The number of cyber attacks against computer networks and systems is constantly increasing. Various countermeasures are currently available to protect computer networks,
such as firewalls or cryptography. Among the available solutions is the intrusion detection system (IDS), which identifies malicious traffic within a computer network
and alerts the network administrator in real time. Based on the study of systems in the literature, several real-time intrusion identification systems were developed
in this work using deep-learning networks. A tuning phase was performed using a reduced dataset in order to be able to choose the best configuration of hyperparameters
for each neural network from those chosen. Finally, the configured systems were tested using a complete dataset so that the best performing network could be identified.
In conclusion, the data obtained show that the best network guaranteeing the best results is the convolutional neural network with fullconnected autoencoder processing.