CS291A (Fall 2021) Introduction to Differential Privacy: Theory, Algorithms and Applications

Syllabus [ link ]

Instructor: Prof. Yu-Xiang Wang

Lecture Section: Monday/Wednesday 1:00-2:40 pm Location: HFH 1132 (also on Zoom, link will be sent to you via email.)

Piazza: https://piazza.com/ucsb/fall2021/cs291/home
Piazza is our main channel of communication. Questions should be posted here.

Gradescope: https://www.gradescope.com/courses/318956
This is where you submit your homeworks and project reports.

Office hours: Instructor: by appointment.

Course evaluation: 45% Homework, 40% Project, 5% for attendance / Participation. 10% for scribing.

Scribing: Please volunteer here, use this latex template


Course Schedule / Scribed Notes

127-Sep Course overview and Privacy challenges [Slides,Annotated, scribe] DR Ch 1-2, Vadhan 5.1, DR Ch 8.1 
229-SepDP Basics I: Definition, interpretation + Laplace Mechanism [Slides,Annotated, scribe] DR Ch 2, Ch 3.1, Ch 3.2,
Ch 3.3 (up to Page 34)
34-OctDP Basics II: Sparse Vector + Private Query Release [Slides, Annotated, scribe]DR Ch 3.6, DR Ch 4.2HW1 out
46-OctDP Basics III: Report-Noisy-Max and Exponential Mechanism [Slides, Annotated, scribe] DR Ch 3.3 (Page 34+), DR Ch 3.4 
511-OctDP Basics IV: Privacy loss RV, Advanced Composition [Slides, Annotated] DR Ch 3.5 
613-OctDP Basics V: Gaussian mechanism, zCDP and RDP[Slides, Annotated, scribe] Bun and Steinke (2018),
Balle and W. (2018)
718-OctDP Basics VI: Privacy profiles and Tradeoff functions [slides, Annotated] Dong et al. (2019)Project proposal due
820-Oct DP Basics VII: Privacy accounting and AutoDP[slides, Annotated, Notebook] W., Balle, Kasiviswanathan (2018),
Zhu, Dong and W. (2021),
autodp tutorial
HW1 due
925-OctDPML I: Introduction and Posterior Sampling [slides, annotated, scribe]Minami (2017) HW2 out, coding question
1027-OctDPML II: Objective Perturbation [slides, annotated, scribe] Chaudhuri et al. , Kifer et al. 
111-NovDPML III: Noisy Gradient Descent [slides, annotated] Bassily et al.
123-NovDPML IV: NoisyGD (Part 2) and NoisySGD [slides, annotated, scribe] Bassily et al.  
138-NovDPML V: NoisySGD (Part 2) and Private Deep Learning [slides, annotated, scribe] Abadi et al , autodp tutorialMidterm report due
1410-NovAdaptive DP I: Smoothed Sensitivity and Median [slides, annotated]Vadhan Ch 3.1.,
Nissim, Raskhodnikova, Smith (2011)
1515-NovAdaptive DP II: Propose-Test-Release [slides, annotated, scribe]Vadhan Ch 3.2HW2 due
1617-NovAdaptive DP III: Generalizing PTR and Data-Adaptive DPML [slides, annotated] W. 2018HW3 out
1722-NovProject consultation I   
1824-NovProject consultation II   
1929-NovIndustry Deployment of DP by Ryan Rogers [Slides]  
2030-NovFederated Learning with DP by Peter Kairouz [Slides]   
211-Dec Mini-Symposium on Practical Differential Privacy Final project report due / HW3 due