IXP scrubber
Matthias Wichtlhuber, Eric Strehle, Daniel Kopp, Lars Prepens, Stefan Stegmueller, Alina Rubina, Christoph Dietzel, Oliver Hohlfeld
Abstract
Distributed Denial-of-Service (DDoS) attacks have plagued the Internet for more than 25 years. Not only have they steadily increased during this period in numbers, but the attacks have also gotten bigger, have been aimed at new targets, and thanks to the proliferation of inexpensive DDoS-as-aservice platforms, attackers with little or no expertise are now able to launch them, be it for political, economic, or societal reasons. Seemingly, some 20+ years of research efforts focused on detecting and mitigating DDoS attacks have been unable to prevent, not to mention eradicate, DDoS attacks as a means for crippling critical network infrastructure, slowing or completely stopping various providers' online services, or demanding ransom payments. However, with Machine Learning (ML) emerging as a promising approach to solving networking problems, there has been renewed hope that with ML-based solutions, it will finally be possible to accurately detect DDoS attacks and stop them in their tracks.