A Lightweight Residual Networks Framework for DDoS Attack Classification Based on Federated Learning
Tian Qin, Cheng Guang, Wenchao Chen, Wu Si
Abstract
With the development of network technology, more and more protocols and devices are used in DDoS reflection and exploitation attacks. Different DDoS attacks often require different responses, so in order to protect against DDoS attacks, it requires not only DDoS detection, but also the classification of the detected DDoS traffic. As DDoS attacks have a wide range of potential targets, the existing algorithms for DDoS detection needs global processing of traffic data from different sources, but such data collection is obviously not conducive to the privacy protection of source users. Therefore, we designed our algorithm for DDoS detection and classification based on federated learning. Deferent terminals only need to pass model gradient parameters rather than directly interact the collected data of them, which can not only reduce communication costs, but protect privacy as well. Considering that network structure should not be too complex for high-rated traffic classification, we use a simplified residual network with fewer parameters for detection and classification.