Authenticating users on mobile devices is particularly challenging because of usability concerns: authentication must be quick and as unobtrusive as possible. Therefore, biometric methods seem well suited for mobile phones. We evaluate both speaker and face recognition methods on off-the-shelf mobile devices concerning their accuracy, suitability for dealing with low-quality recordings, and running with limited resources. Our results show that speaker and face recognition can realistically be used on mobile phones, but that improvements – e.g. in the form of combining multiple methods – are still necessary and subject to future work.