Step Count Print: A Physical Activity-Based Biometric Identifier for User Identification and Authentication

Abstract

Step count is one of the most widely used physical activity data and is easily accessible through smart phones and wearable devices. It records the intensity and happening time of a user’s physical activities, and often reflects a users’ unique way of living. Incorporation of step count into biometric systems may thus offer an opportunity to develop innovative, user-friendly and non-invasive strategies of user identification and authentication. In this paper, we propose Step Count Print (SCP), a physical activity-based novel biometric identifier. Extracted from coarse-grained minute-level physical activity data (step counts), SCP contains features, including user step cadence distribution and average step distribution etc., that reflect an individual’s physical activity behavior. With data collected from 100 users in a five-year long period, we conducted an ablation study to demonstrate the non-redundancy of SCP in user identification and authentication scenarios using commonly used machine learning algorithms. The results show that SCP can achieve a Rank-1 rate of up to 75.0% in user identification scenarios and an average accuracy of 92.3% in user authentication scenarios. In different classification algorithms, the user’s accuracy histogram is drawn to demonstrate the universality of SCP and its effectiveness across a range of scenarios and use cases.