Browse Source

Course planning.

master
Justin Hsu 3 years ago
parent
commit
f19414f909
  1. 11
      website/docs/assignments/gallery.md
  2. 14
      website/docs/assignments/project.md
  3. 12
      website/docs/index.md
  4. 8
      website/docs/org.md
  5. 74
      website/docs/resources/readings.md
  6. 25
      website/docs/resources/software.md
  7. 11
      website/docs/schedule/deadlines.md
  8. 58
      website/docs/schedule/lectures.md
  9. 2
      website/mkdocs.yml

11
website/docs/assignments/gallery.md

@ -1,12 +1,3 @@
# Final Projects
- Yue Gao and Fayi Zhang. *Theory and Optimization of Homomorphic Encryption*.
- Yinglun Zhu. *Answering Evolving Sets of Queries*.
- Yan Nan and Shimaa Ahmed. *Private Voice Transcription*.
- Samuel Drews. *Verifying Decision Tree Stability*.
- Madeleine Berner and Yaman Yu. *Evaluate Adversarial Machine Learning Attacks
on Chinese Character Recognition*.
- Kyrie Zhou and Meghana Moorthy Bhat. *Detecting Fake News with NLP*.
- Hiba Nassereddine and Junxiong Huang. *Adversarial Machine Learning and Autonomous Vehicles*.
- Zichuan Tian and Arjun Kashyap. *pyDiff: Differential Privacy as a Library*.
- Zhiyi Chen and Yiqin Pan. *Detect Compromised Items from Data with Adversarial Attacks*.
TBA

14
website/docs/assignments/project.md

@ -1,8 +1,9 @@
# Project Details
The goal of the course project is to dive more deeply into a particular topic.
The project can be completed in **groups of two or three**. A good project could
lead to some kind of publication. This project could take different forms:
The project can be completed in **groups of three** (or in rare situations,
groups of two). A good project could lead to some kind of publication. This
project could take different forms:
- **Conceptual**: Develop a new technique, extend an existing method, or explore
a new application
@ -36,8 +37,9 @@ should be clear what remains to be done.
be done, along with reach goals to try if things go well.
Besides the milestones, the main deliverable of the project will be a written
final report, around **15-20 pages** in length. Reports should be written in a
research paper style, covering the following areas in some reasonable order:
final report, around **15-20 pages** in length (in some reasonable format).
Reports should be written in a research paper style, covering the following
areas in some reasonable order:
- **Introduce** the problem and the motivation.
- **Review** background and preliminary material.
@ -46,7 +48,3 @@ research paper style, covering the following areas in some reasonable order:
- **Survey** related work.
At the end of the course, each group will give a brief project presentation.
## Deadlines
See [here](../schedule/deadlines.md).

12
website/docs/index.md

@ -1,12 +1,12 @@
# Welcome to CS 763!
This is a graduate-level course covering advanced topics in security and privacy
in data science. We will focus on three core areas at the current research
frontier: **differential privacy**, **adversarial machine learning**, and
**applied cryptography** in machine learning. We will also cover selected
advanced topics; this year, **algorithmic fairness** and **formal verification**
for data science. This is primarily a project-based course, though there will
also be paper presentations and small homework assignments.
in data science. The field is eclectic, and so is this course. We will start
with three core areas: **differential privacy**, **adversarial machine
learning**, and **applied cryptography** in machine learning. Then, we will
cover two advanced topic areas; this year, **algorithmic fairness** and **formal
verification** for data science. This is primarily a project-based course,
though there will also be paper presentations and small homework assignments.
## Logistics
- **Course**: CS 763, Fall 2019

8
website/docs/org.md

@ -49,10 +49,10 @@ time-consuming and they will not be graded in detail.
### Homeworks
After each of the first three core modules, we will assign a small homework
assignment. These assignments are not weighed heavily---though they will be
graded---but they are mostly for you to check that you have grasped the
material.
There will be three small homework assignments, one for each of the core
modules. You will play with software implementations of the methods we cover in
class. These assignments are not weighted heavily, though they will be lightly
graded; the goal is to give you a chance to write some code.
### Course Project

