Enabling Learning on Confidential Data
- Data Security and Governance
- Moscone South | Upper Mezzanine | 152
- 35 min
Multiple organizations often wish to aggregate their confidential data and learn from it, but they cannot do so because they cannot share their data with each other. For example, banks wish to train models jointly over their aggregate transaction data to detect money launderers more efficiently because criminals hide their traces across different banks. To address such problems, we developed MC^2 at UC Berkeley, an open-source framework for multi-party confidential computation, on top of Apache Spark. MC^2 enables organizations to share *encrypted data* and perform analytics and machine learning on the encrypted data without any organization or the cloud seeing the data. Our company Opaque brings the MC^2 technology in an easy-to-use form to organizations in the financial, medical, ad tech, and other sectors.