A Conceptual Approach for the Cybersecurity of Student Information in Artificial Intelligence-Enhanced Educational Settings

The use of artificial intelligence (AI) in education is rapidly transforming personalized learning and systemic performance. However, the adoption of AI technology into educational settings raises serious cybersecurity issues, especially for Student Information Systems (SIS) that process extremely sensitive behavioral, academic, and personally identifiable student information. This paper proposes a cybersecurity framework combining Federated Adversarial Reinforcement Learning (FARL) with Artificial Neural Networks (ANNs) in a novel way. The ANN will be trained to develop dynamic and adaptive vulnerability scores of actors by examining user rights, their locations, their cybersecurity knowledge, their threat exposure, and their behavior regarding access. The model for training and evaluation involved a synthetic dataset of 9,887 records representing plausible education facing threats and threat types. The FARL framework combines three powerful AI techniques: reinforcement learning ensures the environment is able to continuously adapt, adversarial learning seeks to minimize the model's ability to be manipulated or fooled by malevolent actors, and federated learning ensures data privacy by only processing data locally on user devices. The results of subsequent experimentation indicate that the overall threat detection accuracy of the new FARL-based model was 94.3 %, and the FARL-based security model detection accuracy was better than SIS security systems regarding behavioral anomalous detection and cyber threats. The model also meets all global data protection standard regulations, including FERPA and GDPR regulations, providing a scalable, intelligent, and privacy-preserving lesson in cybersecurity education with AI in educational demonstrative environments.