Advancing the Frontier of Privacy-Preserving Technology in Healthcare
Datavant’s mission is to connect the world’s health data, and make it accessible for valuable research and analytics. The traditional approach of aggregating data creates a direct tradeoff between individual privacy and analytical capability. With health data, this isn’t good enough. The stakes are high — data-driven decisions can save lives, just as data breaches can ruin them. The only way we can achieve our mission at scale is by simultaneously increasing the connectivity of health data, and also the privacy of individual health records. This will demand more from our technology.
This morning, we announced Datavant’s collaboration with Boston University (PRNewswire) to pioneer the use of multi-party computing to share insights from health data. Cutting-edge privacy-preserving technologies such as Secure Multiparty Computation (SMC), Homomorphic Encryption, and Differential Privacy enable the utilization of data to be safer than ever before. SMC and homomorphic encryption enable secure programs that can perform computations on data without even requiring access to the underlying data. This fundamentally changes the tradeoff between privacy and analytics. For example, using these protocols, a group of employees can compute their average salary without anyone divulging their own salary, not even to a third-party. If you extend this from salaries to sensitive health conditions, it’s easy to see how transformative this is. Patients can contribute their genetic information to help researchers train a new model, without having to actually divulge their sensitive data to a third-party.
While there have been notable large-scale deployments of these technologies in the past several years, their widespread adoption is still blocked by lack of standardization, lack of mainstream education, technical complexity of implementation, and performance bottlenecks, among other factors. Dozens of open-source SMC software projects have emerged in the past several years, indicating widespread interest. Processing speeds for secure computation have improved by a factor of 1,000,000 since 2011, and we’ve seen major deployments from Apple, Google, and Uber. But success in real-world applications remains difficult, even for these pioneers.
Datavant chose to be a founding member of the Boston University Data Privacy Collaborative to advance the adoption of frontier privacy-preserving technologies in real-world applications. The goals of the Collaborative include:
- Bringing privacy-enhancing technologies such as secure multiparty computation and differential privacy to bear on real-world problems.
- Integrating state-of-the-art cryptographic techniques focused on security and privacy into modern big data workflows.
- Enhancing accessibility and usability of software tools and platforms focused on data privacy.
If we’re successful, in 5 years it should be easy for applications to draw sensitive information from multiple sources in a secure way under the highest privacy standards. We expect the benefits to patients and the healthcare industry to be massive, eliminating the tradeoff between privacy and improved analytics. Let’s build that world!
Primarily authored by Victor Cai. Thanks to David Gold and Bob Borek for feedback on earlier drafts.