Healthcare

Three ways Confidential Computing will Transform the Healthcare Industry

Author: Chris Gough, Global General Manager, Health & Life Sciences, Intel Corporation

Intel

Published in

Intel Tech

6 min readJan 13, 2022

The well-known issue of a global shortage of caregivers is already affecting the healthcare industry, but there are two other problems looming large. The first is ensuring the privacy of the data that is gathered, stored, shared, and analyzed across multiple sites and geographical locations. The second is protecting the intellectual property (IP) of the algorithm owners, as those algorithms are frequently hosted on third-party servers. The problems are exacerbated by the trends: approximately 30% of the world’s digital data is generated by the healthcare industry and capabilities for data analysis are evolving at a breakneck pace.

These data management, analysis, and algorithm IP problems must be solved to “move the needle” on the triple aim of improved health outcomes, better patient experience, and lower cost. Confidential computing may be the solution.

The Patient Data Privacy Problem

The problem with ensuring patient data privacy is that the data we would like to analyze often sits within numerous organizations that span geographic boundaries and have different regulatory requirements for protecting personal health information. Moving this data to a central location for analysis brings an additional set of challenges, such as data replication, de-identification, synchronization, and IP protection that add cost and complexity, preventing many of these projects from getting off the ground.

Ideally, there would be a way to analyze this data where it resides, preserving the privacy of the data and security of the IP, while eliminating the cost and complexity of replicating the source data in a central location. Enter multi-party analytics enabled by confidential computing.

Confidential Computing:

While the industry has had a robust set of security controls to protect data at rest (e.g., encryption) and in transit (e.g., TLS/Transport Layer Security), protection for data in use (i.e., while it is being operated on in memory) has been a gap. Per the Confidential Computing Consortium, confidential computing “…protects data in use by performing computation in a hardware-based Trusted Execution Environment. These secure and isolated environments prevent unauthorized access or modification of applications and data while in use, thereby increasing the security assurances for organizations that manage sensitive and regulated data.”

A pattern that is starting to emerge in the industry is that a Trusted Execution Environment (TEE) is created in an environment controlled/maintained by the owner of a dataset. An application or algorithm created by a solution provider is encrypted (in a container or virtual machine), then sent to the TEE over an encrypted channel for secure execution. In this way, the data steward can’t see the algorithm, the algorithm developer can’t see the data, and only the derived results are returned to the owner of the dataset.

As a founding member of the Confidential Computing Consortium, Intel recognizes the challenges of working with sensitive and regulated data.

Healthcare Transformation:

I’d like to briefly describe three examples of how Confidential Computing is being applied in healthcare today, specifically focusing on examples that enable multi-party analytics:

Example #1: Accelerating Development of Generalizable AI for Healthcare with BeeKeeper AI

The healthcare AI market is hampered by the limited number of generalizable algorithms that have been approved by the FDA. Generalizable AI algorithms require access to primary data raises privacy concerns so it can take from 36 months to 5 years and cost up to $5 million, just to validate one model.[SK1]

To cut the cycle time and secure the algorithm, BeeKeeperAI, Inc. developed an Intel® technology-based confidential computing platform that allows developers to submit their algorithm. It’s then containerized and brought to the data owner’s environment. Once there, it runs on the data in a secure encrypted environment and a confidential report is generated and, importantly, the data and algorithm are destroyed.

Using the system, BeeKeeperAI has already worked to validate three different clinical models: a hemodynamic stability index, a COVID-19 detection tool, and a treatment stratification tool for diabetic retinopathy.

Example #2 Secure Contact Tracing with Leidos

Covid-19 has highlighted the need for a more efficient way to collect and transmit contact-tracing data gathered by public health employees. Doing so securely — using blockchain technology — is the specialty of MicrobeTrace Next.

A collaborative effort between Intel and Leidos, MicrobeTrace Next is a digital public health system that combines mobile data entry with enhanced data visualization through a self-service dashboard.

To ensure no personally identifiable information (PII) is shared, all PII is secured with two blockchain keys. Using this ledger-based encryption scheme, all data access and data movements are fully auditable and traceable, and all transactions are immutable. Also, using role-based security control, accounts tied to non-jurisdiction users will not have the option to be given access to the PII.

The system is deployable at the regional or state level as a service, using existing endpoints, or with hardware bundles for both field and remote call centers. It is also network resilient, in that it can work with spotty connectivity, or store data until a downed connection is restored.

This blockchain platform relies upon the secure compute technology at the core of Intel Xeon Scalable processor platforms.

Example #3: Improve Accuracy of Medical Imaging AI Models in Spanish Hospitals

Another COVID-19 related example of the importance of enabling secure collaboration is centered around Federated Learning, a privacy preserving technique that allows collaborating entities to share a common AI model for medical images diagnosis without sharing PII. However, such an algorithm is only as good as the security, reliability, and performance of the network and hardware upon which it runs.

Last year, Intel and Cisco worked with Capgemini Engineering, Vodaphone Spain, and Gilead to ensure reliable and secure compute and networking to allow three Spanish hospitals — Ramón y Cajal, 12 de Octubre, and Sant Pau — to exchange expert knowledge to enhance COVID-19 diagnosis.

Capgemini Engineering worked on the AI algorithm, Gilead supported the project with a view toward developing better prevention and treatment, Vodaphone Spain supplied the connectivity, and Cisco provided server compute nodes to each hospital. Each node is powered by Intel Xeon Scalable processors with built-in AI acceleration and security features.

Each hospital locally trained its models and sent those to the cloud where a central server aggregated the data to continuously improve the model. Sensitive underlying data or patient information remains inaccessible to all parties, and the model IP is protected, again using Intel security technology.

Summary:

Confidential computing enables privacy-preserving distributed computation that protects the IP of AI algorithm developers without costly and complex centralized approaches that require the replication, synchronization, de-identification, movement, and storage of large data sets. I’m excited to see what future use cases for healthcare, life sciences, and beyond will be unlocked with these capabilities.

For more stories on Intel’s technology innovation, make sure to follow Intel Tech on Medium.

Notices and Disclosures:

Intel technologies may require enabled hardware, software or service activation.

No product or component can be absolutely secure. Your costs and results may vary.

Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.

Intel is committed to respecting human rights and avoiding complicity in human rights abuses. See Intel’s Global Human Rights Principles. Intel’s products and software are intended only to be used in applications that do not cause or contribute to a violation of an internationally recognized human right.