Keystroke dynamics, a biometric template that identifies individuals based on their typing, has been tempting researchers as an easy way to verify a person: All that’s necessary for verification is a keyboard. But the technology has also been facing obstacles as a person’s typing may change over time.
A new paper published by IEEE Access, a journal run by the Institute of Electrical and Electronics Engineers, proposes a new framework to benchmark biometric verification based on keystroke dynamics.
Similar to how voice, signature, giant and touch gestures recognize behavioral biometric traits, keystroke dynamics, also called keystroke recognition, uses a unique biometric template to identify individuals based on typing pattern, rhythm and speed.
Compared to face or fingerprint recognition, however, behavioral biometrics can be challenging since it has to take into account that people change their style of typing with time and place. Progress in the field of keystroke recognition has also been slow because of the limited size of databases and the use of different protocols and metrics by researchers.
The new experimental benchmarking framework could be a solution to these problems, the researchers write.
The benchmarking framework is based on publicly available Aalto Keystroke Databases. It was devised from different tweet-long texts from over 185,000 people typed out both on desktop and mobile keyboards. The framework runs on CodaLab in the form of the Keystroke Verification Challenge (KVC).
The researchers claim that the benchmarking process not only allows for assessing verification but also offers insights related to biometric fairness and bias based on the comparison scores. The paper introduces a new fairness metric called the Skewed Impostor Ratio (SIR), created to capture bias patterns within and outside demographic groups in the verification score.
The research titled Keystroke Verification Challenge (KVC): Biometric and Fairness Benchmark Evaluation was authored by a team of researchers from the Biometrics and Data Pattern Analytics (BiDA) Laboratory in Spain and the Fraunhofer Institute for Computer Graphics Research in Germany.
Keystroke research has also seen other breakthroughs in recent months. A team of scientists from the Brigham Young University Marriott School of Business in Utah, U.S. have developed a fraud detection keystroke tracking system that correctly identifies when a user types in someone else’s information with 95.5 percent accuracy.