A lightweight web server anomaly detection method based on transductive scheme and genetic algorithms

World Wide Web (WWW) is one of the most popular applications currently running on the Internet and web server is a crucial component for this application. However, network anomalies especially Distributed Denial-of-Service (DDoS) attacks bombard web server, degrade its Quality of Service (QoS) and even deny the legitimate users' requests. Traditional network anomaly detection methods often lead to high false positives and expensive computational cost, thus unqualified for real-time web server anomaly detection. To solve these problems, in this paper we first propose an efficient network anomaly detection method based on Transductive Confidence Machines for K-Nearest Neighbors (TCM-KNN) algorithm. Secondly, we integrate a lot of objective and efficient anomalies impact metrics from the perceptions of the end users into TCM-KNN algorithm to build a robust web sever anomaly detection mechanism. Finally, Genetic Algorithm (GA) based instance selection method is introduced to boost the real-time detection performance of our method. We evaluate our method on a series of experiments both on well-known KDD Cup 1999 dataset and concrete dataset collected from real network traffic. The results demonstrate our methods are actually effective and lightweight for real-time web server anomaly detection.