In recent years, computing devices have become widely distributed, and the accumulated data from these devices are growing rapidly, especially as they are increasingly equipped with various sensors and RF communication capabilities. Data science, including machine learning technology, has contributed to the better handling the large amounts of data and feature selection techniques have been a useful strategy. As data amounts continue to grow, scaling features will became crucial in data science. In this paper, we propose a novel filter-based feature-selection method in the context of keystroke dynamics authentication. In particular, we propose a new feature-scoring method and apply it to keystroke-dynamics-based authentications. We implement keystroke-dynamics-based authentications multi-factored with PIN-based authentications and collect data from actual users' testing experiments. Then, we apply our feature-selection method and compare the performance with that when using all of the features and existing feature-selection methods. Our experimental results show that the classification performance by the proposed method is superior to those of the other methods by up to 21.8%. Moreover, our method provides security to other users' data sets, as the method utilizes only mean values from imposter data. Our feature-selection method contributes to improving the quality of keystroke dynamics authentications without user privacy issues. More generally, our method can also be applied to other data-mining data sets, such as IoT sensor data sets.

Personal Identification Numbers (PINs) and pattern drawing have been used as common authentication methods especially on smartphones. Such methods, however, are very vulnerable to the shoulder surfing attack. Thus, keystroke dynamics that authenticate legitimate users based on their typing manner have been studied for years. However, many of the studies have focused on PC keyboard keystrokes. More studies on mobile and smartphones keystroke dynamics are warranted; as smartphones make progress in both hardware and software, features from smartphones have been diversified. In this paper, using various features including keystroke data such as time interval and motion data such as accelerometers and rotation values, we evaluate features with motion data and without motion data. We also compare 5 formulas for motion data, respectively. We also demonstrate that opposite gender match between a legitimate user and impostors has influence on authenticating by our experiment results.

No abstract

Authentication methods using personal identification number (PIN) and unlock patterns are widely used in smartphone user authentication. However, these authentication methods are vulnerable to shoulder-surfing attacks, and PIN authentication, in particular, is poor in terms of security because PINs are short in length with just four to six digits. A wide range of research is currently underway to examine various biometric authentication methods, for example, using the user’s face, fingerprint, or iris information. However, such authentication methods provide PIN-based authentication as a type of backup authentication to prepare for when the maximum set number of authentication failures is exceeded during the authentication process such that the security of biometric authentication equates to the security of PIN-based authentication. In order to overcome this limitation, research has been conducted on keystroke dynamics-based authentication, where users are classified by analyzing their typing patterns while they are entering their PIN. As a result, a wide range of methods for improving the ability to distinguish the normal user from abnormal ones have been proposed, using the typing patterns captured during the user’s PIN input. In this paper, we propose unique keypads that are assigned to and used by only normal users of smartphones to improve the user classification performance capabilities of existing keypads. The proposed keypads are formed by randomly generated numbers based on the Mersenne Twister algorithm. In an attempt to demonstrate the superior classification performance of the proposed unique keypad compared to existing keypads, all tests except for the keypad type were conducted under the same conditions in earlier work, including collection-related features and feature selection methods. Our experimental results show that when the filtering rates are 10%, 20%, 30%, 40%, and 50%, the corresponding equal error rates (EERs) for the proposed keypads are improved by 4.15%, 3.11%, 2.77%, 3.37% and 3.53% on average compared to the classification performance outcomes in earlier work.

No abstract