Categorical classification of distributed denial-of-service (DDoS) attacks using fine-tune EfficientNetB0

The paper demonstrates how Fine-Tuned EfficientNetB0 functions in recognising Distributed Denial-of-Service (DDoS) attacks, which are important threats to cybersecurity. The risks from DDoS attacks become substantial because harmful traffic floods enable disruption of online services and network health operations. The research team engineered a transformation process to turn tabular network traffic data into 64x64x3-dimensioned images for efficient NetB0 to analyze them despite its original design purpose of standard image classification. Its excellent performance, combined with resource efficiency, recommends the use of this model for the required operation. The research optimisations of the pre-trained EfficientNetB0 model derive from different DDoS attack types and normal traffic data in a transformed format to apply image processing for network traffic data pattern identification. Fine-tuned EfficientNetB0 delivers detection precision at 99%, matching above both decision trees and support vector machines in classic machine learning applications regarding detection speed and accuracy. Merging deep learning with image-based network data processing enables real-time DDoS attack detection.