Analyzing web application behavior to detect malicious requests

Abstract

A system is configured for protecting web applications at a host by analyzing web application behavior to detect malicious client requests. Example embodiments described herein include a proxy configured to handle network traffic between a host and clients. The proxy includes two request classification mechanisms, first a list of known clients, malicious and non-malicious, for identifying known malicious and known non-malicious requests and second a web application firewall for determining a classification for unknown requests (e.g., not originating from a known client). The classification itself may be distributed. The proxy determines whether a request is known non-malicious, known malicious, or unknown. The proxy collects request attributes for the known malicious and known non-malicious requests for the generation of a model based on the attributes of the known requests. The proxy passes the unknown requests to the WAF for determining a classification based on their attributes using the model.