KSDSLD — A tool for keystroke dynamics synthesis & liveness detection

Abstract

Keystroke dynamics is a behavioral biometrics modality that employs the characteristic typing patterns of users to verify their identity, generally as a part of a multifactor authentication scheme. Naïve implementations of keystroke dynamics verification are susceptible to presentation attacks with synthesized samples, leveraging partial or complete knowledge of the legitimate users’ typing patterns to forge accurate imitations of their behavior. The companion article has presented several novel methods for the synthesis of forged keystroke dynamics samples given one or more authentic samples of free text by the legitimate user, and a liveness detection method that employs the latter as adversaries to train a classification model, which can distinguish human-written samples from synthetic forgeries. Here, we provide the implementation and its source code.