Clarkson University students develop keystroke authentication methods, datasets
Three Clarkson University ECE undergraduate researchers — Christopher Murphy, Timothy Law, and Adam Sherwin — have developed methods and datasets for keystrokes authentication, which they have detailed in two research papers.
The research papers, which the students co-authored, have both been peer reviewed and accepted by two renowned international conferences.
The first paper, “Shared Dataset on Natural Human-Computer Interaction to Support Continuous Authentication Research,” by Christopher Murphy, Jiaju Huang (CU PhD student), Daqing Hou, and Stephanie Schuckers, has been presented and included in the proceedings of the IEEE/IAPR International Joint Conference on Biometrics 2017 (IJCB 2017), which was held in early October in Denver, Colorado.
The second paper, “Benchmarking Keystroke Authentication Algorithms,“ by Jiaju Huang, Daqing Hou, Stephanie Schuckers, Timothy Law, and Adam Sherwin, has been accepted by the IEEE 2017 International Workshop on Information Forensics and Security (WIFS 2017), to be held at Inria Rennes, France, in December.
The research group, led by Stephanie Schuckers, Director of Center for Identification Technology Research (CITeR) and Paynter-Krigman Endowed Professor in Engineering Science, and Daqing Hou, Associate Professor and Director of Software Engineering, is designed to improve user authentication.
Conventional one-stop authentication of a computer terminal is done at a user’s initial sign-on.
In comparison, continuous, active authentication safeguards users against intruders hijacking an authenticated terminal or accessing sign-on credentials as a result of identity theft.
In addition, many behavioral biometric methods can provide continuous authentication without needing additional hardware.
For further advancement, researchers require benchmarking existing algorithms against large, shared datasets.
The researchers provide the largest known dataset that includes keystrokes, mouse events and active programs collected using passive logging software to monitor user interactions with the keyboard, mouse and software programs.
Data was collected from 103 users in a completely unconstrained, natural setting, over two-and-a-half years.
The first paper, led by lead author Murphy, documents in detail the data collection protocol, characteristics of the dataset, and its novelty.
The new Clarkson dataset is important for reproducible research that could lead to improved methods for user authentication, which is a key contributor to the long-term progress in cybersecurity research.
The second paper, co-authored by ECE majors, Law and Sherwin, benchmarks three algorithms against four public datasets, including the new dataset from Clarkson University.
The research shows that the new Clarkson dataset presents new quality factors that challenge the advanced algorithms that are benchmarked.
Funded by NSF Award CNS-1314792, the project was conducted through close collaboration between the researchers and Clarkson’s Office of Information Technology.