Whole-network IPv6 active address detection method and device based on transfer learning

The invention provides a whole-network IPv6 active address detection method and device based on transfer learning, and the method proposes a novel generative neural network model in order to accurately extract a seed address mode, can automatically divide a seed with a mixed mode into a plurality of clusters, and each cluster corresponds to the same address mode, so that the efficiency of the whole-network IPv6 active address detection is improved. Each mode is accurately described in a fine-grained mode, and all seed information is fully utilized; a mode of combining small-scale initial detection and large-scale iterative detection is provided, continuous dynamic optimization of the detection process is realized based on two migration methods, and the hit rate of the whole detection process is ensured; according to the invention, a lightweight alias detection algorithm is designed, real-time alias prefix detection can be completed quickly and efficiently at low overhead, and the influence of alias prefixes o