You Are as You Type: Investigating the Influence of Timestamp Accuracy on the Robustness of Keystroke Biometrics

Abstract

Keystroke dynamics are behavioral biometric traits that are frequently proposed to be used in novel authentication systems. Keystroke dynamics are based on the analysis of intervals between keystroke events originating from user input and thus directly depend on available timing information. However, modern web browsers limit the accuracy of timestamps to improve security and privacy. This study systematically investigates the impact of limited timestamp accuracies on the performance of keystroke dynamic analysis. By conducting multiple experiments with popular web browsers, we demonstrate that the minor timestamp modifications that mitigate timing side-channel attacks do not interfere with the effectiveness of keystroke dynamics analysis algorithms. Furthermore, they are surprisingly resilient to larger timestamp modifications, which results in a serious threat to users' privacy. This research provides fundamental knowledge enabling researchers, privacy engineers, and browser vendors to study risks in keystroke dynamics-related systems and to develop mitigations against tracking methods that fingerprint users instead of devices or browsers.