AuraVote: A Secure, Decentralized Voting System with AI-Powered Liveness and Identity Verification Using ArcFace and Blockchain

Abstract

Secure and transparent electronic voting remains a major challenge due to vulnerabilities in centralized infrastructures and insufficient voter authentication methods. Existing systems frequently depend on weak biometric checks or static credential verification, making them susceptible to spoofing, identity duplication, and manipulation of centralized databases. AuraVote introduces a next-generation decentralized voting framework that integrates real-time artificial intelligence with blockchain-based immutability to guarantee voter authenticity and system integrity. The proposed model adopts a robust architecture including RetinaFace for face localization, ArcFace for generating highly discriminative 512-dimensional embeddings, and cosine-similarity-based identity verification. Alongside facial verification, the system executes multi-modal liveness detection incorporating blink dynamics, natural micro-movement estimation, and texture-frequency analysis to counter photo, video, and digital screen presentation attacks. A FastAPI-based verification backend processes live video streams from the browser and communicates verification results to a Next.js decentralized application. Verified users cast votes on an Ethereum Proof-of-Authority (PoA) blockchain using MetaMask, ensuring tamper-proof vote recording and cryptographically signed voter participation. This paper details the architecture, AI algorithms, blockchain workflow, system security properties, and implementation strategy, demonstrating how combining modern deep-learning authentication with decentralized ledgers establishes a secure, scalable, and trustworthy e-voting ecosystem.