Detection of DDoS Attacks in Computer Networks Using Deep Learning

Nowadays the security of communication networks has gained increasing significance with the advancement and expansion of large-scale networks like the Internet. Consequently, developing new and efficient systems for detecting malicious traffic, particularly that generated by Distributed Denial of Service attacks among the most severe threats has become a key research focus. This study presents the design and implementation of a system for classifying malicious traffic using Deep Learning architectures, specifically focusing on the CICIDS2017 public dataset. Despite working with sophisticated architectures, the Multilayer Perceptron Artificial Neural Network and its simplicity demonstrated the best classification performance over this dataset. Experimental results indicate that the classifier system achieves satisfactory outcomes, as evidenced by the F1 Score metric, following data analysis and preprocessing steps like class balancing before applying the Deep Learning model. F1 Score was chosen as the primary metric for evaluating model efficiency due to its connection with other crucial metrics, such as Precision and Recall, and the necessity of avoiding false positives and negatives due to the network traffic.