School of Computer Science
Technical Workshop Series: Membership Inference Attacks on Machine Learning Models
Presenter: Ali Abbasi Tadi, Ph.D. Candidate
Date: Friday, November 3rd, 2023
Time: 3:00 PM -4:00PM
Location: 4th Floor (Workshop space) at 300 Ouellette Avenue (School of Computer Science Advanced Computing Hub)
LATECOMERS WILL NOT BE ADMITTED once the presentation has begun.
Abstract:
In today’s world, where machine learning has revolutionized all industries, the concerns about the privacy of ML models have increased as well, one of these concerns is the membership inference attack (MIA). MIA allows an adversary to query a trained machine-learning model to predict whether or not a particular example was contained in the model's training dataset. In this workshop, we detail MIA and figure out the mitigation techniques that are currently in the literature. In this talk, we explore differential privacy, trusted execution environment, homomorphic encryption, and architectural approaches.
Workshop Outline:
Attacks that target machine learning models.
What is MIA? And what are the reasons for that?
How can we mitigate MIA?
Prerequisites:
Neural Networks
Biography:
Ali is pursuing his Ph.D. in computer science at the University of Windsor. His main research interest is security/privacy in machine learning. He has publications on private computing in top-tier conferences and peer-reviewed journals. He has received various scholarships from the University of Windsor and got 5th place in the iDash security 2022 competition. He has been invited as a speaker at the Advanced Computing Hub at the University of Windsor. He is serving on the program committee and the editorial board of top-tier conferences and journals. He is developing a secure framework for highly qualified parallel clustering for sensitive genomic data.