Applying Conditional Random Fields to payload anomaly detection with CRFPAD

Payload anomaly detection is a relatively young branch of intrusion detection focusing on the analysis of packet payloads to identify attacks in the application layer. Previous payload anomaly detection systems have used models such as Support Vector Machines and various Markov models. The model proposed in this paper for analysis is the Conditional Random Fields model. Conditional Random Fields have been successfully applied to anomaly detection methods other than payload anomaly detection. In this paper, a prototype payload anomaly detection system is discussed then tested against a public data set. The results of the testing show that the Conditional Random Fields model is a very powerful model for payload anomaly detection.