All-Packets-Based Multi-Rate DDoS Attack Detection Method in ISP Layer

Distributed denial of service (DDoS) is a fundamental security problem in the ISP layer internet things. However, most existing DDoS detection methods are based on NetFlow data, which cannot handle huge delay flow generation and massive network traffic. Besides, it extremely hard to obtain real attack traffic construct traditional supervised binary classification model. To solve these problems, this paper proposes novel all-packets-based method (APDD). Firstly, new probabilistic storage model square sketch designed, has structural characteristics parallelization, accumulation, recompression. The its conducive fast efficient compression. All packets mapped into sketch, compressed obtained. Secondly, order overcome poor samples, only according recompressed normal network, one-class classifier constructed by generative adversarial networks form likelihood score obtained judge whether or not belongs state. Finally, two data sets high-throughput utilized evaluate proposed method. Compared with several methods, experimental results show that APDD good time efficiency performance.