Biometric authentication has become a popular alternative to passwords on mobile devices. However, most implementations do not incorporate any mechanisms to ascertain whether the originally authenticated user is still in control of the mobile device. Thus, the user has to re-scan for any subsequent device access, which may lead to biometric scan fatigue. One solution to this problem is to re-authenticate the user via ancillary surrogates of identity that are likely to be stable and unique in the short term and easier to acquire compared to the primary biometric modality, such as opportunistically captured clothing information. The aim of this paper is to investigate such clothing information as a soft biometrics for short-term mobile user re-authentication. To this aim, we propose a novel method for segmentation and matching of clothing ROI from images captured via front-facing camera of mobile devices, without explicitly requiring the face to be present. Experimental investigations on a large-scale mobile dataset show error rates as low as 2.5% using this method.