74
website/docs/resources/readings.md

@ -25,6 +25,29 @@
- Matthew Joseph, Aaron Roth, Jonathan Ullman, and Bo Waggoner.
[*Local Differential Privacy for Evolving Data*](https://arxiv.org/abs/1802.07128).
### Adversarial Machine Learning
- Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, and Rob Fergus.
[*Intriguing Properties of Neural Networks*](https://arxiv.org/pdf/1312.6199.pdf).
ICLR 2014.
- Ian J. Goodfellow, Jonathon Shlens, and Christian Szegedy.
[*Explaining and Harnessing Adversarial Examples*](https://arxiv.org/abs/1412.6572).
ICLR 2015.
- Nicholas Carlini and David Wagner.
[*Towards Evaluating the Robustness of Neural Networks*](https://arxiv.org/pdf/1608.04644.pdf).
S&P 2017.
- Kevin Eykholt, Ivan Evtimov, Earlence Fernandes, Bo Li, Amir Rahmati, Chaowei Xiao, Atul Prakash, Tadayoshi Kohno, and Dawn Song.
[*Robust Physical-World Attacks on Deep Learning Models*](https://arxiv.org/pdf/1707.08945.pdf).
CVPR 2018.
- Nicholas Carlini and David Wagner.
[*Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection Methods*](https://arxiv.org/pdf/1705.07263.pdf).
AISec 2017.
- Jacob Steinhardt, Pang Wei Koh, and Percy Liang.
[*Certified Defenses for Data Poisoning Attacks*](https://arxiv.org/pdf/1706.03691.pdf).
NIPS 2017.
- Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu.
[*Towards Deep Learning Models Resistant to Adversarial Attacks*](https://arxiv.org/pdf/1706.06083.pdf).
ICLR 2018.
### Applied Cryptography
- Benjamin Braun, Ariel J. Feldman, Zuocheng Ren, Srinath Setty, Andrew J. Blumberg, and Michael Walfish.
[*Verifying Computations with State*](https://eprint.iacr.org/2013/356.pdf).
@ -51,7 +74,33 @@
[*Verifiable Differential Privacy*](https://www.cis.upenn.edu/~ahae/papers/verdp-eurosys2015.pdf).
EUROSYS 2015.
### Language-Based Security
### Algorithmic Fairness
- Cynthia Dwork, Moritz Hardt, Toniann Pitassi, Omer Reingold, and Rich Zemel.
[*Fairness through Awarness*](https://arxiv.org/pdf/1104.3913).
ITCS 2012.
- Moritz Hardt, Eric Price, and Nathan Srebro.
[*Equality of Opportunity in Supervised Learning*](https://arxiv.org/pdf/1610.02413).
NIPS 2016.
- Tolga Bolukbasi, Kai-Wei Chang, James Zou, Venkatesh Saligrama, and Adam Kalai.
[*Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings*](https://arxiv.org/pdf/1607.06520).
NIPS 2016.
- Jon Kleinberg, Sendhil Mullainathan, and Manish Raghavan.
[*Inherent Trade-Offs in the Fair Determination of Risk Scores*](https://arxiv.org/pdf/1609.05807).
ITCS 2017.
- Úrsula Hébert-Johnson, Michael P. Kim, Omer Reingold, and Guy N. Rothblum.
[*Multicalibration: Calibration for the (Computationally-Identifiable) Masses*](https://arxiv.org/pdf/1711.08513.pdf).
ICML 2018.
- Michael Kearns, Seth Neel, Aaron Roth, and Zhiwei Steven Wu.
[*Preventing Fairness Gerrymandering: Auditing and Learning for Subgroup Fairness*](https://arxiv.org/pdf/1711.05144).
ICML 2018.
- Alekh Agarwal, Alina Beygelzimer, Miroslav Dudík, John Langford, and Hanna Wallach.
[*A Reductions Approach to Fair Classification*](https://arxiv.org/pdf/1803.02453).
ICML 2019.
- Ben Hutchinson and Margaret Mitchell.
[*50 Years of Test (Un)fairness: Lessons for Machine Learning*](https://arxiv.org/pdf/1811.10104).
FAT\* 2019.
### Programming Languages and Verification
- Martín Abadi and Andrew D. Gordon.
[*A Calculus for Cryptographic Protocols: The Spi Calculus*](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/11/ic99spi.pdf).
Information and Computation, 1999.
@ -83,29 +132,6 @@
[*Verification of a Practical Hardware Security Architecture Through Static Information Flow Analysis*](http://www.cse.psu.edu/~dbz5017/pub/asplos17.pdf).
ASPLOS 2017.
### Adversarial Machine Learning
- Christian Szegedy, Wojciech Zaremba, Ilya Sutskever, Joan Bruna, Dumitru Erhan, Ian Goodfellow, and Rob Fergus.
[*Intriguing Properties of Neural Networks*](https://arxiv.org/pdf/1312.6199.pdf).
ICLR 2014.
- Ian J. Goodfellow, Jonathon Shlens, and Christian Szegedy.
[*Explaining and Harnessing Adversarial Examples*](https://arxiv.org/abs/1412.6572).
ICLR 2015.
- Nicholas Carlini and David Wagner.
[*Towards Evaluating the Robustness of Neural Networks*](https://arxiv.org/pdf/1608.04644.pdf).
S&P 2017.
- Kevin Eykholt, Ivan Evtimov, Earlence Fernandes, Bo Li, Amir Rahmati, Chaowei Xiao, Atul Prakash, Tadayoshi Kohno, and Dawn Song.
[*Robust Physical-World Attacks on Deep Learning Models*](https://arxiv.org/pdf/1707.08945.pdf).
CVPR 2018.
- Nicholas Carlini and David Wagner.
[*Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection Methods*](https://arxiv.org/pdf/1705.07263.pdf).
AISec 2017.
- Jacob Steinhardt, Pang Wei Koh, and Percy Liang.
[*Certified Defenses for Data Poisoning Attacks*](https://arxiv.org/pdf/1706.03691.pdf).
NIPS 2017.
- Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu.
[*Towards Deep Learning Models Resistant to Adversarial Attacks*](https://arxiv.org/pdf/1706.06083.pdf).
ICLR 2018.
# Supplemental Material
- Cynthia Dwork and Aaron Roth.
[*Algorithmic Foundations of Data Privacy*](https://www.cis.upenn.edu/~aaroth/Papers/privacybook.pdf).

