Google Cloud has introduced its latest Confidential VM and Assured Workloads to provide more data encryption and restrict storage locations to a particular location. Highly regulated companies and government agencies are targeted.
Google Cloud has taken important steps to secure customer data and prevent potentially unauthorized access to sensitive workloads. In order to meet this imperative as well as the compliance requirements required in particular in the highly regulated sectors as well as government institutions and their subcontractors, the American giant has raised the curtain on Confidental VM and Assured Workloads for Government. Many solutions are available for encrypting data stored in databases or transiting the Internet between servers and clients. However, the biggest security challenge is to protect information while it is processed at the application level in system memory. This has always been a difficult problem to solve because applications need to decrypt the data in memory before processing operations, making it difficult at this stage to guard against an unauthorized third party having control over the OS or other application. A hole in the racket into which malware specialized in the theft of bank information is thrown via RAM-scraping aimed at point of sale terminals.
To protect against these risks, a means exists such as the encryption of data in memory of the manufacturers of processors who have included dedicated security circuits in their sets of components. Known as trusted execution environments (TEEs), they can be used as secure enclaves thanks to their allocated memory which is isolated from that of the main OS. ARM CPUs are offered with TrustZone, Intel CPUs with software guard extensions (SGX), and AMD models include SEV (secure encrypted virtualization) technology.
Encrypt the entire memory of cloud virtual machines
To provide an additional layer of security, Google Cloud announced Confidential VM, initially available in beta and the first product in the latest Confidential Computing range, designed to support the second generation SEV function for the AMD Epyc architecture. . This technology provides businesses with the ability to encrypt the entire memory of their cloud virtual machines with unique keys generated in the CPU that are not exportable. Google Cloud is not the first provider to offer this level of privacy. This is also the case for Microsoft, which has added secure enclaves for Intel SGX-based processors in Azure since 2018. However, there are significant differences between how Intel SGX and AMD SEV work and in what types of scenarios. ‘attacks they respond.
Intel SGX was designed to allow encryption of the memory used by an application and to remain secure even in the event that the operating system itself is compromised. It is a function where the application initializes an SGX enclave and switches to it as much as necessary to decrypt and process operations on the data. With the benefit of the fact that these data never leave their enclave in an unencrypted state, and conversely the applications that must be modified are able to use this capacity in the first place. For its part, AMD SEV focuses on encrypting the memory of all virtual machines rather than applications. Its purpose is to protect the security of client data inside a VM in the event that the hypervisor or the host operating system itself is compromised. The flip side is that the data is not protected if an attacker manages to gain control over the protected VMs or the host OS and the applications running in it.
No application recompilation on AMD SEV
“The main benefit of starting with SEV for AMD is not having to recompile the application,” said Sunil Potti, vice president and general manager of Google Cloud, to our CSO colleague. “This is the most important concern of our customers to which we had to respond to ensure the mass adoption of Confidental VM. With our technology, you just have to take and move your worlkloads on virtual machines and take advantage of the new security technologies that we will introduce over time to take advantage of them. ”
Google had previously partially responded to the guest OS compromise scenario by using virtual machine images enhanced with integrity encryption verification to prevent kernel-level malware and escalation of privileges. It was part of the catalog of the Google Cloud offer under the name Shielded VM. But today Google Cloud goes further: “We built Confidential VM on top of Shielded VM to harden the OS image and verify the integrity of firmware, kernel binaries and drivers,” said Google Cloud in a note. “The OS images supported by Google include Ubuntu v18.04, 20.04, Container Optimized OS (COS v81) and RHEL 8.2. We are working on CentOS, Debian, and other distributions to provide additional confidential OS images. ”
Google has also worked closely with AMD to ensure that the performance measures of confidential virtual machines are as close as possible to the others. Powerful open source drivers for high speed storage and network traffic were also developed. Confidential virtual machines run on the Google Cloud N2D series machine types that support up to 224 vCPU with 8 GB of memory per vCPU.
Assured Workloads for Government currently confined to the United States
In addition to Confidental VM, Google Cloud also announced the Assured Workloads for Government solution, which should allow government institutions and their subcontractors to meet their security and compliance needs which they need to go to the cloud. public. Traditionally, cloud solutions for governments have relied on separate data centers and isolated environments that do not take advantage of all the functionality and computational resources of traditional offerings.
In order to bridge this gap, Assured Workloads for Government allows users to restrict the location of stored data and cloud resources in specific regions, but so far only in the United States. This solution also makes it possible to prevent the risks linked to configuration errors by selecting predefined security controls coupled with organizational rules. Customers can also choose to restrict Google to workloads based on various parameters such as the citizenship of individuals, their geographic location, etc. Insured government workloads allow clients to meet standards set by the Department of Defense (IL4), the FBI’s Criminal Justice Information Services (CJIS) division and the federal management program Risk and Clearance (FedRAMP), Google said.