Confidential collaborative machine learning (ML) enables multiple mutually distrusted data holders to jointly train an ML model while preserving the confidentiality of their private datasets due to regulatory or competitive reasons. However, existing works need frequent data and model exchanges during training via slower conventional links. They face increasing challenges due to the exponentially growing sizes of models and datasets in modern training workloads like large language models (LLMs), resulting in prohibitively high communication costs. In this paper, we propose a novel mechanism called GPU Travelling that leverages recently emerged confidential GPUs. With our rigorous design, the GPU can securely travel to the specific data holder to load the dataset directly into the GPU's protected memory and then return for training, eliminating the need for data transmission while ensuring confidentiality up to a data-centre level. We developed a prototype using Intel TDX and NVIDIA H100 and evaluated its performance on llm.c, a CUDA-based LLM training project, and demonstrated the performance and feasibility while maintaining strong security guarantees. The results showed at least 4x speed improvement when transmitting a 512 MiB dataset chunk versus conventional transmission.

The growing need of trusted execution environment (TEE) has boomed the development of hardware enclaves. However, current TEEs and their applications are tightly bound to the hardware implementation, hindering their compatibility across different platforms. This paper presents vSGX, a novel system to virtualize the execution of an Intel SGX enclave atop AMD SEV. The key idea is to interpose the execution of enclave instructions transparently to support the SGX ISA extensions, consolidate encrypted virtual memory of separated SEV virtual machines to create a single virtualized SGX-like address space, and provide attestations for the authenticity of the TEE and the integrity of enclave software with a trust chain rooted in the SEV hardware. By design, vSGX achieves a comparable level of security guarantees on SEV as that on Intel SGX. We have implemented vSGX and demonstrated it imposes reasonable performance overhead for SGX enclave execution.