Performance Comparison of Machine Learning and Deep Learning Models in DDoS Attack Detection

In today’s world of the internet and growing network technology, along with the advancement of IoT technologies and automation of features, the security and privacy preservation of the network have become one of the main concerns in today’s world. Due to the internet’s ever-expanding network of gadgets, network intrusion has become a trivial problem to identify and prevent network threats from affecting them. There are numerous types of network attacks, among which Distributed Denial of Service (DDoS) attacks are one of the most dangerous and precarious ones as the user is denied any access to the server resources or the sites connected to it due to flooding of the network traffic by malicious packets of the attacker. In this paper, we have put forward a framework for the detection of DDoS attacks using different machine learning (ML) and deep learning (DL) algorithms. We have made use of the Software–Defined Network (SDN) dataset to identify the attack. First, the dataset was imported, and then the data underwent a data preprocessing stage where all the null values have been filtered from the dataset. Then the encoding and normalization of features have been performed, and the data are split for training and testing of the dataset. The data are then fed to the mentioned classifiers and the accuracy of each classifier is calculated, and the best baseline classifier with the highest accuracy is selected to detect the DDoS attacks. It is observed that the Deep Neural Network (DNN) classifier has an accuracy of 99.18% and is the most suitable classifier for the detection of attacks. After finding the best model, we further improved its performance to 99.21% by visualizing and tuning the hyperparameters of the model.