Cyber Attack Classification using a Multilayer Perceptron

Abstract

The continued evolution of Distributed Denial-of-Service (DDoS) attacks poses a serious threat to network security. Numerous studies have employed machine learning and deep learning techniques for DDoS detection; however, most have focused on improving overall accuracy without adequately addressing the class imbalance characteristic of the CICDDoS2019 dataset. Moreover, the exploration of the Focal Loss function within a Multilayer Perceptron (MLP) architecture for multiclass classification remains limited. This study aims to develop an MLP-based cyberattack classification model using Focal Loss to address class imbalance. The research process includes data preprocessing, feature normalization, class weighting, and model training using two hidden layers with ReLU activation and a Softmax output layer. The model was evaluated using accuracy, precision, recall, F1-score, and confusion matrix analysis. The experimental results show that the MLP model achieved an accuracy of 98.80%, with a weighted F1-score of 0.99 and a macro F1-score of 0.87. Although Random Forest achieved a higher overall accuracy of 99.18%, the MLP model demonstrated competitive and stable performance in handling imbalanced data. This study provides preliminary evidence of the potential of an MLP model with Focal Loss for the development of a deep learning-based Intrusion Detection System (IDS), although further isolated testing is required.