This will be a graduate level seminar-style course introducing differential privacy and some of its applications. Differential privacy is a promising approach to the privacy-preserving release of data: it offers a strong guaranteed bound on the increase in harm that a user incurs as a result of participating in a differentially private data analysis. Several mechanisms and software tools have been developed to ensure differential privacy for a wide range of data analysis tasks, such as combinatorial optimization, machine learning, answering distributed queries, etc. In this class we will focus on some fundamental results about the theory and practice of differential privacy and how to use it in concrete applications. Part of the course will also focus on the use of differential privacy for purposes different from protecting privacy like for instance as a technique to prevent false discoveries in experimental science. The course will consist of readings on advanced topics in differential privacy and applications from the programming language, algorithmic, database, machine learning, security, and systems perspective.
The course is offered as 1, 2 or 3 credits. This will be seminar-style course where each student will select and present an article from the literature on differential privacy and applications. Students are expected to read and comment the presented papers previous to class by using NB. Every student will also be invited to engage on one project and to present the results at the end of the course. Discussion about all the aspects of the course will also take place on Piazza.
The grading will be based on the presentation of one paper, participation in class and in the NB and Piazza discussions, and a final project. Students who are not interested in a project can instead present a second article.
The final grade will be composed as:
Projects can take different forms depending on the interest of each student but all the project must have a research component. Some examples of what would constitute a good project are:
| Date | Topic | Presenter |
|---|---|---|
| 1/28 | Introduction to Differential Privacy - basic definitions and mechanisms Optional reading: Chapter 1 and 2 of The Algorithmic Foundations of Differential Privacy, Dwork and Roth, 2014. |
Marco Gaboardi Class introduction Notes |
| 2/04 | Continuing the Introduction to Differential Privacy - basic definitions and mechanisms | Marco Gaboardi Notes |
| 2/11 | Frank McSherry, Privacy Integrated Queries. SIGMOD09. | Akash Mandole |
| Daniel Kifer and Ashwin Machanavajjhala, No free lunch in data privacy. SIGMOD11. | Syed Mohammed Arshad Zaidi | |
| 2/18 | Cynthia Dwork, Guy N. Rothblum, Salil P. Vadhan, Boosting and Differential Privacy. FOCS10. | Swapnil Sudam Auti |
| Zuhe Zhang, Benjamin Rubinstein, Christos Dimitrakakis, On the differential privacy of Bayesian Inference. AAAI16. | Gian Pietro Farina | |
| 2/25 | Prashanth Mohan, Abhradeep Thakurta, Elaine Shi, Dawn Song, David E. Culler, GUPT: privacy preserving data analysis made easy. SIGMOD12. | Namita Marathe |
| Kamalika Chaudhuri, Daniel Hsu, Shuang Song, The Large Margin Mechanism for Differentially Private Maximization. NIPS14. | Shubham Sharma | |
| 3/3 | Cynthia Dwork, Moni Naor, Toniann Pitassi, Guy N. Rothblum, Differential privacy under continual observation. STOC10. | Keval Bhavesh Goradia |
| Moritz Hardt, Katrina Ligett, Frank McSherry, A Simple and Practical Algorithm for Differentially Private Data Release. NIPS12. | Alizishaan Anwar Hussein Khatri | |
| 3/10 | Arjun Narayan, Ariel Feldman, Antonis Papadimitriou, Andreas Haeberlen, Verifiable differential privacy. Eurosys15. | Pranav Pramod Gadekar |
| Konstantinos Chatzikokolakis, Catuscia Palamidessi, Marco Stronati: Geo-indistinguishability: A Principled Approach to Location Privacy. ICDCIT15 | Dhaval Taunk | |
| 3/17 | No class - Spring Break | |
| 3/24 | Florian Tramer, Zhicong Huang, Jean-Pierre Hubaux, Erman Ayday: Differential Privacy with Bounded Priors: Reconciling Utility and Privacy in Genome-Wide Association Studies. CCS15 | Nalin Kumar |
| Frank McSherry, Kunal Talwar: Mechanism Design via Differential Privacy. FOCS07 | Arti Gupta | |
| 3/31 | No class | |
| 4/7 | Michael Hay, Ashwin Machanavajjhala, Gerome Miklau, Yan Chen, and Dan Zhang: Principled Evaluation of Differentially Private Algorithms using DPBench. SIGMOD16 | Kapil Sharma |
| Myrto Arapinis, Diego Figueira, Marco Gaboardi: Sensitivity of Counting Queries. Draft | Desmond Pinto | |
| 4/14 | Yevgeniy Dodis, Adriana López-Alt, Ilya Mironov, Salil P. Vadhan: Differential Privacy with Imperfect Randomness. CRYPTO 2012: 497-516 | Pranav Pramod Gadekar - Namita Balkrishna Marathe |
| Daniel Kifer and Ashwin Machanavajjhala. A Rigorous and Customizable Framework for Privacy. PODS 2012. | Syed Mohammed Arshad Zaidi | |
| 4/21 | Kamalika Chaudhuri, Staal Vinterbo: Stability-based Validation Procedure for Differentially Private Machine Learning. NIPS 2013. | Swapnil Sudam Auti - Kapil Sharma |
| Jason Reed, Benjamin C. Pierce: Distance makes the types grow stronger: a calculus for differential privacy. ICFP 2010: 157-168 | Gian Pietro Farina | |
| 4/28 | Cyrus Shahabi, Liyue Fan, Luciano Nocera, Li Xiong and Ming Li. Privacy-Preserving Inference of Social Relationships from Location Data: A Vision Paper, ACM SIGSPATIAL, 2015 | Akash Mandole - Nalin Kumar |
| Peter Kairouz, Sewoong Oh, Pramod Viswanath: Secure Multi-party Differential Privacy, NIPS 2015 | Shubham Sharma | |
| 5/5 | Samuel Haney, Ashwin Machanavajjhala, Bolin Ding: Secure Design of Policy-Aware Differentially Private Algorithms, PVLDB 2015 | Arti Gupta |
| Georgios Kellaris, Stavros Papadopoulos, Xiaokui Xiao, Dimitris Papadias: Differentially Private Event Sequences over Infinite Streams. PVLDB 7(12): 1155-1166 (2014) | Keval Bhavesh Goradia | |
| Project Presentation | Dhaval Taunk - Desmond Pinto | |
| Project Presentation | Alizishaan Anwar Hussein Khatri |