Detecting network attacks by anomaly search method is to identify behaviors that deviate from established baseline parameters, signaling potential security incidents. In this paper, the authors consider the application of convolutional neural network for network traffic anomaly detection. As part of the study, a convolutional neural network has been developed, trained on the dataset CICIDS2017 dataset and quality assessment has been carried out. Based on the developed neural network, a prototype for anomaly detection in network traffic has been built. Testing and quality assessment of the prototype on the CSE-CIC-IDS2018 dataset has been performed.