#United States #Sunnyvale #Confidential AI #Fortanix #Rafay Systems

Introduction

In a landmark move for artificial intelligence security, Rafay Systems and Fortanix have unveiled a groundbreaking joint solution aimed at empowering neoclouds, sovereign AI clouds, and enterprises. This collaboration facilitates a new era of confidential AI as a service, promising heightened security and data protection for sensitive workloads.

The Solution

The newly introduced technology integrates the Rafay Platform with Fortanix's capabilities, allowing operators to run AI workloads in secure, hardware-isolated environments, all while maintaining encryption on sensitive data. This means that information remains protected even during processing, a significant advancement in AI security.

How It Works

The joint solution utilizes Trusted Execution Environments (TEEs) to guarantee that encryption keys are only released after Fortanix verifies both the environment and AI workload through cryptographic attestation. This is especially crucial for industries where users handle sensitive information, such as healthcare and finance. Developers can now launch AI workloads seamlessly from the Rafay Developer Hub, eliminating the need for support tickets, thus simplifying operations for users.

The Importance of Confidential Computing

Traditional encryption methods, while valuable for securing data at rest and in transit, fall short when it comes to protecting data during processing. Research conducted by Fortanix reveals that although 93% of AI leaders understand this limitation, only half can verify adequate protection during inference. By introducing confidential computing, Rafay and Fortanix are paving the way for organizations to safeguard their proprietary data, thereby mitigating potential risks associated with unauthorized access.

Mohan Atreya, Chief Product Officer at Rafay, emphasized the importance of this partnership, stating that this innovation allows model owners to deploy proprietary AI models securely on third-party infrastructures, without the risk of exposing valuable intellectual property. Moreover, it opens up new revenue opportunities for cloud operators to offer premium AI services focused on security.

Features of the Joint Offering

1. Confidential Containers (CoCo): The solution integrates CoCo, a project supported by the Cloud Native Computing Foundation (CNCF), which applies confidential computing principles at the Kubernetes pod level. Each pod is run in a lightweight confidential virtual machine, ensuring that memory is encrypted, substantially enhancing security.

2. Cryptographic Attestation: The Fortanix Confidential Computing Manager (CCM) ensures a rigorous verification process within the computing environment, checking the integrity of CPUs, GPUs, firmware, and software components before releasing encryption keys. This multi-layered approach guarantees that only attested workloads can access sensitive data.

3. Multi-Tenancy: Operators have the flexibility to onboard GPU infrastructures once and then provide isolated tenants for each customer, enhancing privacy and security for users.

4. Self-service KDEs: Administrators can package workloads conveniently and offer them as SKUs in Rafay's SKU Studio, allowing end users to launch them on demand.

5. Governance: The solution includes comprehensive governance features, including policy compliance, access controls, and quotas across various tenants and clusters, ensuring compliant and secure operations.

Addressing Trust Dilemmas in AI

AI factories often face significant trust challenges. Model owners seek to conceal their weights from both operators and customers, while enterprises require the protection of their data from these same entities. The collaboration between Rafay and Fortanix provides an elegant solution to this dilemma. With just one click, operators can orchestrate the deployment of proprietary models securely while ensuring that the keys are only released to verified workloads.

Practical Applications

This solution is designed to cater to several use cases:

• - Neoclouds and GPU Cloud Providers: They can now offer confidential AI as a premium service backed by attestation evidence, providing peace of mind to users that their workloads are secure.
• - Sovereign AI Clouds: These entities can maintain control of encryption keys securely, attesting to what is running before any data is decrypted, thus preserving confidentiality.
• - Regulated Enterprises: Sectors like finance and healthcare can deploy AI solutions on sensitive data while maintaining confidentiality throughout the processing phase.

Conclusion

The partnership between Rafay Systems and Fortanix marks a significant advancement in delivering confidential AI solutions, addressing a critical need for security in the rapidly evolving AI landscape. Their combined efforts not only enhance protection for sensitive data but also pave the way for organizations to confidently leverage AI for strategic advantages, all while ensuring adherence to stringent security standards.

For further information, visit Rafay's official website to explore the vast possibilities this partnership offers to transform the future of confidential AI computing.