Title: Multiclassification of DDoS attacks using machine and deep learning techniques

Authors: Rashmi Bhatia; Rohini Sharma

Addresses: Department of Computer Science and Applications, Panjab University, Chandigarh, India ' Department of Computer Science and Applications, Panjab University, Chandigarh, India

Abstract: There are very few studies to detect different classes of DDoS attacks. Multiclassification helps network administrators to study individual behaviour. In this study, 82 flow-based features are used to detect 13 types of DDoS attacks using seven machine learning techniques namely naive Bayes, decision tree, multinomial logistic regression, random forest, k-nearest neighbour, AdaBoost and one hidden layer multi-layer perceptron (MLP) and two deep learning techniques namely multiple hidden layers MLP and long short-term memory (LSTM). Different variants of deep learning techniques are compared while fine-tuning hyperparameters. Their performance is analysed using 5-fold cross-validation and compared with existing studies. The experimental results show that random forest performed best with the highest accuracy of 0.7677 followed by one hidden layer MLP with accuracy of 0.7485 and improvements in them can give better results. It is also concluded that appropriate selection of features is important to get higher accuracy with lesser classification time.

Keywords: machine learning; deep learning; multilayer perceptron; MLP; long short-term memory; LSTM; intrusion detection; DDoS attacks.

DOI: 10.1504/IJSN.2024.140268

International Journal of Security and Networks, 2024 Vol.19 No.2, pp.63 - 76

Received: 19 May 2023
Accepted: 24 Sep 2023
Published online: 01 Aug 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article