2020/08/24 by William W. L. Cheung, Sudip Vhaduri, Cheung, William +1 · 1 citation
Computer Science · Engineering · #Biometric Identification and Security #FOS: Computer and information sciences #FOS: Electrical engineering #Gait Recognition and Analysis #Machine Learning (cs.LG) #Signal Processing (eess.SP) #User Authentication and Security Systems #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2008.10779
openalex publication_date 2020/08/24 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
The security of private information is becoming the bedrock of an\nincreasingly digitized society. While the users are flooded with passwords and\nPINs, these gold-standard explicit authentications are becoming less popular\nand valuable. Recent biometric-based authentication methods, such as facial or\nfinger recognition, are getting popular due to their higher accuracy. However,\nthese hard-biometric-based systems require dedicated devices with powerful\nsensors and authentication models, which are often limited to most of the\nmarket wearables. Still, market wearables are collecting various private\ninformation of a user and are becoming an integral part of life: accessing\ncars, bank accounts, etc. Therefore, time demands a burden-free implicit\nauthentication mechanism for wearables using the less-informative\nsoft-biometric data that are easily obtainable from modern market wearables. In\nthis work, we present a context-dependent soft-biometric-based authentication\nsystem for wearables devices using heart rate, gait, and breathing audio\nsignals. From our detailed analysis using the "leave-one-out" validation, we\nfind that a lighter k-Nearest Neighbor (k-NN) model with k = 2 can obtain\nan average accuracy of 0.93 \± 0.06, F1 score 0.93 \± 0.03, and em\nfalse positive rate (FPR) below 0.08 at 50 % level of confidence, which\nshows the promise of this work.\n