How GCP Confidential Computing develops trusted AI

How Trusted AI Is Built on Confidential Computing

GCP Confidential Computing has altered cloud-based sensitive workload management for enterprises. Due to Google’s hardware ecosystem development, Confidential Computing can enable cutting-edge applications like privacy-preserving AI and multi-party data analytics, generating a new wave of adoption.

Google Cloud announces its latest Confidential Computing innovations, showing how its clients use it to protect their most sensitive workloads, including AI workloads.

Most recent developments

A sneak peek at GKE Nodes and private virtual machines with NVIDIA H100 GPUs for AI workloads
Preview of the confidential Vertex AI Workbench
An overview of Confidential Space using the widely accessible NVIDIA H100 and Intel TDX GPUs
Generally accessible, confidential GKE Nodes on C3 computers with integrated acceleration and Intel TDX
General availability of confidential GKE nodes on N2D machines with AMD SEV-SNP
Preview of confidential virtual machines running AMD SEV on C4D machines
Check out Gemini Cloud Assist’s preview to speed up your Confidential Compute trip.

Confidential Computing New Use Cases

The use of GCP Confidential Computing by businesses to unlock commercial breakthroughs is having an influence on all of the major industries.

AiGenomix

With a global network of partners in the public and private sectors, AiGenomix is using Google Cloud Confidential Computing to provide extremely unique infectious disease surveillance, early cancer detection, and medicines intelligence.

Google Ads

Confidential matching has been implemented by Google Ads to safely link first-party consumer data for marketing purposes. Confidential computing is being used for the first time in Google Ads products, and more businesses will eventually incorporate this privacy-enhancing technology.

Swift

Swift is leveraging GCP Confidential Computing to fuel a money laundering detection model while guaranteeing that private information from some of the biggest banks stays totally confidential.

Confidential Computing

Confidential VMs, Confidential GKE, Confidential Dataflow, Confidential Dataproc, and Confidential Space are tools for protecting data while it’s being used.

Protect your data by encrypting it while it’s being processed.
Performance is not sacrificed for a straightforward, user-friendly implementation.
Collaboration in confidence while maintaining data ownership

Benefits

A breakthrough in the field of confidentiality

Customers may now encrypt their data in the cloud as it’s being processed with a revolutionary technology called Confidential Virtual Machines.

Easy for everyone to understand

Google Cloud’s method enables users to encrypt data while it’s being used without requiring them to alter the code of their apps or sacrifice performance.

Opening up new opportunities

Confidential Computing opens up possibilities that were previously unattainable. Companies can work together while maintaining the privacy of their information.

GCP Confidential Computing features

Confidential Computing Platform

Confidential VMs

By encrypting data-in-use during processing, confidential virtual machines (VMs) can safeguard the privacy of data stored in the cloud. Modern CPUs from AMD, Intel, and other manufacturers provide security features that are utilized by confidential virtual machines. Customers can be sure that their data will remain secure and private even when it is processed in the cloud with GCP Confidential Computing.

For AI/ML workloads on Intel, Google Cloud also leverages the Intel AMX, a CPU accelerator that is activated by default on the general-purpose C3 machine series for Confidential VMs. Your AI data and models are protected at the hardware level with Confidential VMs on the C3 machine series, which also significantly improves performance for tasks involving deep learning and inference.

Confidential VMs with H100 GPUs

Confidential virtual machines (VMs) on the accelerator-optimized A3 machine series with NVIDIA H100 GPUs can enable companies to fully utilise AI and machine learning while protecting sensitive information. With H100 GPUs, confidential virtual machines (VMs) help guarantee that data is safe from the time it enters the GPU until the output is produced. This lessens the possibility of unwanted access, even from privileged users or malevolent actors operating within the system. Businesses may work together more freely and securely with partners and outside vendors with Confidential VMs on the A3 machine line, which offer a trusted execution environment for AI applications.

Confidential GKE Nodes

It is possible to encrypt data processed within your GKE cluster while maintaining a high level of performance with Confidential GKE Nodes. Confidential GKE Nodes and Confidential VMs share the same technological underpinnings. This feature enables you to use dedicated, node-specific keys that are created and controlled by the processor to keep data encrypted in memory. Google or other host nodes cannot access the keys since they are produced in hardware upon node construction and only exist in the CPU.

Confidential Space

Ensuring the confidentiality of sensitive data while allowing organisations to benefit from its aggregation and analysis is possible using Confidential Space. With the assurance that their data is safeguarded from all parties, including strengthened protection against cloud service provider access, organisations can carry out tasks like collaborative data analysis and machine learning (ML) model training. In the post-cookie era, privacy-preserving ad campaign analytics and remarketing can be conducted using the Confidential Space integration with Privacy Sandbox, which offers a trusted execution environment.

Confidential Dataflow and Dataproc

A vast array of machine learning and streaming analytics use cases is supported at scale by Dataflow, a fully managed service. Your data pipelines can be processed with Compute Engine Confidential VMs, which offer inline memory encryption, with Dataflow’s support for Confidential VMs.

With fully managed Spark, Hadoop, and other open source technologies and frameworks, Dataproc makes large data processing possible. With Confidential Dataproc, you may set up a Dataproc cluster that encrypts inline memory using Compute Engine Confidential virtual machines. This strengthens security assurances, particularly when handling extremely sensitive data.

All features

Utilizing real-time encryption

Users of Google Cloud can encrypt data while it’s being used, utilising cloud services for secret computing as well as security features provided by contemporary CPUs from AMD, Intel, and others. Clients may rest assured that even during processing, their data will remain confidential and encrypted.

Elevate and relocate confidentially

Google Cloud’s objective is to simplify GCP Confidential Computing. All workloads you currently operate, both new and old, can run as Confidential VMs, making the switch to them smooth. To use Confidential VMs, you don’t need to modify any code in your apps. It’s as easy as checking one box.

Identification of sophisticated persistent attacks

GCP Confidential Computing expands on the rootkit and bootkit defences provided by Shielded virtual machines. This makes it easier to guarantee the integrity of the operating system you decide to use in your private virtual machine.

Increased creativeness

GCP Confidential Computing has the potential to enable previously unattainable computer scenarios. Now, companies may work together on regulated and sensitive data in the cloud while maintaining secrecy.

High performance

The performance of regular N2D virtual machines is comparable to that of confidential virtual machines.