Olya Ohrimenko

Olya helped develop privacy-preserving collaborative learning methods that address information leaks through software’s data-access patterns. At Microsoft Research, this work formed part of a confidential-computing group that contributed to Azure Confidential Computing and the Confidential Consortium Framework. For engineers building cloud and AI systems, the research addresses a practical security requirement: keeping private information protected while computation is taking place.

Olya Ohrimenko is a Professor at the University of Melbourne, researching how to protect information while software processes it. After completing a PhD at Brown University, Olya spent six years at Microsoft Research in Cambridge, UK, rising to principal researcher and leading work on confidential AI and data-access side-channel mitigation, before moving to Melbourne in 2020.

To address privacy, security and integrity in machine learning and cloud platforms, Olya pairs implemented systems with mathematical reasoning about security. Collaborative learning methods developed through this research allow multiple parties to train machine-learning models together without leaking private information through data-access patterns. Because access patterns can expose secrets even when underlying data remains protected, this approach establishes that secure software must safeguard information throughout active computation as well as in storage.

Current projects include leading a 2025 Australian Office of National Intelligence grant on securing data flows in AI systems. Olya also investigates verifiable machine unlearning, developing checks to ensure that a trained model has removed the influence of specified training data.