Transforming Tabular Data into Image Features for Robust DDoS Attack Detection Using Deep Learning

Abstract

This paper introduces an innovative deep learning-based method for robust DDoS attack detection by transforming tabular network traffic data into color image representations. Traditional machine learning approaches often struggle to capture the complex relationships present in tabular data, limiting their effectiveness against sophisticated cyber threats. To address this, the proposed approach converts each network traffic record into a structured color image, enabling the model to learn spatial and chromatic feature correlations. Leveraging EfficientNetB0, a cutting-edge convolutional neural network architecture, the system extracts rich feature representations and performs precise classification. Key preprocessing steps, including feature selection and data augmentation, are applied to improve input quality and model generalization. Experimental results demonstrate that the proposed model achieves outstanding performance, with an accuracy of 99.0%, precision of 99.2%, recall of 98.8%, and an F1-score of 99.0%. These metrics highlight the model’s exceptional ability to accurately identify and distinguish DDoS attacks from normal traffic. By transforming tabular data into informative color images and utilizing advanced deep learning techniques, this study presents a scalable and effective framework for enhancing network security and improving detection capabilities against increasingly complex cyberattacks.