Lightweight generative adversarial network, and encrypted traffic hiding method and device based on network in anonymous network environment

The invention discloses a lightweight generative adversarial network, and an encrypted traffic hiding method and device based on the network in an anonymous network environment. The method comprises the following steps: 1, performing split hierarchical clustering based on feature relevance weighting on common traffic; 2, performing data enhancement on the class cluster in the step 1; 3, according to the class cluster in the step 2, training and using a variational auto-encoder which introduces an orthogonalization weight matrix with a trade-off factor; 4, outputting concentrated hidden variables and fuzzy traffic by using an encoder and a decoder; 5, training the improved lightweight generative adversarial network according to the result of the step 4; and 6, modifying the encrypted traffic of the anonymous network gateway node according to the result of the step 5. According to the method, a large amount of calculation of the generator and the discriminator is shared to the variational auto-encoder which is