Statistical Properties of DDoS Attacks

Despite the evolving practical approaches to avoid DDoS attacks, it still poses a thread against information services. This paper aims to aid researchers in DDoS detection field to choose better statistical methods by examining structured probability distribution functions among frequently used traffic features in DDoS detection. Ten different continuous probability distributions are fitted using three different measures to attack and normal states of traffic feature vectors. Best probability distributions are chosen to perform binary hypothesis testing and likelihood ratio test.