Cryptographically Verified AI/ML Audits -
Is this company's AI/ML model biased? Are its predictions reliable? Are they using my data responsibly? As AI/ML is deployed in sensitive applications, it is increasingly important to audit models to ensure that they uphold societal values. However, AI/ML service providers almost never release their models or data to other parties for auditing due to intellectual property and data privacy issues. My work aims to address this tension through privacy-preserving cryptographic 'contracts' which can bind service providers' models. These contracts use zero-knowledge proofs and other cryptographic tools to guarantee that (i) the model satisfies an important property such as group fairness, robustness, or differential privacy; (ii) outside parties can view the contract to verify whether the model has the property, but they learn no information about the model parameters or data by doing so. Cryptographic verification is powerful but computationally expensive, especially for larger models. In this talk I will introduce a variety of optimization strategies that I've employed in my research to enable this critical emerging approach to AI/ML regulation.

Bio: Olive Franzese-McLaughlin is a distinguished postdoctoral fellow at the Vector Institute & University of Toronto, where she studies cryptographically verifiable AI/ML regulation. She completed her PhD in applied cryptography at Northwestern and worked as a visiting professor at Reed College while ABD, following an MSci and BA in computational biology at University of Maryland and Reed College respectively. In addition to the Vector Distinguished Postdoctoral Fellowship, Olive was awarded an NSF GRFP, an NCI Cancer Research Training Award, and a Top 10 Research Highlight of 2022-23 recognition from the Alan Turing Institute. She has publications in top conferences in machine learning, computer security, and computational biology. Olive has advised projects with several liberal arts college students, and is looking for motivated research assistants to help investigate cryptographically verified AI/ML.