Unlocking the full value of data is at the aim of every enterprise. Yet keeping sensitive data private and secure enroute to results is no simple task. “Many businesses are afraid to use the cloud for critical tasks because they are worried that their data could leak out,” observes Mark Horvath, a senior director of research at Gartner.
Today’s protection methods, which focus heavily on encryption at rest and in motion, don’t always deliver the end-to-end protection organizations require. In addition, homomorphic encryption, which allows users to perform computations without decrypting it, has traditionally been slow and difficult to deploy. As a result, there’s a growing push to take protection to the chip level.
Trusted execution environments (TEEs) take aim at this challenge. The technology physically separates critical code and data from other parts of the CPU or GPU chip -- and the overall computing environment. The data in a TEE is encrypted and the underlying instruction set cannot be altered. The result is end-to-end encrypted data that cannot be altered -- even when the data resides outside the TEE.
Intel, AMD, ARM, and other chip makers now offer TEEs -- and cloud providers such as Google, Microsoft and AWS are building the technology into their frameworks. “These environments provide certain guarantees about the confidentiality and integrity of computing that is taking place within them,” states Sean Peisert, a senior scientist at Berkeley Lab and an adjunct professor of computer science at the University of California, Davis.
Protection Schemes
The idea of building protection directly into chips isn’t new. Since the mid-2000s, TEE technology has been around in one form or another. In 2010, a standard emerged, and the first business cases began to take shape. For example, Netflix used a TEE to protect high definition content on smartphones and tablets.
Since then, the technology has expanded into high-performance clouds, where it’s increasingly used by businesses, governments, and research institutes that require airtight data security. Among the core areas of use: smart factory environments that rely on an array of systems and data, including sensors, instruments and other devices that are part of extended Internet of Things and Industrial Internet of Things (IIoT) frameworks.
TEE cloud environments are particularly attractive because they secure data across multiple applications -- and they support advanced biometric authentication and digital rights management on mobile devices and the Internet of Things (IoT). They’re also able to store data outside the trusted environment -- typically in a separate processing environment or device that’s required to use tokens or keys to gain access. This high level of security and flexibility is baked into the environment.
Not surprisingly, TEEs continue to evolve. Intel’s Software Guard Extensions (SGX) technology, which was introduced in 2015, has a working limit of about 96 megabytes, which makes it difficult to use for many of today’s applications and data sets, says Jason Lowe-Power, an assistant professor in the computer science department at University of California, Davis. His research, which involved benchmarking the SGX technology, found that this can result in a slowdown ranging from 10x to 100x over conventional methods.
Newer technology, such as AMD’s SEV, ARM’s TrustZone, and Intel’s TDX, incorporate a virtualization layer that breaks free of hardware-only memory limitations. Virtualization make it possible to refactor and reprogram systems, thus making them faster and more flexible. “The fusion of hardware and virtualization techniques is superior to each technology individually,” Gartner’s Horvath notes. “The approaches cloud solution providers are using aren’t fundamentally different in terms of basic architecture, but there are some important differences in both execution and how they are optimized.”
For example, AWS offers a solution called Nitro, which offloads virtualization resources to dedicated hardware and software in order to minimize the attack surface. The solution prohibits administrative access, thus eliminating the possibility of human error and tampering. Other cloud providers, including Microsoft and Google, are part of the Confidential Computing Consortium (CCC), which promotes “confidential computing” through a TEE approach.
“While there’s no way to get to 100% certainty about data protection -- in theory it’s possible there could be a flaw or some type of backdoor embedded in the chip -- the encryption keys and the overall encryption management on these systems have advanced pretty considerably over time,” Horvath says. Today, “It’s about the best protection possible for data-sensitive workloads in the cloud.”
Taking TEE Beyond the Enterprise
Already, TEEs are changing the face of computing across numerous industries and fields. In addition, Nvidia has introduced a more advanced confidential computing framework for GPUs. It could prove transformative for deep learning models and other forms of artificial intelligence. TEE technology also integrates well with Blockchain and other digital frameworks. Says Lowe-Power: “Instead of securing small, rarely-executed compute kernels, we are executing entire large-scale sensitive applications in these emerging enclaves.”
To be sure, TEE’s have utility far beyond an individual enterprise. One of the most appealing features is the ability to share data across organizations and entities without revealing sensitive information. “Where it really shines is in secure multi-party computing environments where the parties benefit by accessing records, but the data cannot be in the clear,” Horvath says. This includes a group of financial services firms studying breach data or healthcare companies looking to decipher epidemiological data that spans organizations.
Meanwhile, researchers such as Berkeley Lab’s Peisert are studying ways to extend the functionality of TEEs through open standard RISC-V processors. This would open the black box of chip and BIOS engineering for close examination -- and make it possible for organizations to develop instruction set extensions and other features that address specific security needs. “This approach would add an additional layer of protection because it would be possible to have clear evidence that the processor hardware is secure,” Peisert says.
Make no mistake, TEEs are changing the face of cloud computing -- and they will have an even bigger impact in the months and years ahead. “We will eventually see a full range of processors that support trusted execution environments -- from cloud systems and high-performance computing to mobile devices and IoT devices on the edge of the network,” Peisert concludes. “This end-to-end protection will fundamentally revamp the way we think about trust and change the way we use data.”
What to Read Next:
Data Strategy: Synthetic Data and Other Tech for AI's Next Phase
Can Data Collection Persist Amid Post-Roe Privacy Questions?
.svg)