Aniket Kamat Rose Hills
Secure Protocols for Distributed QR Decomposition
Researchers in medicine and law often attempt to answer causal questions such as whether a new drug reduces a particular symptom or whether a specific policy change reduces crime. The gold standard for answering these questions is a randomized controlled trial, which is often conducted across multiple locations to increase the diversity of the sample. However, when trial data contains sensitive personal information, institutional review board (IRB) policies may restrict the electronic transfer of raw datasets between sites.
This project aims to develop an alternative to traditional computational protocols based on information-theoretic security, which provides a privacy guarantee that isn’t reliant on computational hardness. It specifically aims to design a secure protocol for performing a QR decomposition, which is fundamental in causal inference. Many analyses rely on such computations and current alternatives are not as secure or numerically stable, so they can leak information or introduce extra computational errors. I’ll be developing a new protocol as well as software for researchers to easily implement this in their distributed settings.
Message To Sponsor
Thank you for your support! I’m truly excited to be developing software that makes collaborative research more secure and efficient. I’m grateful for the opportunity to explore this field more deeply over the summer.