P012 Physical inspection capability and cost
What is the per-die cost, time and miss rate of destructive and non-destructive die inspection, and how many chips must be inspected to catch an inserted trojan with high confidence?
Capability, cost and sampling statistics of physical die inspection.
Terms: q: the fraction of chips that carry a hardware trojan.
Status
Analyzed A desk analysis of the sampling clause: inspection cost and sample size against Trojan prevalence, with the observation that sampling 1% of chips may cost more than producing all chips unverified; Aarne & Petrie (2025, flexHEG Part 3) give the same arithmetic for a 1/1000 sampling rate (63% chance of catching a 1,000-device compromised batch). No held document measures per-die cost, time or miss rate of destructive or non-destructive inspection at an advanced node: Harack et al. (2025) state no non-destructive technique is known for a chip with tens of billions of transistors, and Ilhan et al. call inspection 'extremely expensive, slow, and often destructive' without numbers. Cross-references, not evidence for this headline: Agrawal et al. (2007) simulate side-channel screening (no silicon); Thummala et al. (2026) measure a power side-channel screen on one Atmel microcontroller (93.2% of tampered-firmware traces flagged at 1% false positives), firmware tampering rather than die modification.
If only 1 chip in 100,000 is compromised, many thousands must be tested to achieve high confidence of detection. The sampling regime can be tuned based on acceptable risk (e.g., testing 1% of chips vs 0.01%), but this involves significant cost considerations.
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):
One might worry that delayering and scanning a random sample of chips would be too slow for timely detection of Hardware Trojans (or false specs), with modern chips having “up to 100 layers” (ASML) and nanometer-scale features.
(App. A.1, p. 40, L2415–2417) (source states the problem but not the quantity the question asks for)
Also raised in: Harack et al. (2025), Verification for International AI Governance (Oxford AIGI); Ilhan, Withers, Gietz & Harack (2026), Verifiable Semiconductor Manufacturing (Oxford AIGI); Cankaya (2026), A System Overview for Near-Term, Low-Trust AI Compute Verification (v0.2); Proofworks, Living Verification Documents.
Sub-questions
| ID | Sub-question | What the sources report |
|---|---|---|
| P012.1 | What is the most complex logic die recovered by delayering and imaging, and what wall-clock time and cost per die (destructive vs non-destructive) at an advanced node? | open |
| P012.2 | What is the per-die cost and miss rate of SEM/FIB imaging for a single rogue via at an advanced node? | open |
| P012.3 | What is the most advanced node and layer count at which full-die non-destructive imaging (e.g. ptychographic X-ray laminography) has been published? | a source suggests an answer (not checked) |
| P012.4 | Can non-destructive X-ray imaging resolve buried interconnects in a CoWoS/SoIC-class package at rogue-via resolution? | open |
| P012.5 | At trojan prevalence q (e.g. 1-in-100,000), how many chips must be inspected for a stated detection probability, and when does that cost exceed producing all chips unverified? | open |
| P012.6 | Has any inspection method (e.g. dopant-contrast imaging) detected a dopant-level trojan in a blind test? | implied by a source, not stated as open |
| P012.7 | Has side-channel IC fingerprinting been demonstrated on a large-transistor AI accelerator die, at what per-unit cost and false-negative rate as a supply-chain screening tier? | partly answered |
| P012.8 | Can IRIS imaging detect an inserted comparator trojan on a leading-edge die? | open |
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 P012.1: What is the most complex logic die recovered by delayering and imaging, and what wall-clock time and cost per die (destructive vs non-destructive) at an advanced node?
For P012.3: What is the most advanced node and layer count at which full-die non-destructive imaging (e.g. ptychographic X-ray laminography) has been published?
For P012.4: Can non-destructive X-ray imaging resolve buried interconnects in a CoWoS/SoIC-class package at rogue-via resolution?
For P012.6: Has any inspection method (e.g. dopant-contrast imaging) detected a dopant-level trojan in a blind test?
For P012.8: Can IRIS imaging detect an inserted comparator trojan on a leading-edge die?
* 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