Security analysis of three digital credential formats
Researchers evaluated the verification pipelines of SD-JWT VC, W3C Verifiable Credentials Data Model 2.0, and mdoc.
The study evaluated the verification pipelines of three formats. These are SD-JWT VC, W3C Verifiable Credentials Data Model 2.0 with the eddsa-rdfc-2022 Data Integrity cryptosuite (W3C VCDM), and mdoc (ISO/IEC 18013-5).
For W3C VCDM, three structural attack surfaces were identified. These include denial of service (DoS) via poison graphs, context injection through @context, and server-side request forgery (SSRF) via malicious URLs.
A poison graph uses blank-node references to force RDFC-1.0 canonicalization into deep recursion. This can raise costs to 278x the baseline from five blank nodes in 527 bytes. This can also raise costs to 13.5 s from eight blank nodes.
An acyclic input with 1,500 blank nodes stays at 462--714x. Every attack family aborts at 16 or below, but every legitimate credential completes at 1.
SD-JWT VC and mdoc need neither canonicalization nor external references. The verifiers for these formats resisted every evaluated vector.
Mitigations include a static document loader, a context allowlist, and a call limit with a size bound. These address the remote context resolution and RDFC-1.0 canonicalization surfaces.
Key facts
- W3C VCDM showed attack surfaces for denial of service, context injection, and server-side request forgery.
- A poison graph with eight blank nodes can raise costs to 13.5 s.
- SD-JWT VC and mdoc resisted every evaluated attack vector.
- Mitigations include a static document loader and a context allowlist.