%0 Journal Article
%T Keystroke Dynamics User Authentication Based on Gaussian Mixture Model and Deep Belief Nets
%A Yunbin Deng
%A Yu Zhong
%J ISRN Signal Processing
%D 2013
%R 10.1155/2013/565183
%X User authentication using keystroke dynamics offers many advances in the domain of cyber security, including no extra hardware cost, continuous monitoring, and nonintrusiveness. Many algorithms have been proposed in the literature. Here, we introduce two new algorithms to the domain: the Gaussian mixture model with the universal background model (GMM-UBM) and the deep belief nets (DBN). Unlike most existing approaches, which only use genuine users¡¯ data at training time, these two generative model-based approaches leverage data from background users to enhance the model¡¯s discriminative capability without seeing the imposter¡¯s data at training time. These two new algorithms make no assumption about the underlying probability distribution and are fast for training and testing. They can also be extended to free text use cases. Evaluations on the CMU keystroke dynamics benchmark dataset show over 58% reduction in the equal error rate over the best published approaches. 1. Introduction With the ever increasing demand for more secure access control in many of today¡¯s security applications, traditional methods fail to keep up with the challenges because pins, tokens, and passwords are too many to remember. Even carefully crafted user name and password can be hacked, which compromises the system security. On the other hand, biometrics [1¨C5] based on ¡°who¡± the person is or ¡°how¡± the person acts, as compared with what the person has (key) and knows (password), presents a significant security advancement to meet these new challenges. Among them, keystroke dynamics [6] provides a natural choice for secure ¡°password-free¡± computer access with no additional hardware required. Keystroke dynamics refers to the habitual patterns or rhythms an individual exhibits while typing on a keyboard input device. These rhythms and patterns of tapping are idiosyncratic, [7] the same way as handwritings or signatures are, due to their similar governing neurophysiological mechanisms. Back in the 19th century, telegraph operators could recognize each other based on one¡¯s specific tapping style [8]. Recently, it is shown that typing text can be deciphered simply based on the sound of key typing [9]. As such, it is believed that the keystroke dynamics contains enough information to be a good biometrics to ascertain a user at the keyboard. Compared with other biometrics, keystroke biometrics has additional attractiveness for its user-friendliness and nonintrusiveness. Keystroke dynamics data can be collected without a user¡¯s awareness. Continuous authentication is possible using
%U http://www.hindawi.com/journals/isrn.signal.processing/2013/565183/