Original / Google Cloud launches Confidential VMs

exist

Translation / Zhou Yuhan

existGoogle Cloud Next '20[1] At the conference, Google Cloud (Google Cloud) launched a "confidential virtual machine" (Confidential VMs). This new type of virtual machine can use Google's encrypted computing to achieve the privacy of data at rest and in memory.

At a press conference on July 14, Google said,

To ensure ourMulti-tenant Architecture[2], we have applied a series of isolation and sandbox technologies in the cloud infrastructure, which will raise the memory encryption security of confidential virtual machines to a new level. In this way, Google Cloud will further isolate the user's workload when using the virtual machine system while protecting the user's sensitive data. We think this feature is useful for thoseRegulated Industries[3] is very valuable.

On the back end, the Confidential VM uses an AMD-based second-generationEPYC Processor(EPYC) [4]'s secure encryption virtualization technology realizes the encryption of data in use. The key is generated by the CPU TEE and cannot be exported, even Google itself cannot know the key.

In just a few easy steps, developers can swap out existing virtual machines for confidential ones. In addition, Google mentioned that they have launched a hardening service project for virtual machines Shielded VMs[5], can protect the system under root software and other vulnerability attacks.

Raghu Nambiar, vice president of AMD Data Center Ecosystem, said: "AMD 2nd Generation EPYC processors (EPYC) have built-in secure encryption virtualization technology, a hardware-based security innovation that helps protect data in virtual environments. In the latest version of the N2D series of Google Compute Engine confidential virtual machines jointly developed by Google, we hope to improve operating efficiency while protecting user privacy."

The last sentence is important, because in traditional solutions, encryption and decryption techniques often mean sacrificing performance (even if only a small part). But Google said that a new open-source driver they developed in partnership with AMD will enable "confidential virtual machineperformance[6] It can be similar to a conventional virtual machine, and the performance loss is almost negligible. According to the information that has been released so far, the startup time and memory read and write performance of Google's confidential virtual machine are no different from those of general virtual machines.

Based on the judgment that TEE technology will flourish on multiple platforms, the Phala protocol abstracted the SGX or SEV protocol at the early stage of design, and continued to track the progress from the first generation of SEV protocol. Currently, Google Cloud Platform's confidence in AMD is based on the SEV second-generation and third-generation standards, and it is expected that there will be a breakthrough in the commercial availability of the third-generation SEV.Therefore, the Phala protocol will be compatible with the DEV-SNP standard, and will support the TEE protocol of AMD chips from 2020. At that time, the privacy computing cloud supporting multiple TEE standards will become a hot topic in the field of next-generation computing.