SPICE @ IETF 121

Tuesday 2024-11-05 13:00Z

Welcome & What is SPICE - chairs

https://datatracker.ietf.org/doc/slides-121-spice-chair-slides/

The SPICE Working Group session at IETF 121 centered on refining
Selective Disclosure for CBOR Web Tokens (SD-CWT), introducing the
concept of Global Unique Enterprise Identifiers (GLUE), and improving
CBOR data structures for enhanced security and interoperability.
Discussions included selective disclosure mechanics, the complexities of
nested redactions, and risks associated with issuer-verifier collusion
in corporate identity verification.

Topics Discussed

1. Introduction to SPICE and Scope
   SPICE’s mission is to fill gaps in digital credentials, particularly
   by ensuring both human and non-human identities are securely and
   privately represented. The chairs explained that while digital
   credentials are widely discussed, many specifications fail to
   address specific use cases. SPICE aims to bridge these gaps by
   developing profiles tailored to varied needs. The scope covers
   security and privacy, including in non-human contexts (e.g., IoT
   devices), but explicitly excludes key discovery and the creation of
   new cryptographic primitives. New attendees were encouraged to
   explore related groups (e.g., RATS, OAUTH, COSE) to see how SPICE
   integrates with these efforts.

2. Selective Disclosure for CBOR Web Tokens (SD-CWT)
   Rohan Mahy introduced SD-CWT, a method allowing selective disclosure
   of claims in a CBOR Web Token. He described SD-CWT as a solution for
   enabling users to selectively reveal information based on the
   verifier's needs. This structure involves a 'redaction' process,
   where claims are either fully disclosed or hidden (blinded) using a
   16-byte salt for consistency and security.

   • Technical Details: For disclosure, the issuer generates a
     random salt and combines it with the claim, creating a
     three-element array: the salt, claim key, and claim value. This
     ensures that while claims are present, the verifier only sees
     what’s intentionally revealed.
   • Syntax Options: Rohan outlined syntax considerations,
     especially for nested structures, and Orie Steele added insights
     on structuring arrays for hierarchical claims. Discussion
     highlighted syntax compatibility with both JSON and CBOR, aiming
     to prevent data leakage and unwanted covert channels (data
     encoded without user consent).
   • Key Binding: Orie discussed the importance of ‘key binding’
     to counter replay attacks, ensuring that only the intended
     verifier can access specific disclosures. This approach aims to
     bind disclosures directly to the holder's private key, making it
     difficult to duplicate or share unauthorized access.

3. CBOR and Syntax Optimization
   This segment examined ways to handle redactions in CBOR data
   structures without disrupting interoperability with CWTs. Orie
   Steele presented various methods for redacting keys within CBOR maps
   using unique integer identifiers. He cautioned about balancing data
   size and complexity with the risk of hidden data channels.

   • Challenges: Orie and Carsten Bormann discussed concerns
     about potential covert channels, especially when nested data is
     redacted. The group debated if using unique integers for
     redactions could prevent hidden communication between issuer and
     verifier, ultimately leaning towards maintaining transparency by
     limiting data redactions to standardized integer or string
     values.
   • Redaction Techniques: Carsten noted that while JSON often
     uses strings for keys, CBOR’s unique syntax requires different
     considerations. He proposed using pre-set values (e.g.,
     ‘undefined’ in CBOR major type 7) as placeholders to signal
     redactions, which could be easier to parse and less prone to
     unintended use.

4. Global Unique Enterprise Identifiers (GLUE)
   https://datatracker.ietf.org/doc/draft-zundel-spice-glue-id/

   Brent Zundel presented GLUE as a solution to address corporate
   identities by associating identifiers with specific organizations.
   He proposed a structure for consistent, globally unique identifiers
   that can distinguish between corporate entities, supporting
   scenarios where multiple identifiers need to coexist within a single
   domain.

   • Audience Feedback: Attendees raised questions about how GLUE
     differs from existing Uniform Resource Names (URNs). Brent
     explained that GLUE aims to provide more context by combining an
     identifier with a registry, allowing entities to verify details
     about the identity (e.g., what organization it represents).
     Concerns were raised about potential "land grabs" of short
     identifiers and how multiple schemas might conflict within one
     domain. Brent noted that this remains an open question.
   • Next Steps: Attendees suggested further investigation into
     GLUE’s overlap with URN registries, particularly for
     multi-domain usage. The group agreed to explore adopting GLUE as
     a formal working item and refine it to meet broader needs.

