Date of Award

December 2020

Degree Type


Degree Name

Doctor of Philosophy (PhD)


Electrical Engineering and Computer Science


Vir V. Phoha

Second Advisor

Shobha K. Bhatia


Behavioral Biometrics, Gait, Keystrokes, Swipes, User Authentication

Subject Categories



Biometrics eliminate the need for a person to remember and reproduce complex secretive information or carry additional hardware in order to authenticate oneself. Behavioral biometrics is a branch of biometrics that focuses on using a person’s behavior or way of doing a task as means of authentication. These tasks can be any common, day to day tasks like walking, sleeping, talking, typing and so on. As interactions with computers and other smart-devices like phones and tablets have become an essential part of modern life, a person’s style of interaction with them can be used as a powerful means of behavioral biometrics.

In this dissertation, we present insights from the analysis of our proposed set of contextsensitive or word-specific keystroke features on desktop, tablet and phone. We show that the conventional features are not highly discriminatory on desktops and are only marginally better on hand-held devices for user identification. By using information of the context, our proposed word-specific features offer superior discrimination among users on all devices. Classifiers, built using our proposed features, perform user identification with high accuracies in range of 90% to 97%, average precision and recall values of 0.914 and 0.901 respectively. Analysis of the word-based impact factors reveal that four or five character words, words with about 50% vowels, and those that are ranked higher on the frequency lists might give better results for the extraction and use of the proposed features for user identification.

We also examine a large umbrella of behavioral biometric data such as; keystroke latencies, gait and swipe data on desktop, phone and tablet for the assumption of an underlying normal distribution, which is common in many research works. Using suitable nonparametric normality tests (Lilliefors test and Shapiro-Wilk test) we show that a majority of the features from all activities and all devices, do not follow a normal distribution. In most cases less than 25% of the samples that were tested had p values > 0.05. We discuss alternate solutions to address the non-normality in behavioral biometric data.

Openly available datasets did not provide the wide range of modalities and activities required for our research. Therefore, we have collected and shared an open access, large benchmark dataset for behavioral biometrics on IEEEDataport. We describe the collection and analysis of our Syracuse University and Assured Information Security - Behavioral Biometrics Multi-device and multi -Activity data from Same users (SU-AIS BB-MAS) Dataset. Which is an open access dataset on IEEEdataport, with data from 117 subjects for typing (both fixed and free text), gait (walking, upstairs and downstairs) and touch on Desktop, Tablet and Phone. The dataset consists a total of about: 3.5 million keystroke events; 57.1 million data-points for accelerometer and gyroscope each; 1.7 million datapoints for swipes and is listed as one of the most popular datasets on the portal (through IEEE emails to all members on 05/13/2020 and 07/21/2020).

We also show that keystroke dynamics (KD) on a desktop can be used to classify the type of activity, either benign or adversarial, that a text sample originates from. We show the inefficiencies of popular temporal features for this task. With our proposed set of 14 features we achieve high accuracies (93% to 97%) and low Type 1 and Type 2 errors (3% to 8%) in classifying text samples of different sizes. We also present exploratory research in (a) authenticating users through musical notes generated by mapping their keystroke latencies to music and (b) authenticating users through the relationship between their keystroke latencies on multiple devices.


Open Access

Included in

Engineering Commons