Abstract: To address the challenge of scanning the entire IPv6 network due to its vast address space, a GAN-based active IPv6 address generation algorithm is proposed. The algorithm employed a prefix tree structure to eliminate alias addresses, thereby reducing their interference during training and improving hit quality. It utilized generative adversarial networks GANs to learn the probability distribution characteristics of seed address sets, enhancing the hit rate of generated addresses. Experimental results demonstrate that the proposed algorithm achieves higher address quality and hit rate compared to existing methods.