CS 795: Topics in Privacy and Anonymity, Fall 2018

Course Overview

Course Description

Content
This course will cover topics in 3 different areas of computer science research. The heaviest will be on differential privacy, which is a rigorous approach for making data available and useful, while still maintaining privacy for the individual data contributors. We will then look at anonymity, which the aim of understanding how users might interact with online data, without unwittingly sacrificing their own information. Finally, we will take a brief look at fairness in algorithms, to gain a deeper appreciation of why privacy is so important.

Objectives
Students will learn how to carefully read and present technical papers. They will learn how to think rigorously about privacy, and will gain a technical understanding of the inherent tradeoff that is faced when providing and using data.

 

Course Requirements

The course is heavily focused on reading, discussing and presenting of research papers. There will be a reading assignment each week. Students are expected to summarize each reading assignment, to post some questions about each reading, and to answer each other's questions on the Piazza discussion board. Additionally, each student will be assigned to present twice during the semester. In the week that they are presenting, they will be expected to lead the discussion from the white board, presenting the details of technical claims and proofs.

Grading
Presentation: 50%
Paper summaries: 25%
Online discussion: 25%

Presentation Requirements Your responsibility as presenter is to help the class understand the main technical claims of the paper, and their proofs. On the one hand, this does not necessarily mean you have to cover every single Lemma along the way: it is OK to ask us to accept certain things without proof. On the other hand, this does mean you may need to step back and provide some background. It is up to you to determine what background material might be necessary to include in your presentation! It is quite likely you will want to do some related reading, and possibly present some prior work. I suggest the following plan for preparing your presentation, though it is not required.

 

Tentative Schedule

The following schedule is tentative, and is likely to change. Please go here for a schedule that is updated weekly to reflect what we actually cover.

Date Topic Links
Aug. 30 Intro to differential privacy The Algorithmic Foundations of Differential Privacy
Sep. 13 (Thurs) Intro to differential privacy The Algorithmic Foundations of Differential Privacy
Sept. 17 Differential Privacy: Strong Composition The Algorithmic Foundations of Differential Privacy
Sept. 27 (Thurs.) Smooth sensitivity Nissim et al., 2007
Oct. 4 (Thurs.) Differential privacy for graphs Kasiviswanathan et al., 2013
Oct. 9 (Tues.) Differential Privacy for ML: deep learning Abadi et al.
Oct. 15 Differentially oblivious algorithms Chan et al., 2017
Oct. 22 Anonymous routing: Atom Atom
Oct. 29 Anonymous routing: Vuvuzela Vuvuzela
Nov. 5 Anonymous routing: Stadium Stadium
Nov. 12 Fairness: overview Berk et al., 2017
Nov. 19 Fairness Through Awareness Dwork et al.
Nov. 26 Fairness: Inherent trade-offs Kleinberg et al.~,2016