Foot biometrics encloses both physiological and behavioral characteristics. Yet, despite its great potential, foot biometrics remains underexplored. One reason behind this is that it usually involves the removal of shoes and socks. As thermal cameras become ubiquitous, we foresee a new form of capturing footprints. Thermal cameras allow for capturing thermal footprints or heat traces resulting from stepping. These can be used to reconstruct the footprint and identify users while they still wear their shoes and socks. We propose a novel method that combines thermal and visible features of the foot to seamlessly and unobtrusively identify users. We recorded a dataset of users’ steps with shoes and socks in a user study (N=20). Our proposed system achieves an AUC score of up to 0.98 for shoe footprints and 0.92 for socks footprints using a Logistic Regression on the extracted features. Our findings demonstrate the potential of thermal imaging for unobtrusive user identification and pave the way for novel applications for usable security and seamless identification.


Research Questions and Future Work

Research Question 1: Could a multimodal approach of combining thermal and visual more accurately identify people?

Research Question 2: What is the influence of different footwear on identification?



We propose a novel approach for identifying users based on users' thermal and visual footprint. We collected empirical data from 20 participants while wearing shoes and socks. We evaluated the performance of the built classifiers. Our approach identified users based on thermal and visual features and produced AUC scores up to 0.98.



Dr. Yomna Abdelrahman