Publishing Student Grades while Preserving Individual Information Using Bayesian Networks

Privacy-preserving data publishing is an important problem which exists in research and has become increasingly vital in recent years. We come across situations where a data owner wishes to publish data without revealing private information. A known solution to this problem is differential privacy which is a research topic that implements noise injection using the Laplace distribution and building a Bayesian network to retain the user's privacy. In the field of education, tracking a student's progress by making grades and other student performance available to analysts is important and challenging. We perform a study on differential privacy pertaining to the field of education. Our goal considers less accuracy for low-dimensional data such as a list of grades. We study the relationship between variance, data size and accuracy to achieve differential privacy being applied on real data consisting of student grades.