Detecting Anomalies In Internet Traffic Using Autoencoder Method

This study presents a method to detect unusual activity in Internet traffic using an autoencoder, a type of neural network. We used the publicly available UNSW-NB15 dataset, which includes network traffic data and labels indicating hacker attacks. The data was processed using an entropy method to prepare it for the autoencoder. The analysis involves training the autoencoder to reconstruct the input data. By measuring the difference between the original data and the reconstructed data we can identify anomalies. Large differences, or errors, suggest anomalies, which often correspond to malicious activities. This method leverages the autoencoder's ability to learn and represent data efficiently, making it a strong tool for detecting unusual activities, and thus provides a way to enhance network security.