Underwater sensor node trust identification method based on improved variational automatic encoder

The invention discloses an underwater sensor node trust identification method based on an improved variational automatic encoder. The underwater sensor node trust identification method comprises the following steps that: 1, a sink node collects behavior information of each underwater node as a data set; 2, an improved variational automatic encoder VAE is operated, an encoding part learns features of the data and maps the features to a potential space, and a decoding part tries to reconstruct normal data through probability distribution of the potential space; 3, calculating the loss of data reconstruction, evaluating the credibility of the data, and identifying the data with larger loss as abnormal data; 4, adversarial loss is added into reconstruction loss so as to improve the ability of the VAE to capture underwater node data features; and 5, training the improved VAE for identifying the abnormal behavior of the underwater sensor node. The method can effectively improve the accuracy and efficiency of detect