Privacy Enhancing Technologies

Fall 2024
Instructor: Prof. Florian Tramèr
Contact: florian.tramer@inf.ethz.ch
Lectures: Monday 10:15-12:00, HG E 7
Exercise Sessions: Tuesday 17:15-18:00, CAB G 51, and Thursday 17:15-18:00, CAB H 52
Course catalog

Moodle – Gradescope – Schedule & notes – Homework assignments

Description

Privacy is a fundamental human right! And yet, technological advances (in particular in computer science) can often undermine privacy. In this class we will see how to formalize various notions of privacy and how to build systems that preserve privacy, by combining techniques from cryptography and statistics. The later parts of the course will focus on applications to machine learning.

Assignments

You must use LaTeX to write up your problem set. You must use the provided template to typset your assignment (the provided preamble can be handy for common notations).

You must submit your problem set via Gradescope. Please use the course code provided on Moodle to sign up. Do not forget to add your Student ID to your Gradescope account (the one that looks like 11-111-111 and is on your Legi card). Note that Gradescope requires that the solution to each problem starts on a new page.

Please read the collaboration policy!You may discuss the problem sets with other students and you may work together to come up with solutions to the problems. If you do so, you must list the name of your collaborators on the first page of your submission. Each student must write up their problem set independently.

Problem Set 1 (source, preamble) – due Friday, Oct 18 (out: Sep 23)
Problem Set 2 (source, preamble) – due Friday, Nov 8 (out: Oct 21)
Problem Set 3 (source, preamble) – due Friday, Nov 29 (out: Nov 11)
Problem Set 4 – due Friday, Dec 20 (out: Dec 2)

Schedule

Date	Topic	Optional Readings
Mo, Sep 23	Lecture Notes, Class Notes Logistics Commitment schemes	“The Moral Character of Cryptographic Work” [Rogaway, 2015] Boneh-Shoup (§3.12, §8.10.2) Refresher on complexity theory, cryptography, and probability theory which will be helpful for the course.
Mo, Sep 30	Lecture Notes, Class Notes Secret Sharing Secure Multiparty Computation	“How to Share a Secret” [Shamir, CACM'79] “A Pragmatic Introduction to Secure Multi-Party Computation” [Evans et al., 2018] (Chapters 2 and 3)
Mo, Oct 7	Lecture Notes, Class Notes Zero-Knowledge Proofs Non-interactive and succinct proofs	Boneh-Shoup (§20.1-20.3) Susan Hohenberger’s lecture notes on Fiat-Shamir Vitalik Buterin on “applications of ZK-SNARKs for preserving privacy”
Mo, Oct 14	Lecture Notes, Class Notes Polynomial IOPs The PLONK SNARG	“Interactive Oracle Proofs” [Ben-Sasson et al., TCC'16] Vitalik Buterin’s “explainer on PLONK”
Fr, Oct 18	Problem Set 1 Due at 11:59pm via Gradescope
Mo, Oct 21	Lecture Notes, Class Notes Private Information Retrieval	“A Survey of Single-Database PIR” [Ostrovsky and Skeith, 2007] “Private Information Retrieval with Sublinear Online Time [Corrigan-Gibbs and Kogan, EUROCRYPT'22]” “Apple’s deployment of PIR in production”
Mo, Oct 28	Lecture Notes, Class Notes Oblivious RAM	The original Goldreich-Ostrovsky paper Elaine Shi’s lecture notes
Mo, Nov 4	Lecture Notes, Class Notes Computing Private Statistics Linear PCPs	Henry Corrigan Gibbs’ PhD Defense which describes PRIO. Deployment notes by Mozilla and Apple & Google
Fr, Nov 8	Problem Set 2 Due at 11:59pm via Gradescope
Mo, Nov 11	Lecture Notes, Class Notes Data Privacy Data Reconstruction Attacks	“Linear Program Reconstruction in Practice” [Cohen & Nissim, TPDP'18] Section 8.1 of Dwork and Roth A Tweet thread by John Abowd on reconstruction attacks on the US Census
Mo, Nov 18	Lecture Notes, Class Notes Randomized Response Differential Privacy	“RAPPOR: Randomized Aggregatable Privacy-Preserving Ordinal Response” [Erlingsson et al., CCS'14] Slides on the use of “DP in US 2020 Census” Chapter 1 of Dwork and Roth, “The Promise of Differential Privacy”
Mo, Nov 25	Lecture Notes, Class Notes Laplace Mechanism Approximate Differential Privacy	Damien Desfontaines’ blog on the “Privacy Loss Random Variable” and “Why not differential privacy” Rogaway’s critique of DP in “The Moral Character of Cryptographic Work”
Fr, Nov 29	Problem Set 3 Due at 11:59pm via Gradescope
Mo, Dec 2	Lecture Notes, Class Notes Private learning DP-SGD	Frank McSherry on “Statistical inference considered harmful” Papernot and Goodfellow on “Privacy and machine learning: two unexpected allies?”
Mo, Dec 9	Lecture Notes, Class Notes Membership Inference Attacks Privacy Auditing	“Resolving Individuals Contributing Trace Amounts of DNA to Highly Complex Mixtures Using High-Density SNP Genotyping Microarrays” [Homer et al., PLoS genetics'08.] “Membership Inference Attacks From First Principles” [Carlini et al., IEEE S&P'22] “Debugging Differential Privacy: A Case Study for Privacy Auditing” [Tramèr et al., 2022]
Mo, Dec 16	Lecture Notes, Class Notes Privacy across the Machine Learning pipeline Federated Learning	Apple’s System for “Learning with Privacy at Scale” Blog on “Stealing Part of a Production Language Model” [Carlini et al., ICML'24]
Fr, Dec 20	Problem Set 4 Due at 11:59pm via Gradescope

Feedback

This is the first time we teach this class, so we would love to get feedback on how to improve it! You can use this form to give us anonymous feedback throughout the semester. You can fill it out as often as you want. And we make mistakes! If something looks wrong or impossible, please let us know.

Course Staff

Jie Zhang (Head TA)
Laura Hetz (1st course half on crypto)
Daniel Paleka (1st course half on crypto)
Michael Aerni (2nd course half on DP)
Hana Farid