Keystroke Dynamics Authentication using MLP Mixers

Passwords and pin codes are commonly used to secure a wide range of digital applications; however, their vulnerability to security breaches poses significant challenges. With over three billion passwords compromised each year, enhancing traditional security methods has become imperative. Keystroke dynamics, a behavioral biometric trait that identifies individuals based on their typing patterns, has emerged as a promising non-intrusive secondary verification technique. This paper introduces an application of MLP mixers to keystroke dynamics. By integrating recent advancements in free-text keystroke biometric authentication with the streamlined architecture of MLP Mixers, this work achieves competitive accuracy when compared to LSTM and transformer-based models. Additionally, it offers benefits in terms of runtime efficiency and reduced model complexity.