User Profiling Analysis based Anomaly Detection System
We propose an anomaly detection system based on user profiling analysis to promote its performance. It presents a method to extract user's behavior by carefully choosing features of the user's data. It improves performance of Principle Component Analysis (PCA) algorithm by sampling and online PCA methods. In this way it fits the demands of anomaly detection systems and thus improves the efficiency and stability to achieve a better user profiling. In addition, we apply our model into applications of databases and web browsers. Using data mining methods we successfully deduce the parameters in our model and then deploy it in real environment.