Development of a Neural Network Module for Detecting Network Threats in Traffic Based on Convolutional and Recurrent Neural Networks

Increased accuracy of detecting different types of network attacks is possible by using neural network algorithms. For Intrusion Detection Systems and Intrusion Prevention Systems there are a lot of datasets such as LYCOSIDS2017 containing records of normal traffic and network attacks. The training of models on preprocessed data made it possible to achieve an accuracy of detecting attacks in network traffic of at least 90% when working with known threats and at least 80% when new types of threats appear which indicates the possibility of using this method to provide protection in telecommunication systems.