DDoS Attacks Detection Using a Deep Neural Network Model

In today’s world of globalization, with the exponential increase in the use of Internet-based services, there is also an increase in network traffic and a huge risk of hacking and cyber-attacks. One of the cyber-attacks is the Distributed Denial of Service (DDoS) attack which simply sends the requests to the target for the denial of service for the legitimate users. To detect the DDoS attacks from the vast volume of network traffic, there is a need of a deep learning model. Also, there is the requirement for a lightweight model for today’s high-volume traffic and time critical applications. Therefore, in this paper, a deep neural network (DNN) model has been proposed and validated over the CICDDoS 2019 and PVAMUDDoS-2020 datasets with all features and reduced (important) features. The hyperparameters of the DNN model have been tuned and the best model has an input layer, five hidden layers, and an output layer. The proposed DNN model with the same hyperparameters values performed well over the CICDDoS 2019 and PVAMUDDoS-2020 datasets with all features and reduced features. The results are comparable in both cases and the evaluation with reduced features shows less training and testing time which would be helpful in the time-critical applications and the large volume of network traffic.