Non-Parametric Early Warning Signals from Volumetric DDoS Attacks

Distributed Denial of Service (DDoS) is a classic type of Cybercrime and can still strongly damage company reputation and increase costs. Attackers have continuously improved their strategies, and the amount of unleashed communication requests has doubled in volume, size and frequency. This has occurred through different isolated hosts, leading them to resource exhaustion. Previous studies have concentrated efforts in detecting or mitigating ongoing DDoS attacks. However, addressing DDoS when it is already in place may be too late. In this article, we attract the attention for the crucial role and importance of the early prediction of attack trends in order to support network resilience. We suggest the use of statistical and non-parametric leading indicators for identifying trends of volumetric DDoS attacks and we report promising results over real dataset from CAIDA.