Date: Thursday February 6, 2020

Time:  12:00 pm – 1:00 pm

Location: LT 3105

With the recent advances in Data Analytics and Machine Learning, organizations are becoming more and more interested in utilizing these techniques to generate insights from the data they have. But the biggest hurdle, especially for those organizations who collect private data is that it becomes very difficult to share their data with data analysts without compromising the privacy of the data. In 2006, Cynthia Dwork et al. came up with something called Differential Privacy. What this essentially does is that it makes sure that the answers to the queries of the data analyst does not compromise the privacy of any of the participants of the data set. Even though this seems very promising, but it still poses a lot of burden on data analysts to understand Differential Privacy and manage privacy budgets accordingly. Moreover, this also doesn't give any accuracy guarantees to the data analyst. Keeping this in mind, Chang Ge et al. introduced a novel system called APEx in May 2019 which allows data analysts to run sequence of queries keeping privacy and accuracy in place. Change Ge et al. in their work implemented APEx (Accuracy-Aware Differentially Private Data Exploration) for only one table in the database. As we know that in real-world, relational databases never contain all of the data in only one table as it violates the database normalization concepts. So, in order to test APEx on real-world scenarios, in this research it will be extended and evaluated on multiple table queries. Also, various other extensions to the original version of APEx will be implemented.

Thesis Committee: 

 

MSc Thesis Proposal Announcement