BlockProctor: Multi-Stage AI and Blockchain-Based Online Examination System with Trust-First Enrollment

Abstract

Online education platforms require robust examination systems that balance academic integrity with student privacy. Traditional centralized examination repositories are vulnerable to data tampering, whereas existing proctoring solutions typically employ invasive surveillance practices. This paper presents BlockProctor, a multi-stage architecture integrating AI-driven behavioral monitoring with blockchain-based integrity verification. The system implements a "Trust-First" enrollment workflow requiring administrative approval and biometric verification before examination access. A lightweight browser-based AI engine employs dual-interval monitoring: MediaPipe FaceMesh at 30 FPS for head pose tracking and TinyFaceDetector at 1.5-second intervals for identity verification. The framework detects seven behavioral anomalies, including 3D head orientation deviation, temporal absence patterns, multi-face presence, and impersonation attempts. A progressive trust scoring system quantifies examination integrity across multiple dimensions. All examination data, student submissions, and proctoring logs undergo SHA-256 cryptographic hashing before immutable storage via Ethereum smart contracts. Preliminary validation under controlled conditions achieved 99.8% accuracy for normal sessions and 100% detection for absence, multi-face, and head pose violations. Blockchain transaction confirmation averaged 170ms on the local testnet. The proposed system provides a cost-effective, privacy-preserving solution for digital assessments, eliminating the need for video recording or external data transmission.