AnonCreds

AnonCreds (Anonymous Credentials) is an opinionated verifiable-credential format built from the ground up around zero-knowledge proofs, with the explicit design goal of maximizing holder privacy and unlinkability. Where most credential systems bolt privacy on afterward — a disclosure flag here, a hashed claim there — AnonCreds treats it as the default posture: a presentation reveals as little as cryptographically possible, and two presentations of the same credential cannot be correlated with each other.

Core privacy machinery

AnonCreds v1 rests on CL signatures, named for the Camenisch–Lysyanskaya line of cryptographic work that underpins them, and three techniques combine at presentation time. First, the holder never shows the issuer's signature itself; they present a zero-knowledge proof of possession of a valid signature, freshly randomized for each verifier, so there is no static signature value for verifiers to match across presentations — the correlation handle that plagues simpler formats simply never exists. Second, a link secret binds credentials to the holder in blinded form: the same secret sits inside every credential the holder ever receives, proving to each verifier that the credentials being combined belong to one person, yet without ever becoming a shared tracking identifier, because no issuer or verifier ever sees its raw value. Third, predicate proofs let a holder prove a statement about a value without disclosing the value — that a birthdate implies age over 18, that a number falls within a range — which removes entire categories of correlatable personal data from circulation. A bartender learns you are of age; they do not learn your birthday, and they could not recognize you by your credential if you returned tomorrow.

Evolution and trade-offs

AnonCreds v2 generalizes the architecture: the CL-signature core becomes pluggable, so schemes such as BBS signatures can slot in, and revocation — historically the format's heaviest machinery — gains a more scalable design. The strengths are unambiguous: uncompromising unlinkability and predicate support that lighter formats like SD-JWT do not attempt. The historical costs are equally real: heavier cryptography on constrained devices, a credential-definition and revocation-registry model originally tied to specific ledgers, and reliance on cryptographic assumptions and key sizes that diverge from the mainstream JOSE toolchain. That divergence is one reason large public deployments have gravitated toward simpler salted-hash formats first — deployability beat privacy, a pattern privacy engineers will find wearily familiar.

Why Bitcoiners should pay attention

The problem AnonCreds attacks — proving facts about yourself without becoming trackable — is the identity-layer twin of Bitcoin's payment-privacy problem, and the design philosophies rhyme. The link secret plays a role analogous to a private key: never revealed, only proven. Predicate proofs are the identity equivalent of not broadcasting your whole balance to make one payment. And the unlinkability goal mirrors what output linking resistance pursues on-chain: interactions that cannot be stitched into a profile. As digital-ID mandates spread, the difference between credential systems that are private by construction and those that are private by policy will matter enormously — policy can be amended; mathematics cannot.

Where it fits

AnonCreds remains among the strongest practical expressions of selective disclosure available for a privacy-respecting verifiable presentation. If SD-JWT is the pragmatic floor of the credential-privacy spectrum, AnonCreds is the demonstration of how high the ceiling goes — and a benchmark to hold every "privacy-preserving" digital-ID proposal against. It also teaches a broader lesson worth carrying into every protocol evaluation: privacy that exists by construction survives changes of management, jurisdiction, and business model, while privacy that exists by configuration is one policy update away from gone. That one test — construction or configuration? — sorts the credential landscape quickly.