SVM-based Novelty Detection Approach for Password Typing

摘要

As the most widely used identify verification method, password is vulnerable to be attacked by imposters. Some researchers have found that password typing patterns of each individual were different from others. These patterns include some features such as user's manner and rhythm etc. When a user types the password, the keystroke pattern can be characterized by a timing vector, including the durations of keystrokes and the intervals between them. The owner's timing vectors are acquired to build a model that discriminates between the owner and imposters. This paper introduces an approach for password typing novelty detection based on support vector machines, which find a global minimum of the actual risk upper bound using structural risk minimization. The results of experiments demonstrate that the approach can discriminate the owner and imposters effectively.