Research Project

Overview

The research project is a major component of this course. The project requires the students to execute research in the area of security and privacy of machine learning. Please be realistic about what can be accomplished in a single term. Start planning early since projects executed in the closing days of the term are unlikely to be well received. The grade will be based on the following factors: novelty, depth, clarity of presentation, and effort. As an example, you can first implement an existing privacy attack and then come up with a new attack or propose a novel solution. Projects will be conducted in groups of 2 people (please contact me for possible exceptions). The projects will be due in several stages as detailed below.


Important Dates (all deadlines are 11:59:59 pm)

September 24: Project group information due

October 1: Project proposal due (meet the instructor at office hours or by appointment in the week of January 18-22 to discuss the proposal if needed)

October 14: Midterm project presentation

November 9/11: Final project presentation

November 18: Project report due


Project Proposal

Maximum 3 pages. Please use this ACM template on Overleaf for project related submissions.

What: A name for the project. A clear description of your goal (e.g., proposing a new attack, proposing a new defense, improving an existing attack/defense, etc.).

Why: Why should we be interested in the result of your project? A detailed view of motivation and most related work.

How: What resources will you use? Is there an existing code-base of any related work? How will you evaluate your results?


Midterm Project Presentations

In general, the presentation should consist of the following components (and other components where applicable):

(1) Problem and motivation

(2) Detailed related work

(3) Threat model

(4) Demonstration/result of existing work which you plan to extend (if applicable)

(5) The next steps of your project



Final Project Presentations

The presentation slides should include the following components:

  1. Problem motivation (why is the problem important? use examples)

  2. Background (concepts required to understand the technical contribution of your work)

  3. Related work (what is the state-of-the-art?)

  4. Research objective (what is the goal of the work?)

  5. Threat model

  6. Challenges you had to overcome

  7. Methodology

  8. Dataset description

  9. Results

  10. Limitations

  11. Future work

  12. Share your most important learning(s) from this project

If you are presenting a PowerPoint/Keynote file, please upload the original file, i.e., not the pdf.



Final Project Report

  • Template: All project reports should be submitted using this ACM template on Overleaf.

  • Organization: If you are working towards extending an existing research paper, that particular paper's organization might be helpful. The following is a general list of sections for all the reports. Feel free to add more sections as required by the project.

      1. Introduction

        • Problem motivation

        • Challenges

        • Main contributions

      2. Preliminaries

      3. Related work

      4. Threat model

      5. Methodology

      6. Evaluation

        • Dataset description

        • Experiment setup

        • Evaluation metrics

        • Experiment results

        • Comparison with existing work (if applicable)

      7. Limitations (& any discussion if applicable)

      8. Future work

      9. Conclusion

  • Report length:

    • At least 12 pages not including references.

  • Publication and authorship: Our aim is to write an excellent report from this course that we can eventually submit as a paper for review and potential publication. Given that this is a class project, the instructor will serve as a research advisor for every project (individual/team) and as such will be listed as the last author on any papers submitted from this course.