Mahalanobis Distance Map approach for Anomaly Detection of web-based attacks

Web servers and web-based applications are commonly used as attack targets. The main issues are how to prevent unauthorised access and to protect web servers from the attack. Intrusion Detection Systems (IDSs) are widely used security tools to detect cyber-attacks and malicious activities in computer systems and networks. In this paper, we focus on the detection of various web-based attacks using Geometrical Structure Anomaly Detection (GSAD) model and we also propose a novel algorithm for the selection of most discriminating features to improve the computational complexity of payload-based GSAD model. Linear Discriminant method (LDA) is used for the feature reduction and classification of the incoming network traffic. GSAD model is based on a pattern recognition technique used in image processing. It analyses the correlations between various payload features and uses Mahalanobis Distance Map (MDM) to calculate the difference between normal and abnormal network traffic. We focus on the detection of generic attacks, shell code attacks, polymorphic attacks and polymorphic blending attacks. We evaluate accuracy of GSAD model experimentally on the real-world attacks dataset created at Georgia Institute of Technology. We conducted preliminary experiments on the DARPA 99 dataset to evaluate the accuracy of feature reduction.