Method for Adjusting Artificial Neural Network Hyperparameters for Network Intrusion Detection Systems

The paper discusses an algorithm for tuning the hyperparameters of the artificial neural network (topology, activation function, learning algorithm - hyperparameters), which acts as a classifier of attack types in a computer network. It is shown that properly tuned artificial neural network can significantly improve the quality of attack type detection and provides a sufficiently high degree of detection of new types of attacks (even those that were not in the training sample).