Hybrid Intelligence for DDoS Defense
Year: 2025
Type: article
Abstract: Recent advances in machine learning, deep learning, and large language models enable the design of refined and complex algorithms to detect and prevent cybersecurity attacks. In this paper, we present a hybrid fusion approach combining Generative AI, ADASYN, Recursive Feature Elimination (RFE), and boosting algorithms to detect DDoS attacks. RFE was employed to optimize feature selection, enhancing model interpretability and performance by reducing dimensionality. The proposed model leverages (1) Packet Capture (pcap) data generated from virtual networks as real data, (2) synthetic data generated by the Synthetic Data Vault, (3) ADASYN to balance the data, and (4) boosting algorithms for training and testing. The results obtained from this hybrid-fusion model provided an accuracy of 97–98%, indicating that the model is robust and reliable. Cross-validation of the model further validated the results
Authors: Lakshmi Prayaga, Chandra Prayaga, Rhys Misstle, Mariah Zuanazzi, Sri Satya Harsha Pola
Institution: University of West Florida
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4407960111