Template protection for biometric gait data

Abstract

Biometric gait recognition is a well suited method for authentication on mobile devices as it is unobtrusive and concurrent. Hence, in contrast to PIN authentication it is no extra-effort for the user. The characteristic gait of a subject can be recorded using accelerometers which are nowadays already contained in many mobile devices. From this data biometric feature vectors can be extracted and stored as reference data on the device. Only if the user is not recognized by his walk an active authentication via PIN is necessary. As the number of attacks on mobile devices increases it cannot be assumed that the data stored on the device is under constant control of the subject. Therefore, template protection techniques should be applied to secure biometric data. As biometric gait recognition is a new field of research no specific template protection methods have been developed so far. This paper describes a new method for securing biometric gait features based on histograms and using the earth mover's distance for comparison. The method is tested with gait data of 48 subjects recorded using a mobile phone and the results are compared to the ones obtained without template protection.