In the past few years, mobile devices with touch screen features have been used by more and more users. In the year 2013, over 1 billion smartphones had been sold. Smartphone users range from teenagers to presidents, from civilian to military personnel. These devices contain lots of private, personal, and business information, and sometimes are used to remotely access information critical to national security. However, these devices have no physical security protections like the traditional PC and workstation have. As such, the security of these mobile devices poses greater challenge than traditional office equipment. For example, a stolen unlocked smartphone could potentially leak all critical information stored on the phone and some other remotely accessible data.

On the bright side, these mobile devices are equipped with many sensing modalities not available on traditional computing equipment. Sensors suite on a modern device can include: touch screen, accelerometer, gyroscope, magnetometer, camera, finger scanner, microphone, GPS, proximity sensorhear rate sensor, gesture sensor, barometer, etc. Among them, touch screen sensors, accelerometer, and gyroscope can be used to strengthen keystroke dynamics authentication on mobile devices [21, 31]. The touch sensor can sense not only the event of touching, but also the size and pressure of touching. As such, mobile devices have great potential to achieve high security if proper technologies are developed to exploit these rich sensing modalities. To meet the new challenges for mobile keystroke dynamics biometrics, we present a deep learning approach [12], which is a very powerful advanced machine leaning method. Since its introduction, it has dominated the speech analysis field, drastically improving performance records on many tough problems that have been kept unsolved for years. It is starting to take over other challenging research fields as well, eg, face recognition [27]. We have previously applied deep learning method to static keystroke dynamics biometrics authentication for desktop computers, where it has demonstrated superior performance compared to the state-of-the-art. In this chapter we investigate the deep learning approach to keystroke dynamics authentication on mobile devices [6], and explore additional sensory data available on mobile devices for augmented keystroke dynamics biometrics. We will evaluate our approach on mobile keystroke dynamics dataset and compare it with the state-of-the-art. The rest of this chapter is organized as follows. Section 4.2 gives a review of mobile keystroke biometrics. Section 4.3 introduces deep learning approach to keystroke dynamics authentication for mobile devices. Section 4.4 describes user verification experiments and performance of the proposed algorithms on public data set. We draw conclusions and layout future work in Section 4.5.