An Artificial Intelligence Model for Real-Time Detection of Online Examination Fraud

Abstract

The rapid adoption of online examination systems has transformed academic assessment by enhancing accessibility and flexibility; however, it has simultaneously introduced significant challenges related to examination integrity, fraud detection, and data security. Traditional online proctoring systems are largely centralized, rule-based, and heavily dependent on human invigilators, making them susceptible to manipulation, limited scalability, and high false-positive rates. Furthermore, existing approaches often treat artificial intelligence-based fraud detection and blockchain-based data security as separate domains, resulting in a critical gap in achieving both real-time detection and tamper-proof evidence management. This study proposes a novel Artificial Intelligence--Blockchain integrated framework for improved real-time detection of online examination fraud. The system is designed using the Object-Oriented Analysis and Design Methodology (OOADM) and follows a multi-layer architecture comprising data acquisition, preprocessing, AI-based analysis, decision-making, and blockchain-based storage. The proposed model employs Convolutional Neural Networks (CNNs) for spatial feature extraction from video and screen data,YOLOv8 object detection, MediaPipe gaze tracking, LSTM for temporal behavior modeling capable of capturing both instantaneous and sequential fraudulent patterns, and VGGish audio analysis, combined into a unified intelligent fraud detection ecosystem. The upgraded multimodal system achieved 98.2% accuracy, 97.5% precision, 98.9% recall, and 98.2% F1-score, substantially outperforming conventional CNN-only architectures. Detected anomalies are processed through a decision engine that assigns confidence scores and classifies events as normal or suspicious. High-risk events are securely recorded on a permissioned blockchain network using cryptographic hashing and smart contracts, ensuring immutability, transparency, and non-repudiation of examination records.