Enhancing Inclusivity in Keystroke-Based Authentication through Automated Input Detection

Kurzfassung

In the digital age, secure authentication is increasingly important. To enhance the security of existing password authentication systems, various enhancements have been developed, one of which is Keystroke Dynamics (KD) authentication. In KD authentication, next to verifying that the provided password is correct, it is also analyzed how the password was typed. Over time, accurate user profiles of typing characteristics can be created, which adds a behavioral component to classic password authentication, without requiring any hardware other than a keyboard. However, since KD systems often depend on the user manually typing the password, they are inherently incompatible with password automation tools (PATs), such as password managers. When a user does not enter their password manually, KD authentication can not work properly. To address this limitation, this thesis proposes an alternative KD authentication flow, which utilizes a classifier able to identify PAT usage. The created classifier had a low false positive rate of 0.38%, while all PATs it was tested with were correctly classified. Using this alternative KD flow in combination with the classifier, KD authentication is more inclusive towards users who regularly use PATs.