Multi-Modal Noise-Robust DDoS Attack Detection Architecture in Large-Scale Networks Based on Tensor SVD

Abstract

Distributed Denial of Service (DDoS) attacks consumes the resources of traditional or cloud computing networks, resulting in the network unable to provide normal services. Therefore, accurate detection of DDoS attacks can avoid greater losses and provide an important guarantee for network space security. But with the rapid development of the Internet, the network scale is becoming larger and larger, and the structure is becoming more and more complex. Network data shows large-scale heterogeneous characteristics, which lead to data processing becomes more difficult and the traditional algorithms cannot accurately identify attack traffic. Therefore, how to accurately and efficiently detect DDoS attacks in large-scale networks has become a new challenge. To deal this problem, this paper proposes a novel DDoS attack detection framework. Which has mainly made three contributions: (i) Representation of large-scale heterogeneous network data by tensor; (ii) A multi-modal denoising algorithm based on tensor SVD is proposed; (iii) An efficient anomaly detection architecture suitable for large-scale networks is proposed, which combines (i), (ii) and XGBoost classification model. Experiments show that the framework can achieve a high detection rate of 98.84%, and has the characteristics of well extendable, strong noise-robust and fast detection speed.