25
website/docs/resources/software.md

@ -1,15 +1,22 @@
## Differential Privacy
## Core software
- [TensorFlow](https://www.tensorflow.org/overview): Framework for writing,
training, and testing neural networks. You should at least work through the
[first tutorial](https://www.tensorflow.org/tutorials/keras/basic_classification).
- [TensorFlow Privacy](https://github.com/tensorflow/privacy): Extensions to TF
for differentially-private training.
- [CleverHans](https://github.com/tensorflow/cleverhans): Extensions to TF for
adversarial attacks and defenses on ML models.
- [MPyC](https://github.com/lschoe/mpyc): Python libraries for Secure Multiparty
Computation.
## Other tools
### Differential Privacy
- [DFuzz](https://github.com/ejgallego/dfuzz)
- [HOARe2](https://github.com/ejgallego/HOARe2)
## Cryptography
### Cryptography
- [HELib](https://github.com/shaih/HElib)
- [Obliv-C](https://oblivc.org/)
- [ObliVM](http://oblivm.com/download.html)
## Language-Based Security
- [Jif](https://www.cs.cornell.edu/jif/)
- [FlowCaml](https://opam.ocaml.org/packages/flowcaml/flowcaml.1.07/)
## Adversarial Machine Learning
- [CleverHans](https://github.com/tensorflow/cleverhans)

