Title: Collaborative 'many to many' DDoS detection in cloud

Authors: Siqi Ma; David Lo; Ning Xi

Addresses: School of Information System, Singapore Management University, Singapore 178902, Singapore ' School of Information System, Singapore Management University, Singapore 178902, Singapore ' School of Computer Science and Technology, Xidian University, Xi'an 710071, Shaanxi, China

Abstract: Cloud computing provides a scalable and cost-effective environment for users to store and process data through the internet. However, it also causes distributed denial-of-service (DDoS) attacks. DDoS attacks risk systems outage and intend to disable the service to legitimate users. In this paper, due to the nature of its large-scale and coordinated attacks, we propose a collaborative prediction approach for detecting DDoS. Our approach provides a clean and direct solution to attack defense. The DDoS attacks follow certain patterns when employing a large number of compromised machines to request for service from the servers in the victim system. So we construct an attacker-server utility matrix by the number of packets and adopt matrix factorisation to detect potential attackers collaboratively. We derive the latent attacker vectors and latent server vectors to predict the unknown entries in the matrix. Experimental results on the NS-2 simulation networks demonstrate the superiority of our approach.

Keywords: cloud computing; DDoS detection; collaborative detection; matrix factorisation; cloud security; distributed DoS; denial of service; DDoS attacks; simulation.

DOI: 10.1504/IJAHUC.2016.079269

International Journal of Ad Hoc and Ubiquitous Computing, 2016 Vol.23 No.3/4, pp.192 - 202

Received: 26 Jan 2015
Accepted: 15 Jun 2015
Published online: 26 Sep 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article