Skip to content

Separation of Concerns for IP Protection in USD - #107

Open
asluk wants to merge 11 commits into
PixarAnimationStudios:mainfrom
asluk:aluk/ip-protection
Open

Separation of Concerns for IP Protection in USD#107
asluk wants to merge 11 commits into
PixarAnimationStudios:mainfrom
asluk:aluk/ip-protection

Conversation

@asluk

@asluk asluk commented May 14, 2026

Copy link
Copy Markdown

Description of Proposal

Summary

Document the IP protection requirements that arise when USD is used in cross-organization workflows -- manufacturing, data center infrastructure, M&E, aerospace -- and separate them into four distinct concerns with different threat models and different likely solutions.

Problem statement

"IP protection" is used as a single label for at least four separate things: protection against dissemination, ephemeral data and runtime protection, copyright and attribution, and provenance and compositional integrity. When they get lumped together, discussions stall because participants are effectively trying to solve different problems at the same time. This proposal provides a clear problem statement and separation of concerns.

Glossary

  • Protection through omission -- The approach of curating exports to include only what the recipient should see, rather than relying on runtime enforcement.
  • Tiered representations -- Multiple representations of the same asset curated for different audiences (full access, customer simulation, public/layout), distinct from LODs.
  • Living provenance -- Provenance that invalidates when the composition is changed, as opposed to a static origin stamp.
  • The recombination problem -- USD's composition model allows components from different validated assemblies to be rearranged into configurations never sanctioned by the original supplier.

Reference Links

Details

The proposal frames IP protection as four independent concerns, each with different characteristics and likely solutions. It documents existing USD mechanisms (asset resolvers, variable expressions, composition arcs, customData), identifies what USD can address vs. what requires external systems vs. what requires guidance, and provides practical guidance on structuring USD assets for protectability. No changes to core USD are assumed. Where future USD enhancements might help, those are identified and scoped for potential follow-up proposals contingent on community discussion.

Risks

  • Scope creep into solution space (the proposal is a problem statement, not a solution proposal)
  • Incomplete industry coverage (currently weighted toward manufacturing and infrastructure)

Alternate solutions

N/A -- this is a problem-statement proposal, not a solution proposal.

Out of Scope

DRM and encryption schemes, specific access control implementations, cloud streaming and pixel-only delivery, and concrete schema proposals for copyright metadata or flatten-protection annotations. Each is identified as downstream work or as out of scope for USD entirely.

Link to Rendered Proposal

Supporting Materials

None at this time.

Contributing

asluk and others added 11 commits April 8, 2026 10:15
Joint proposal with Stephen Prideaux-Ghee (PTC), Steve Blackwell (Vertiv),
and Aaron Luk (NVIDIA). Establishes four distinct IP protection concerns
(dissemination, ephemeral data, copyright, provenance), provides industry
use cases, and frames guidance on asset structuring for protectability.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Rework phrasing to reduce AI-isms and diverge from identifiers
  proposal voice
- Break walls of text into bullets and sub-bullets throughout
- Add Media & Entertainment industry use case section (multi-vendor
  VFX, pre-release confidentiality, cross-studio collaboration)
- Reference M&E alongside other industries in intro, motivation,
  and design principles
- Link identifiers proposal to PR PixarAnimationStudios#105 instead of relative path
- Fix line widths to ~75 chars
- Drop NVIDIA platform vendor label, drop Omniverse mention

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Soften language throughout to avoid reading as proposed USD changes:
- Table "USD role" column: point to Open Questions instead of naming solutions
- Concern 3 (Copyright): add illustrative disclaimers on code blocks, hedge
  flatten/accumulate language
- "What USD can address": replace "low-risk enhancement" with conditional
  framing, defer specifics to community discussion
- Next steps: make follow-up proposals conditional on community input
- Broaden privilege boundaries guidance to cover all external-layer arcs
- Fix fact-check: `final` is not a current USD keyword
- Replace "variant" with "approach" to avoid USD variant confusion

