P034 Attributing captured traffic to declared workloads
Can analysis of captured traffic attribute it to declared workloads and detect an undeclared workload embedded in a declared one, at stated error rates?
Attributing captured traffic to declared workloads.
Status
Analyzed A protocol design with a security argument: an auditor-owned passive tap hashes inter-node collective payloads into a Merkle tree that is checked against the declared training at challenge time; the authors list the canonical wire-to-tensor mapping ('not a stable public interface today'), auditing that mapping without trusting the trainer's host, and the absence of any NVLink witness as open problems; the paper's prototype concerns the proof system, not the tap or the attribution. Cankaya (2026) separately proposes per-transfer metadata headers (job ID, tensor type, layer, step) so captured samples can be matched to declared workloads, with negligible overhead asserted, not measured. Detecting an undeclared workload embedded in a declared one at stated error rates: no measurement on real traffic; Baker et al. (2025) call the analysis an unsolved problem.
Network anchor. Catches any deviation that changes wire-bound content; deterministic (not statistical) detection on any sampled wire-bound tensor.
Posed by
- 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):
Analyzes the kernels, other logged data, and Prover declarations to verify that the compute cluster executed (only) the declared high-level code. This is an unsolved problem.
(App. A.3, p. 43, L2597–2599; Table 10 2A, p. 34, L2064–2065)
Also raised in: Cankaya (2026), Research Note: The Fundamentals and Feasibility of Secure Network Taps for Verifying AI Datacenter Use; Peigné-Lefebvre, Nguyen & Wang (2026), Zero Knowledge Verification for Frontier AI Training Is Possible, arXiv 2606.05433.
Sub-questions
| ID | Sub-question | What the sources report |
|---|---|---|
| P034.1 | Can analysis of logged kernels and inter-accelerator traffic detect an undeclared workload embedded in a declared one, with reported false-accept and false-alarm rates? | open |
| P034.2 | Can a canonicalisation shim reproduce identical hashes of collective (NCCL ring-allreduce) traffic across configurations, i.e. a stable map from wire bytes to tensors? | open |
| P034.3 | What byte overhead do per-transfer metadata headers (job ID, tensor type, layer, step) add relative to tensor size? | partly answered |
Sub-questions not yet stated precisely enough to check
| ID | Sub-question |
|---|---|
| P034.4 | How can an auditor verify the queue-pair-to-tensor mapping without trusting the trainer's host? |
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 P034.1: Can analysis of logged kernels and inter-accelerator traffic detect an undeclared workload embedded in a declared one, with reported false-accept and false-alarm rates?
For P034.2: Can a canonicalisation shim reproduce identical hashes of collective (NCCL ring-allreduce) traffic across configurations, i.e. a stable map from wire bytes to tensors?
* Full text not reviewed; more likely to contain errors.
Data
- problems.csv — one row per problem
- problem_status.csv — status, scope and evidence per problem
- problem_sources.csv — the posed-by quote and other documents per problem
- leaves.csv — all sub-questions, including those without a checkable form
- leaf_sources.csv — every source row behind each sub-question