Margit Antal

An Evaluation of One-Class and Two-Class Classification Algorithms for Keystroke Dynamics Authentication on Mobile Devices

In this paper we study keystroke dynamics as an authentication mechanism for touch screen based devices. The authentication process decides whether the identity of a given person is accepted or rejected. This can be easily implemented by using a two-class classifier which operates with the help of positive samples (belonging to the authentic person) and negative ones. However, collecting negative samples is not always a viable option. In such cases a one-class classification algorithm can be used to characterize the target class and distinguish it from the outliers. We implemented an authentication test-framework that is capable of working with both one-class and two-class classification algorithms. The framework was evaluated on our dataset containing keystroke samples from 42 users, collected from touch screen-based Android devices. Experimental results yield an Equal Error Rate (EER) of 3% (two-class) and 7% (one-class) respectively.

Information revealed from scrolling interactions on mobile devices

The information content of touchscreen gestures are analyzed.Besides user identity, gender and touchscreen experience level of the user can be revealed from gestures.Sequences of 10 strokes are appropriate for high accuracy user identity, gender and touchscreen experience level recognition. The aim of this study is to analyze information that can be revealed from simple touch gestures such as horizontal and vertical scrolling. Touch gestures contain identity information, they can reflect the users experience using touchscreen and they can infer the gender of the user. The statements are based on measurements on a large touch dataset collected from 71 users using 8 different mobile devices, both tablets and phones. Touch data were divided in strokes and classification measurements were investigated based on single and multiple strokes. Classification results based on single stroke are inaccurate, which can be improved by using multiple strokes. Measurements prove that identity, gender and users touchscreen experience level can be accurately predicted from a sequence of 10 strokes. In addition to the different classification results we present statistical analysis of the collected data in order to reveal basic differences between male and female users as well as for less and more experienced touchscreen users.

Gender recognition from mobile biometric data

This paper investigates gender recognition from keystroke dynamics data and from touchscreen swipes. Classification measurements were performed using 10-fold cross-validation and leave-one-user-out cross-validation (LOUOCV). We show that when the target is unseen user data classification, only the second approach is viable. Based on our limited datasets, we show that gender cannot be reliably predicted. The best results were 64.76% for the keystroke dataset and 57.16% for the swipes dataset. However, the classification accuracy is over 80% for more than half of the users in the case of keystroke dynamics dataset.

On-line verification of finger drawn signatures

The current proliferation of mobile devices raises an important issue: that of intelligent authentication mechanisms based on sensors within handheld devices. On-line signatures can now be easily captured with the fingertip on mobile devices. We study the performance of on-line signature authentication systems on touchscreen-based mobile devices. Performance evaluations were carried out on the DooDB database, which contains both doodles and pseudo-signatures from 100 users. Besides the results of the function-based methods presented by the database creators, we also evaluated this database using feature-based anomaly detectors. Moreover, we present performance evaluations using different Equal Error Rate (EER) estimation methods and show that there are significant differences among the obtained EER values.

The MOBIKEY Keystroke Dynamics Password Database: Benchmark Results

In this paper we study keystroke dynamics as an authentication mechanism for touchscreen based devices. A data collection application was designed and implemented for Android devices in order to collect several types of password. Besides easy and strong passwords we propose a new type of password—logical strong—which is a strong password, but easy to remember due to the logic behind the password’s characters. Three main types of feature were used in the evaluation: time-based, touch-based and accelerometer-based. We propose a novel feature set—secondorder—which is independent of the length of the password. The preliminary results show that the lowest equal error rate (EER) is achieved by the logical strong password, followed by the strong password. The worst performance was achieved by the easy password; suggesting that the strong password is the best choice even in the case of keystroke dynamics based authentication systems.

Online Signature Verification on MOBISIG Finger-Drawn Signature Corpus

We present MOBISIG, a pseudosignature dataset containing finger-drawn signatures from 83 users captured with a capacitive touchscreen-based mobile device. The database was captured in three sessions resulting in 45 genuine signatures and 20 skilled forgeries for each user. The database was evaluated by two state-of-the-art methods: a function-based system using local features and a feature-based system using global features. Two types of equal error rate computations are performed: one using a global threshold and the other using user-specific thresholds. The lowest equal error rate was 0.01% against random forgeries and 5.81% against skilled forgeries using user-specific thresholds that were computed a posteriori. However, these equal error rates were significantly raised to 1.68% (random forgeries case) and 14.31% (skilled forgeries case) using global thresholds. The same evaluation protocol was performed on the DooDB publicly available dataset. Besides verification performance evaluations conducted on the two finger-drawn datasets, we evaluated the quality of the samples and the users of the two datasets using basic quality measures. The results show that finger-drawn signatures can be used by biometric systems with reasonable accuracy.

Finger or Stylus: Their Impact on the Performance of On-line Signature Verification Systems.

Abstract The widespread use of smartphones and the ability of these devices to digitize signatures have made it possible to sign electronic documents in this way. In this paper we compared two on-line signature databases in terms of verification performance: the MCYT containing signatures drawn by stylus pen, and MOBISIG containing finger drawn signatures. Performance evaluations were performed using both local and global systems. In the case of global systems, we evaluated the performance of a novel information theory features set. Little improvement was achieved by this feature set. There were large differences between the two databases in terms of performance. Finger drawn signatures collected by mobile device were proved inferior to signatures collected by digitizing tablet and its stylus.

Some remarks on a set of information theory features used for on-line signature verification

Recently a new set consisting of six information theory features was proposed for on-line signature verification by Rosso, Ospina and Frery. The proposed features were evaluated on the MCYT-100 on-line signature database resulting in the best performance ever measured on that dataset. In this paper we repeat their measurements and show that their result is erroneous. In addition, we evaluate the performance of the same on-line signature verification system using exactly the same number of state-of-the-art features. State-of-the-art features always outperform the information theory related features, regardless of the classification method used.

Predicting User Identity and Personality Traits from Mobile Sensor Data

Several types of information can be revealed from data provided by mobile sensors. In this study touchscreen and accelerometer data was collected from a group of 98 volunteers during filling in the Eysenck Personality Questionnaire on a tablet computer. Subjects performed swipes on the touchscreen in order to answer the questions. Touchscreen swipes have been already used for user authentication. We show that our constrained swipes contain enough user specific information to be utilized for the same task. Moreover, we have studied the predictability of personality traits such as extraversion, and neuroticism from the collected data. Extraversion was found to be the most reliably predictable personality trait.