5. OpenID Connect Standard Claims Registration for CBOR Web Tokens
   https://datatracker.ietf.org/doc/draft-maldant-spice-oidc-cwt/

   Beltram Maldant presented his work on aligning OpenID Connect (OIDC)
   standard claims for CBOR Web Tokens (CWT), emphasizing the need for
   a streamlined approach to personal data claims within CWT. He
   highlighted the gap in current CWT and ANI registries where standard
   claims related to personal information (as defined by OIDC) are
   missing, which impacts use cases involving Personally Identifiable
   Information (PII).

   • Draft Proposal: Maldant proposed registering 19 of the 20
     OIDC-defined claims in the CWT registry (subject claim already
     exists), aiming for two-byte range identifiers due to CBOR’s
     payload size constraints. This registration is intended to make
     frequently used claims more efficient in CWT.

   • Feedback and Suggestions: Mike Jones expressed support,
     noting that standardizing these claims in CBOR would enhance
     utility across multiple use cases. Philip raised the question of
     whether address claims were intentionally excluded, to which
     Maldant confirmed they were retained in the draft. Orie Steele
     clarified the history of these claims in JWT and suggested that
     although they traditionally focused on human users, they might
     be adapted for other entities (e.g., devices or organizations).

   • Further Considerations: Discussion touched on expanding
     claims to support corporate or device identities while
     maintaining the intended semantics. Participants suggested that
     prefixing these claims (e.g., “OIDC”) could help clarify their
     usage in CBOR. Roy Williams raised the need for clear
     differentiation between user and corporate identifiers, as some
     claims like “name” may apply differently to individuals versus
     organizations.

6. Audience Q&A and Discussions
   Throughout the session, audience members engaged with various edge
   cases in selective disclosure. Points raised included:

   • Nested Disclosure Complexity: One attendee highlighted the
     challenges of nested claims, asking if claims could be disclosed
     selectively at different hierarchy levels. Orie confirmed that
     this would require careful structuring to ensure clarity across
     levels without compromising data integrity.
   • Differences from SD-JWT: Some participants questioned if
     SD-CWT’s syntax would diverge from Selective Disclosure JWT
     (SD-JWT) standards, which use similar selective disclosure
     mechanisms. Orie explained that SD-CWT aims for more versatility
     in CBOR environments, potentially using unique CBOR tags to mark
     redactions without impacting compatibility.
   • Preventing Collusion: Mike Prorock raised concerns about
     issuer-verifier collusion, especially where hidden data or
     patterns might lead to covert exchanges. He recommended clear
     syntax standards that limit discretionary control over data
     elements to mitigate this risk.
   • Registry and Adoption: Attendees requested a clearer
     understanding of registry structures and whether GLUE would rely
     on a URN-based system or require new conventions. Some suggested
     that SPICE adopt GLUE to enable more feedback-driven
     improvements.

Speaker Contributions

• Rohan Mahy: Introduced SD-CWT concepts, discussed selective
  disclosure mechanics, and outlined syntax for salt-based arrays,
  emphasizing the role of selective redactions in user-controlled data
  sharing.
• Orie Steele: Provided insights into CBOR syntax challenges,
  introduced GLUE concepts, and addressed security considerations for
  structuring nested disclosures.
• Brent Zundel: Presented GLUE, discussed its relevance for
  corporate identity schemas, and addressed audience questions on
  registry overlap and multi-domain management.
• Mike Prorock: Voiced concerns about issuer-verifier collusion,
  and offered insights on maintaining consistent standards for SD-JWT
  syntax comparisons.

Open Questions and Action Items

1. CBOR Redactions: Continue discussions on effective methods for
   CBOR data redactions without compromising compatibility with CWT and
   preventing covert data channels.
2. GLUE Registry Exploration: Further explore possible registry
   structures for GLUE identifiers, considering potential overlaps with
   existing URN conventions and avoiding “land grab” issues.
3. Syntax Feedback: Gather community feedback on specific syntax
   options, particularly for handling redactions and nested disclosures
   within CBOR environments.
4. Collusion Prevention: Address security concerns regarding
   issuer-verifier collusion by developing syntax standards that limit
   discretionary data control.