Multi-terminal identity authentication method, device and system based on feature perception

The invention provides a multi-terminal identity authentication method, device and system based on feature perception. The method comprises the following steps: acquiring a behavior feature data set of a plurality of cross-network communication terminals; inputting the data set into a trained BP neural network model, and determining an identity authentication result of each cross-network communication terminal based on the output of the BP neural network model; the model is obtained by training based on a confidence interval upper bound algorithm and comprises a plurality of hidden layers, the BP neural network model takes the server as a decision maker, the reciprocal of a loss function between every two times of training is taken as a reward value in the training process, the number combination of neurons of each hidden layer is taken as a rocker arm, and learning is carried out through a multi-arm bandit algorithm. According to the invention, continuous identity authentication of cross-network communicatio