FedExamGuard: A Privacy-Preserving AI Framework for Intelligent Examination Governance

Abstract

Examination governance in higher education is increasingly challenged by unauthorized examination participation, examination malpractice, inefficient venue allocation, collusion among students, and concerns over student data privacy. Existing examination management systems are predominantly rule-based and provide limited support for intelligent fraud detection, privacy-preserving collaboration, and transparent decision-making. This paper proposes FedExamGuard, a Privacy-Preserving Artificial Intelligence framework for intelligent examination governance. The framework integrates Federated Learning (FL), Multi-Agent Systems (MAS), Explainable Artificial Intelligence (XAI), Differential Privacy (DP), and Graph-Based Optimization to support secure examination eligibility verification, fraud detection, intelligent venue allocation, and anti-collusion seat assignment across higher education institutions. The framework introduces three key components: the Collaborative Fraud Intelligence Network (CFI-Net) for privacy-preserving fraud detection, the XExam-Agent for autonomous examination governance, and the Privacy-Preserving Anti-Collusion Seat Optimization (PACSO) model for intelligent seating arrangements. A simulation-based proof-of-concept evaluation was conducted using a synthetically generated multi-institutional examination dataset that emulates realistic examination processes, including fee payment verification, course registration, examination scheduling, and student behavioural patterns. Simulation results indicate that the proposed framework achieved 97.8% eligibility verification accuracy, 98.4% fraud detection accuracy, 95.6% venue utilization efficiency, and an estimated 80.6% reduction in collusion risk under the defined experimental conditions. The Explainable AI component further enhanced transparency by providing interpretable explanations for automated decisions. The findings demonstrate the feasibility of FedExamGuard as a scalable and privacy-preserving framework for next-generation examination governance and provide a foundation for future validation using real institutional datasets.