11
website/docs/schedule/deadlines.md

@ -1,11 +1,12 @@
The first key date is **September 16**. By this date, you should:
The first key date is **September 9**. By this date, you should:
- **Check in** with me briefly.
- **Sign up** to present a paper.
- **Choose** a project topic and form groups. This is not a firm commitment, but
you should have an initial direction.
- **Form project groups** of three.
- **Brainstorm** project topics. Try to come up with **1-2 sentences**
describing your initial direction. This is not a firm commitment---you can
change your topic as you learn more.
## Project Deadlines
- Milestone 1: **October 7**
- Milestone 1: **October 11**
- Milestone 2: **November 8**
- Final writeup and presentation: **December 11** (TBD)

58
website/docs/schedule/lectures.md

@ -3,38 +3,38 @@
Date | Topic | Notes
:----:|-------|:---------:
| <center> <h4> **Differential Privacy** </h4> </center> |
9/4 | [Course welcome](../resources/slides/lecture-welcome.html) <br> **Paper:** Keshav. [*How to Read a Paper*](https://web.stanford.edu/class/ee384m/Handouts/HowtoReadPaper.pdf). |
9/6 | |
9/9 | |
9/11 | |
9/13 | |
9/4 | [Course welcome](../resources/slides/lecture-welcome.html) <br> **Reading:** Keshav. [*How to Read a Paper*](https://web.stanford.edu/class/ee384m/Handouts/HowtoReadPaper.pdf). | HW1 Out
9/6 | Basic private mechanisms <br> **Reading:** AFDP 3.2-4 |
9/9 | Composition and closure properties <br> **Reading:** AFDP 3.5 | Signups
9/11 | What does differential privacy actually mean? <br> **Reading:** McSherry. [Lunchtime for Differential Privacy](https://github.com/frankmcsherry/blog/blob/master/posts/2016-08-16.md) |
9/13 | Paper presentations | HW1 Due
| <center> <h4> **Adversarial Machine Learning** </h4> </center> |
9/16 | |
9/18 | |
9/20 | |
9/23 | |
9/25 | |
9/27 | |
9/16 | Overview and Basic attacks | HW2 Out
9/18 | More attacks |
9/20 | Paper presentations |
9/23 | Defense: Adversarial training |
9/25 | Defense: Certified defenses |
9/27 | Paper presentations | HW2 Due
| <center> <h4> **Applied Cryptography** </h4> </center> |
9/30 | |
10/2 | |
10/4 | |
10/7 | |
10/9 | |
10/11 | |
9/30 | Overview and basic constructions | HW3 Out
10/2 | Secure Multiparty Computation |
10/4 | Paper presentations |
10/7 | Homomorphic Encryption |
10/9 | Oblivious computing and side channels |
10/11 | Paper presentations | HW3 Due <br> MS1 Due
| <center> <h4> **Advanced Topic: Algorithmic Fairness** </h4> </center> |
10/14 | |
10/16 | |
10/18 | |
10/21 | |
10/23 | |
10/25 | |
10/14 | Overview and basic notions |
10/16 | Individual and group fairness |
10/18 | Paper presentations |
10/21 | Repairing fairness |
10/23 | Challenges in defining fairness |
10/25 | Paper presentations |
| <center> <h4> **Advanced Topic: PL and Verification** </h4> </center> |
10/28 | |
10/30 | |
11/1 | |
11/4 | |
11/6 | |
11/8 | |
10/28 | Overview and basic notions |
10/30 | Programming languages for differential privacy |
11/1 | Paper presentations |
11/4 | Probabilistic programming languages |
11/6 | Verifying probabilistic programs |
11/8 | Paper presentations | MS2 Due
| <center> <h4> **No Lectures: Work on Projects** </h4> </center> |
12/11 (TBD) | Project Presentations |

2
website/mkdocs.yml

@ -29,5 +29,5 @@ nav:
- Related Courses: 'resources/related.md'
- Assignments:
- Presentations: 'assignments/presentations.md'
- Project: 'assignments/project.md'
- Projects: 'assignments/project.md'
- Gallery: 'assignments/gallery.md'
Loading…
Cancel
Save