Course Overview

Instructor: Shagufta Mehnaz (Shagufta.Mehnaz@Dartmouth.edu)

Location: Zoom (see Canvas for details)

Meeting times: 2A Slot (Tuesdays and Thursdays 2:25 - 4:15 pm)

Office hours: Thursdays 4:30-5:30pm and Fridays 3:30-4:30 pm, Zoom (see Canvas for details).

TA: Ehsanul Kabir (ehsanul.kabir.gr@dartmouth.edu)

TA office hours: Monday 4:30-5:30 pm (In-person) and Wednesday 4:00-5:00 pm (In-person + Zoom). In-person Location: Lab 003 at Sudikoff. Zoom Link: https://dartmouth.zoom.us/j/2543954514

Objective: Today we see applications of machine learning almost everywhere we look - in the domains of autonomous driving, medical diagnosis, fraud detection, etc. While the use of machine learning is increasing in our day-to-day lives, these techniques also pose significant threats to security and data privacy. This course will explore recent academic research at the intersection of machine learning, security, and data privacy that demonstrates the risks adversaries pose to machine learning systems. The research papers explored in this course would cover attacks on machine learning systems as well as defense techniques to mitigate such attacks. At the end of this course, students will:

(1) acquire a solid background on recent developments in the area of security and privacy of machine learning

(2) be able to identify the security and privacy threats by rigorously analyzing systems that leverage machine learning, and finally,

(3) be motivated to conduct research in this emerging area.

Prerequisites: COSC 74 (Machine Learning and Statistical Data Analysis) or instructor's permission.

Course Expectations

Please note that this course will be conducted like a seminar. The teaching method will be a combination of traditional instructor-led lectures and student-led research paper presentations. Every class period will include presentations of 1-2 research papers. Each student is expected to present 2-3 papers throughout the course. The number of presentations per student will be decided based on class size. All students must read the assigned research papers before the class. They will submit summaries (2 pages maximum per paper) of 8 papers of their choice throughout the term (must cover 1 paper from each topic). They are also expected to actively participate in discussions during lectures and presentations, e.g., by asking questions. This is important since participation will also be assessed for grading. Finally, the students are expected to propose and work on a term-long research project, present the findings, and write a project report.