From Network Package Flow to Images: How to Accurately Detect Anomalies in Computer Networks

With the increasing adoption of data networks, attacks will cause more and more damage, including financial losses to organizations and governments. Artificial intelligence, combined with Deep Learning techniques such as Convolutional Neural Networks (CNN), helps organizations detect early attacks on their infrastructures. A CNN is a specialist system in extracting pieces of information about images. This work presents a method that converts data from network connections into images. We use a CNN with three convolutional layers, which will use the generated images as input and extract features based on their weights to perform attack detection. We reached promising results detecting attacks using the NSL-KDD dataset as the input to create the images.