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Renderers need geometry, not policy: carrying IP privilege metadata
  compromises render performance and still cannot enforce protection
- Note that customLayerData does not survive Flatten, citing the
  Revise Use of Layer Metadata proposal

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Reorder open questions by priority (scope first, streaming last)
- Condense resolver precedent paragraph; note that OAuth/PAM/JAAS
  standardize auth rules while ArResolver intentionally does not
- Unpack authorship traceability vs compositional provenance
- Rename "Defense in depth" to "No single mechanism is sufficient"
- Remove internal JIRA reference; reorder contributor credits

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add factory planner requirement to hide existence of references (not
just content), property-level protection as two distinct requirements
(export control vs consumer visibility), operational data tiers as
independent axis from geometry, vendor management cost of tiered
delivery, lowest-privilege tier must serve a workflow, display
watermarking prior art, view-dependent obfuscation prior art, and
streaming rationale from mesh surface reconstruction risk.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Eliminate verbose "Existing mechanisms" section, replace with
lightweight cross-reference. Collapse "Separation of concerns"
subsections into table. Fold "Relationship to other proposals"
inline. Compress two-personas lists, copyright code blocks, next
steps. Merge principles 5+6. Remove redundant paragraph after
four-concerns table and generic "Why now" point.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Contributors won't feed valuable IP into shared pools without
confidence it's protected — making this the root driver behind
all four concerns. Scoped to industrial USD; M&E noted separately.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Covers all 13 inline comments + summary review.

- Glossary (M&E / PLM / DRM)
- ArResolver caveats: usdcat, multi-vendor, composition impact
- Streaming reframed as out-of-band; out of scope for USD
- `final` -> `unflattenable`; unenforceability framing strengthened
  (open-data-access bullet, Presto `locked` reference, DRM-history
  closing)
- C2PA reference + AuthorshipAPI cross-link in provenance
- Variable Expressions as package-and-deliver alternative
- Provenance: arbitrary modification, per-prim checksum scope,
  scenegraph markup as signature-scope complication
- Open questions: resolver missing-credentials behavior, virality /
  aggregation as cross-proposal question

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

## Concern 4: Provenance and compositional integrity

Provenance -- tracking where content came from -- gets complicated in

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's not just where it comes from, but who created/edited it, when they did it, etc.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It may include authoring, ownership/copyright, etc.

1. Reference both product C+D and product E+F.
2. Extract component C from the first and component F from the second.
3. Compose them into a new structure C+F.
4. Save this as a new asset that looks like a valid product.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

and it may well be a valid product - but not from the same producer as the originals. And that's OK.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

well technically it isnt a valid product because they've taken my product and changed it and if i'm not allowing that (copyright), what they create will never be a valid product. clearly we need to get terminology right. we should chat.


If product C+D carries a provenance marker indicating it is a validated
configuration from vendor X:
- Extracting C and composing it with F from a different assembly should

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It doesn't break the existing provenance - but that the existing provenance may/does carry into the new asset. It depends on how "vendor X" define things - and they could setup the provenance to be only over the entire thing OR as a series of "ingredients" that can be used independantly.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fair point; that is assuming of course that there's aa mechanism to do that. interested to learn more about your stuff.

- The new assembly C+F should not inherit the validated status of its
source components.

This is "living" or "dynamic" provenance -- not a static stamp, but a

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is provenance. It's not any more living or dynamic than any other understanding of provenance.

components -- invalidates the signature. This is conceptually clean
but requires infrastructure for key management and signature
verification. The
[Coalition for Content Provenance and Authenticity (C2PA)](https://c2pa.org/)

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

And I (as chair of the C2PA's TWG) have a proposal for how to bring C2PA's Content Credentials into USD(Z)

combined and whether the combination has been validated.
- Copyright and provenance may benefit from overlapping mechanisms --
e.g., digital signatures that attest to both origin and integrity.
- An honest limit: provenance signals are best understood as

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We've done a lot of work at C2PA to allow these features - if you want them in your system...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: Draft

Development

Successfully merging this pull request may close these issues.

4 participants