Detecting Age Groups using Keystroke Dynamics

The purpose of this document is to guide the readers of the paper “Distinguishing Child Users from Adults Using Keystroke Dynamics”.

In this page you can find the files and functions necessary for replicating the tests together with the information about the test procedure. The files are in compressed format, so you need suitable software to extract it.

The implementation environment

We developed the test platform in Visual Studio 2008. The test implementations for the algorithms are developed in MATLAB (version 7.6.0) numerical computing environment.

The dataset

We collected a brand new dataset that contains typing samples from adult and child participants using two different typing phrases.

1. Turkish phrase (“Mercan Otu”)

2. Password phrase (“.tie5Roanl”)

In order to promote similar studies, we present the test collection application executable (together with the database). You can also download the full source code for the test application, which is implemented in Visual C# programming language.

We have two data subsets for two typing phrases mentioned above. In our datasets, each row consists of 5 header fields and the feature vector for one typing session of a test participant. The dataset is in the following structure in tab separated format:

User ID

Gender

User Group

Year Of Birth

Session Number

Feature 1

Feature 2

…

Feature N

The description of record is presented as follows:

- User ID: An integer identifier that is unique for each user or participant.

- Gender: 1: male, 2: female

- User Group: g < 20: children, g > 20: adults, g = 20: impostors

- Year Of Birth

- Session Number

- Feature 1: First feature value

- Feature 2: Second feature value

- Feature N: Last feature value

Note that gender and year of birth fields are only informative, and not necessarily used for running the test.

Test scripts

Here are the MATLAB scripts for running the tests, which are mentioned in the study. The scripts include the main files, distance measurement files and some utilities we use.

Running the test procedure

After downloading the source package, to run the test procedure:

1. Open MATLAB environment.

2. Navigate to the source code directory.

3. Run the script file with the name “scr_Test”

The code script runs the tests for all the training algorithms and lists the results.

Contact

You can reach the authors via e-mail for any questions:

bicakci (at) etu.edu.tr

yuzun (at) epdk.org.tr