce the classification accuracy of the BP neural network from 96.67% to 26.67%, verifying the effectiveness of the proposed QASGA algorithm.