P058 GPU TEE security evidence

What independent, non-vendor security evidence exists for GPU secure enclaves (TEEs) against physical and side-channel attack, and how many in-scope breaks have been found?

Independent security evidence for GPU TEEs, incl. physical and side-channel attacks.

Terms: TEE (trusted execution environment): an isolated, hardware-protected area of a processor whose code and data the machine's owner cannot read or alter.

Status

Started One independent test of NVIDIA GPU confidential computing under physical attack, and it found a break. Academic researchers built a DDR5 memory-bus probe for under $1,000, extracted Intel's attestation key from a fully trusted Xeon in about 15 minutes, and ran a workload on an RTX 3060 outside any secure enclave while passing H100 confidential-computing attestation relayed from a rented H100, because the GPU attestation is not tied to a specific VM. The break goes through the CPU; the H100's own memory and enclave were not attacked, and both vendors call physical attacks out of scope. An independent reverse-engineering of H100 confidential computing (Gu et al. 2025) finds the boot chain and attestation sound within NVIDIA's threat model, but host-visible plaintext command metadata that allows partial reordering or dropping of firmware calls, unprotected command queues, a transfer-size timing channel and still-readable registers: leaks and partial integrity loss, not a break. No physical attack on a data-centre GPU's HBM is held. A Rowhammer attack on workstation GPUs (Lin et al. 2026, GPUBreach) is a cross-reference: its authors list confidential-computing mode as untested.

proves the CPU interfaces directly with a genuine CC GPU. However, the attestation does not identify what the GPU is running.
Chuang et al. (2026), TEE.fail: Breaking Trusted Execution Environments via DDR5 Memory Bus Interposition (IEEE S&P '26) (2026 (public October 2025)), L1187–1189

Posed by

Also raised in: Harack et al. (2025), Verification for International AI Governance (Oxford AIGI); Baker, Kulp, Marks, Brundage & Heim (2025), Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment (RAND WR-A4077-1); Scher & Thiergart (2024), Mechanisms to Verify International Agreements About AI Development (MIRI); Petrie & Aarne (2025), Technical Options for Flexible Hardware-Enabled Guarantees (flexHEG Part 2); Aarne, Fist & Withers (2024), Secure, Governable Chips (CNAS); Ilhan, Withers, Gietz & Harack (2026), Verifiable Semiconductor Manufacturing (Oxford AIGI); Cankaya (2026), Catching Misreporting About ML Hardware Use by Turning Noise into Signal, Part II; Cankaya (2026), Compute Verification FAQs; Cankaya (2026), Research Note: The Fundamentals and Feasibility of Secure Network Taps for Verifying AI Datacenter Use; Ansari (2026), Hardware-Level Governance of AI Compute: A Feasibility Taxonomy for Regulatory Compliance and Treaty Verification; Harack (2026), AI Verification: Infrastructure for Prosperity, Governance, and Peace (Lawfare); Hausenloy & Li (2026), The Compute Verification Post; Kulp et al. (2024), Hardware-Enabled Governance Mechanisms (RAND WR-A3056-1); Jin et al. (2026), Trusting What You Cannot See: Auditable Fine-Tuning and Inference for Proprietary AI, arXiv 2603.07466; AI 2040 (2026), Get Involved in Verification; AI Security Forum, Requests for Discussion; Proofworks, Living Verification Documents.

Sub-questions

IDSub-questionWhat the sources report
P058.1How many independent (non-vendor) security evaluations of GPU confidential computing have been published, how many in-threat-model breaks have they found, and how soon after release?open
P058.2Has the GPU system processor (GSP) firmware / RoT been the subject of a published independent security analysis?open
P058.3Has a physical-access attack (e.g. HBM bus or interposer probing) broken the confidential-computing mode of a current data-centre GPU, and at what cost — given vendors treat such attacks as out of scope?a source suggests an answer (not checked)

Sub-questions not yet stated precisely enough to check

IDSub-question
P058.4Are GPU TEEs vulnerable to power/timing side channels (which leave no tamper evidence), and do their crypto primitives lack agility?
P058.5Can a CC/TEE stack built by one vendor be trusted internationally, and which attestation mechanisms are nearest to that?
P058.6Which microcode/hardware changes to CC would make it fit for verification?
P058.7How can governance-grade remote attestation survive an attacker with physical, side-channel, fault and invasive capability, given secure boot and attestation cover software only?

Possibly relevant work found by automated search

Found by AI agents searching for this problem's sub-questions, mostly by keyword and citation in the OpenAlex scholarly database and on the web; a second AI pass reviewed each one, and we spot-checked some by hand. They seemed reasonably related, as far as one can tell from outside each paper's field. Some come from neighbouring fields, such as arms control and financial audit, and bear on the problem by analogy. Not reviewed against the status above; a paper listed here does not change that status. 928 of 1052 quotes checked word for word against the paper's text.

For P058.1: How many independent (non-vendor) security evaluations of GPU confidential computing have been published, how many in-threat-model breaks have they found, and how soon after release?

For P058.3: Has a physical-access attack (e.g. HBM bus or interposer probing) broken the confidential-computing mode of a current data-centre GPU, and at what cost — given vendors treat such attacks as out of scope?

For P058.5: Can a CC/TEE stack built by one vendor be trusted internationally, and which attestation mechanisms are nearest to that?

Data