An optimized neural network-based IDS against DDoS attacks

Abstract

As cyberattacks become increasingly sophisticated, the identification of harmful assaults like Distributed Denial of Service (DDoS) poses a significant challenge. DDoS attacks have emerged as a critical threat to the safety of information systems and the dependability of computer networks, which serve as vital infrastructures in the contemporary world. Detecting DDoS attacks necessitates a complex undertaking that demands immediate attention to avert substantial damage to networks and services. To tackle this challenge, security researchers propose leveraging the strengths of the Intrusion Detection Systems (IDSs) and Neural Network (NN) models to safeguard networks against these threats. However, developing an efficient Neural network-based IDS is not a straightforward task due to the intricate architecture of NN models and the large number of hyper-parameters that need to be tuned. In this context, this paper presents a Neural Network -based IDS for DDoS attacks, employing Convolutional Neural Network (CNN) and Multilayer Perceptron (MLP) models optimized using the Bayesian optimization (BO) technique. Experimental results on the CIC-DDoS2019 dataset demonstrate that the CNN achieves an accuracy of 99.71% with a loss value of 0.002. These findings underscore the efficacy of the suggested Intrusion Detection System (IDS) in identifying DDoS attacks within network environments.