Distributed Artificial Intelligence for Privacy Preservation and Cyber Resilience in Educational Ecosystems
Abstract
The digitization of education has concentrated vast volumes of sensitive learner data within centralized platforms that are priority targets for cyber attackers. It is argued that the dominant centralized model of educational artificial intelligence is structurally misaligned with modern privacy and resilience needs, and an alternative grounded in Distributed Artificial Intelligence is proposed. A layered architecture is presented that couples Federated Learning with Belief-Desire-Intention extended (BDIx) agents to deliver privacy-preserving analytics and autonomous cyber resilience without centralizing raw data. Learner records remain on institutional premises; model updates are protected by differential privacy and secure aggregation; and collaborative threat detection and self-healing are performed by distributed BDIx agents. BDIx-FL reached 99.17% accuracy, 98.93% macro-F1 and 99.76% macro AUC in simulation. Controversies are examined, deployment recommendations are offered and open directions are identified for educational ecosystems that are trustworthy by design.
