Edge computing technology can effectively reduce the pressure of the cloud center and improve the efficiency of data processing, and its data security issues have attracted wide attention. This study focuses on the detection of cloning attacks in edge computing. First, the clone nodes were identified by channel authentication, then the training set was established to train Back Propagation Neural Network (BPNN), and finally unknown nodes were detected by the trained model. The experiment showed that the method achieved the highest recognition accuracy when the number of neurons was 3, and the accuracy of detecting the testing test set was 89.6%. Compared with Efficient Distributed Detection (EDD) algorithm, the method had higher average accuracy, 87.77%, and lower energy consumption, 44 J, which suggests the method, was reliable. This research provides a new method for cloning attack detection, which is conducive to improving the security of edge computing and promoting the higher and faster development of edge computing.