Face Recognition-Based Voter Authentication for Online Voting: A Prototype and Security Analysis

Addiga Raj Kumar
Department of Computer Science & Artificial Intelligence Central University of Andhra Pradesh

Mr. D. Ashok
Assistant Professor

Department of Computer Science & Artificial Intelligence Central University of Andhra Pradesh

Abstract—With the increasing prevalence of online and elec-tronic services there is a growing need to develop trustworthy, secure, transparent, and reliable on-line voting systems. Single layer password based authentication is simply not viable in a high assurance electoral system, which calls for biometric authentica-tion to increase security. This paper provides a web-based proto-type application that uses face recognition to authenticate voters for on-line voting. The multi-layered security pipeline included in the system contains, for starters, credential validation, K-Nearest Neighbors (KNN) based face recognition, and finally blink-based liveness detection. A token generation system separates the voter from the cast ballot in order to maintain the anonymity of voters whilst ensuring that no more than one vote is cast per person. A verification tracker and public bulletin board are used for auditing post-vote to prevent vote stuffing, without compromising voter privacy. The system was built using Python/Flask for back end services, HTML/CSS/JavaScript for the front-end, and SQLite as a secure data repository. From experimentation the system achieved roughly 92%accuracy on authentication, with a 4% false acceptance rate (FAR), and an 8% false rejection rate (FRR), under ideal conditions. Insertion of the vote into the database took 0.5 sec. While retrieving vote details took under a second. Security checks showed that the system could not be fooled by a replay or impersonation attack, nor a photo-spoofing attack. This proposed system is scalable, low cost, software-only system capable of building the foundations for a next generation on-line voting system that could potentially include deep learning biometrics, blockchain and/or cloud deployment. Index Terms-Biometric authentication, Face recognition, K Nearest Neighbors, KNN, liveness detection, on-line voting.Index Terms—Biometric authentication, face recognition, K-Nearest Neighbours, KNN, liveness detection, online voting, token-based voting, voter anonymity.