SCALE-AI Chair in Data-Driven Supply Chains,
Professor at the Department of Mathematics and Industrial Engineering (MAGI) of Polytechnique Montréal, Canada
Abstract:
The deployment of machine learning models in high-stakes domains (e.g., finance, medicine) raises profound questions about the privacy of the data used to train them. In this talk, I will present a broader perspective on how operations research can provide a rigorous methodological backbone for analyzing, quantifying, and ultimately mitigating privacy risks in modern ML pipelines.
I will first discuss a white-box reconstruction attack that formulates the recovery of a random forest training data as a maximum-likelihood combinatorial problem solved with constraint programming. Remarkably, this approach can often reconstruct entire datasets, even from forests with only a few trees.
Next, we turn to black-box access and explainability-driven interfaces. Counterfactual explanations (now increasingly needed and exposed through ML APIs) represent a powerful attack surface.
Using tools from online optimization and competitive analysis, we derive tight bounds on the number of counterfactual queries required to exactly extract tree-based models and introduce new algorithms achieving provably perfect fidelity.
Finally, I will examine the protection offered by differential privacy. Focusing on ε-DP random forests, we demonstrate that even models satisfying strict DP guarantees can still leak meaningful, dataset-specific information in practice, unless the privacy noise is increased to the point where the model loses most of its predictive value.
Together, the talk highlights critical tensions between predictive performance, explainability, and privacy protection, and showcases OR-based techniques as powerful instruments for navigating these trade-offs.
Register at here to receive zoom link
*****************************
We also invite you to our upcoming seminars:
************************************************
Optimization Over Trained Neural Networks: What, Why, and How?
Speaker: Prof Thiago Serra Azevedo Silva, Assistant Professor of Business Analytics, Tippie College of Business, University of Iowa, USA
Beyond MILP: A Hybrid Approach to Large-Scale Real-World Optimization with Hexaly
Speaker: Dr Fred Gardi, Founder & CEO of Hexaly, France
Machine Learning for Faster Matheuristics: Perspectives and Advances
Speaker: Prof Emma Frejinger, Université de Montréal, Canada
For more information, visit: https://euroorml.euro-online.org/