The dominant architecture of digital identity and reputation is brittle by design: centralised custodians accumulate trust on behalf of users, monetise it opaquely, and offer no meaningful recourse when it fails. Shadowgraph proposes an epistemic refoundation of digital identity — a distributed trust fabric in which belief, uncertainty, and reputation are first-class mathematical objects, dynamically updated through cryptographically verifiable interactions. The system is built around the equation Trust = f(Evidence, Uncertainty, Time) and implements this across six architectural layers: a P2P overlay network organised by reputation-space topology rather than physical addressing; a decentralised public-key infrastructure (DPKI) with social key recovery and zero-knowledge attestation; a self-sovereign identity layer representing identity as an evidential probability distribution rather than a fixed credential; an Evidence-Based Subjective Logic (EBSL) engine that forms the mathematical substrate for belief fusion and uncertainty propagation; a zero-knowledge data marketplace enabling privacy-preserving exchange of provable information; and a predictive intelligence layer using zero-knowledge machine learning (ZKML) and graph neural networks to forecast reputation trajectories and detect anomalies. Shadowgraph distinguishes itself from prior work in decentralised identity (W3C DIDs, Verifiable Credentials) and trust systems (EigenTrust, TrustRank) by treating every edge in the social graph as an evidential assertion with explicit uncertainty structure, and by enforcing that all trust computations remain privately verifiable. The result is a system where reputation is measurable but not ownable, where privacy is absolute yet participation remains verifiable, and where the mathematical structure of belief constitutes the foundation of coordination. Applications include reputation-gated token distribution, verifiable credentials, peer-to-peer credit assessment, and decentralised governance mechanisms.

Whitepaper v2.1 — October 2025.