An improved differential privacy-preserving technique for educational dataset

Large amount of digital educational data is becoming more widely available, which offers an opportunity to gain better insights into the educational system. Sharing this data with data analysts raises serious privacy issues. K-anonymization and $$l$$ l -diversity are popular existing privacy-preserving techniques used by many researchers. The disadvantages of these techniques are that they are vulnerable to background knowledge, homogeneity and similarity attacks. In response to these challenges the Differential privacy has emerged as another popular privacy preserving technique. It allows researchers and scientists to analyse the data efficiently without invading privacy of individual. Unlike other techniques, Differential privacy does not control the access of data analyst. In the proposed work an improved privacy preserving approach is suggested which uses ε-differential privacy for securing the data to maximize the privacy while maintaining the utility of data. In comparison with other existing privacy techniques ε-Differential Privacy does not partition the dataset into different attributes and it is also difficult for the adversary to apply background knowledge attack, homogeneity and similarity attack. The results of our experiments demonstrate that the proposed approach achieves higher privacy while maintaining competitive levels of data utility as compared to existing approaches. An educational dataset is used for experimentation and results are compared with existing work available in literature. Experiments are conducted for different values of ε which facilitate in deciding the appropriate value of ε to maintain proper balance in privacy and utility of data.