Steps for Security and Privacy Protection in NLP-based Marking Systems

Abstract

This paper provides an overview of the methods and techniques used to ensure the security and privacy protection of Natural Language Processing (NLP) based test scoring systems. NLPs improve the accuracy and efficiency of correction systems. However, these systems process sensitive data such as student responses, which raises security and privacy concerns. We examine the components of such a system and then propose measures such as access controls, homomorphic encryption, firewalls and blockchain mixed together to secure the system. Next, we safeguard privacy through methods such as differential privacy protection, anonymization and pseudonymization of data. In addition, we insist on the integration of a browser monitoring module to detect any cheating during composition. In this article we partly present a system called "GestStudent New Generation" in which we integrate most of the security concepts to secure the whole system and guarantee privacy protection. Finally, we conclude by stressing the importance of continuous evaluation of these security and privacy measures to ensure the trust and reliability of NLP-based examination marking systems.