Trustworthy and Private AI/ML
Jaideep Vaidya, Rutgers Business School, United States
Privacy Risks in Machine Learning: Truths and Myths
Josep Domingo-Ferrer, Rovira i Virgili University, Spain
Trustworthy and Private AI/ML
Jaideep Vaidya
Rutgers Business School
United States
Brief Bio
Jaideep Vaidya is a Distinguished Professor of Computer Information Systems at Rutgers University and the Director of the Rutgers Institute for Data Science, Learning, and Applications. His research focuses on the intersection of privacy, security, data mining, data management, and artificial intelligence, with a strong emphasis on real-world applications and interdisciplinary impact. He has authored over 200 peer-reviewed publications and received best paper awards across leading venues in data mining, databases, digital government, cybersecurity, and healthcare informatics. He is a Fellow of the AAAS, ACMI, AIMBE, IAHSI, IEEE, and IFIP, and an ACM Distinguished Scientist. He served as Editor-in-Chief of the IEEE Transactions on Dependable and Secure Computing and is currently the Editor-in-Chief of the ACM Transactions on Internet Technology.
Abstract
In the era of pervasive data collection and AI-driven decision making, ensuring both privacy and trust in machine learning systems is more critical than ever. This talk explores how we can build AI/ML systems that are not only effective but also respectful of individual privacy and accountable in their decision-making. Drawing on a wide range of real-world applications – from pandemic surveillance to financial anomaly detection – we delve into recent advances in privacy-preserving techniques, including federated learning, differential privacy, secure multiparty computation, and synthetic data generation. We also introduce the concept of Sensitive Privacy, a novel approach for protecting anomalous records, and discuss how these innovations can be practically implemented to support secure, equitable, and trustworthy AI. By grounding the discussion in real-world systems and interdisciplinary collaboration, we aim to provide a roadmap for building AI/ML systems that are both trustworthy and private.
Privacy Risks in Machine Learning: Truths and Myths
Josep Domingo-Ferrer
Rovira i Virgili University
Spain
http://crises-deim.urv.cat/jdomingo
Brief Bio
Josep Domingo-Ferrer (Fellow, IEEE and Distinguished Scientist, ACM) received BSc-MSc and PhD degrees in computer science (Autonomous University of Barcelona), a BSc-MSc in mathematics (UNED) and an MA in philosophy (U. Paris Nanterre). He is a distinguished full professor of computer science and an ICREA-Acadèmia research professor at Universitat Rovira i Virgili, Tarragona, Catalonia, where he also leads CYBERCAT (Center for Cybersecurity Research of Catalonia). He is currently also affiliated as an invited professor with LAAS-CNRS, Toulouse, France. His research interests include data privacy, data security, trustworthy machine learning, and ethics in IT.
Abstract
The privacy risks inherent to machine learning are mainly connected to potential leakage of sensitive data using to train a model. The basic attack against privacy is a membership disclosure attack (MIA) that can be used as a building block to mount more sophisticated attacks such as attribute disclosure or reconstruction attacks. In this talk, I will analyze the chances that a MIA yields unambiguous disclosure in the real world. I will also focus on the particular case of MIAs used to attack the right to be forgotten pursued in machine unlearning.