Using statistical analysis and support vector machine classification to detect complicated attacks

Anomaly detection systems can detect unknown attacks, but they have a high false alarm rate. This article introduces our prototype that uses statistical analysis and support vector machine classifier to detect complicated attacks. We research the sampling methods of statistical analysis techniques, and propose a new statistical model named Smooth K-Windows. An improved support vector machine classifier that has higher accuracy is proposed after analyzing the reason why support vector machine makes misclassifications. The experimental results show that the prototype system can detect complicated attacks in which the attackers stash their behavior by changing it gradually.