Federated Blockchain for Secure AI-Proctored Examination Systems: Architecture, Implementation, and Evaluation

Abstract

Remote examination platforms have experienced exponential growth, yet centralized architectures remain susceptible to data manipulation, unauthorized record alteration, and deficient audit mechanisms. This work introduces a federated, permissioned blockchain framework built upon Hyperledger Fabric, integrated within an AI-driven online examination platform designated as Evalon. The proposed architecture distributes ledger maintenance across multiple authorized institutional peers, recording cryptographic digests of examination lifecycle events—including candidate authentication, session boundaries, proctoring anomalies, and grade finalization—without exposing personally identifiable information on-chain. A Byzantine fault-tolerant ordering service coupled with endorsement policies ensures that no single administrative entity can unilaterally modify committed records. The blockchain substrate operates alongside a microservices backend deployed on serverless cloud infrastructure, facilitating real-time event validation through RESTful APIs and deterministic smart contracts. Complementing the integrity layer, computer vision models perform continuous behavioral analysis, detecting multi-face presence, gaze deviation, and anomalous motion patterns during live sessions. Experimental evaluation across 12,000 simulated examination sessions demonstrates a 99.7% hash verification success rate, sub-second ledger commit latency under concurrent loads of 500 transactions per second, and a 34% reduction in undetected integrity violations compared with conventional centralized logging. The combined framework establishes a tamper-resistant, auditable, and scalable ecosystem suitable for academic, certification, and enterprise assessment deployments.