MemCCA: Novel Near-Memory Support for Confidential Compute Architectures
Lead Participant:
ETHICRONICS LTD
Abstract
The project "Novel Near-Memory Support for Confidential Compute Architectures" addresses the emerging challenges in heterogeneous computing environments by leveraging Compute eXpress Link (CXL) technology. CXL serves as a memory interconnect protocol, facilitating the scaling of memory resources in disaggregated computing environments. Recent advancements have seen CXL memory devices with added compute capacity, opening up opportunities for near-memory computation while mitigating latency and bandwidth issues associated with accessing distant memory resources.
However, existing CPU-centric Trusted Execution Environment (TEE) solutions lack confidential computing (CC) capabilities in such disaggregated environments. To bridge this gap, our project aims to expand the trust boundaries of current CC models to include CXL devices, enabling near-memory confidential computing for enhanced security and performance benefits.
To achieve this, we will integrate trusted computing principles, such as attestation, into CXL chiplet devices using cryptographic and hardware assurance methods. By extending trust boundaries to memory devices, we enable the adaptation of confidential computing practices in settings where memory-intensive applications with high-performance requirements cannot effectively fit into existing CPU-defined models.
Our innovative approach addresses the growing demand for secure and efficient computing in heterogeneous environments, unlocking new possibilities for confidential computing in memory-intensive applications. By leveraging CXL technology and incorporating trusted computing principles, we aim to pioneer novel solutions that enhance security, performance, and scalability in modern computing architectures.
However, existing CPU-centric Trusted Execution Environment (TEE) solutions lack confidential computing (CC) capabilities in such disaggregated environments. To bridge this gap, our project aims to expand the trust boundaries of current CC models to include CXL devices, enabling near-memory confidential computing for enhanced security and performance benefits.
To achieve this, we will integrate trusted computing principles, such as attestation, into CXL chiplet devices using cryptographic and hardware assurance methods. By extending trust boundaries to memory devices, we enable the adaptation of confidential computing practices in settings where memory-intensive applications with high-performance requirements cannot effectively fit into existing CPU-defined models.
Our innovative approach addresses the growing demand for secure and efficient computing in heterogeneous environments, unlocking new possibilities for confidential computing in memory-intensive applications. By leveraging CXL technology and incorporating trusted computing principles, we aim to pioneer novel solutions that enhance security, performance, and scalability in modern computing architectures.
Lead Participant | Project Cost | Grant Offer |
---|---|---|
ETHICRONICS LTD | £900,000 | £ 630,000 |
  | ||
Participant |
||
UNIVERSITY OF OXFORD | £269,972 | £ 269,972 |
People |
ORCID iD |
Franck Courbon (Project Manager) |