ADVISOR EVIDENCE PORTFOLIO

Zero-Trust Consent Architecture

Dr. Gail-Joon Ahn — Center for Cybersecurity & Trusted Foundations, ASU SCAI

EIP-191 wallet-signed consent · zero-custodial design · bilateral authentication research · documented threat model

EIP-191 + EIP-1193 No Custodial SDKs Phase 5 — Specified Patent Pending 64/063,037 · 64/106,748

EIP-191 Wallet-Signed Consent Flow

Six-step authentication chain on encrypthealth.io. No passwords, no email, no OAuth. The wallet is the pseudonymous identity; its signature is the proof.

1
eth_requestAccounts — The browser discovers wallets through EIP-6963 and requests accounts from the selected EIP-1193 provider. User approves address disclosure.
2
Nonce generation — Server generates cryptographically random nonce bound to wallet address. Prevents replay attacks.
3
personal_sign — EIP-191 plain-language message signed by the wallet private key. Message includes address, nonce, timestamp, and consent text. User reviews and approves in wallet UI.
4
ethers.verifyMessage — Server recovers signer address from signature. If recovered address matches claimed address, control of that address is cryptographically proven. The server also checks the consent text and accepts each nonce once, within 10 minutes; the timestamp is informational.
5
JWT issuance — HS256 JSON Web Token issued with 24-hour TTL. Contains wallet address and role (Guide or Participant). Set as a client-set cookie (SameSite=Lax; not HttpOnly).
6
jose.jwtVerify — Middleware validates JWT on every protected route. Expired or tampered tokens are rejected. No session database — stateless verification.

Zero-Custodial-SDK Design

0
Custodial SDKs
0
Key Custody
0
Third-Party Key Mgmt
1
Trust Boundary

Architecture Decision

Browser-extension wallets connect with zero SDKs or relays (EIP-6963). An optional WalletConnect v2 path (Reown AppKit) is offered for mobile/multi-device access; it routes encrypted messages through the WalletConnect relay, which cannot read session material. No Privy, no third-party key custody. Private keys never leave the user’s device on either path. WalletConnect signing is not yet reliable; desktop browser-extension wallets are the supported path for this proof of concept.

Trust Minimization Rationale

Privy holds key material; OAuth providers control identity revocation. In FSL’s architecture, key custody rests with the user’s wallet alone: the FSL server issues session tokens (HS256 JWT), enforces consent-gated reads, and submits deployer-signed attestations, and Phase 5 moves attestation signing to the two parties. The WalletConnect relay is an accepted operational dependency for mobile access — it transports encrypted messages but holds no keys and cannot read session content. The benefit is zero third-party key custody across all connection paths.

Role-Based JWT Architecture

Two-Tier Credential Model

Sovereign Guide (practitioner) — initiates sessions and records Guide-signed SessionStarted/SessionEnded attestations on the SovereignSession contract through the session app’s Guide controls (on chain, any wallet can start a session, and the contract records the caller as the Guide); manages participant relationships.
Participant — controls own data, grants/revokes access, holds decryption keys for personal records.

JWT Configuration

Cookie: fsl_jwt · Algorithm: HS256 · Max-age: 86400 (24 hours) · SameSite: Lax · Path: /
JWT payload contains wallet address, role, and issuance timestamp. No PII stored in token. Stateless verification — no session table required.

Three-Layer Revocation Model

Layer 1: JWT Expiry

24-hour TTL. Automatic expiration without server-side session invalidation. Forces re-authentication via wallet signature. No stale sessions possible beyond 24-hour window.

Layer 2: Contract Pause

SovereignLedger and BenevolenceFund contracts include pause() capability. Owner can halt all contract operations in response to detected compromise. On-chain circuit breaker.

Layer 3: Ownership Transfer

Five owner-controlled contracts (HNT v2, EHT v2, SovereignLedger v3, BenevolenceFund v3, FSL Sovereign Achievement) expose transferOwnership(), so administrative control can move to a new key in a planned rotation without redeployment. HypnoNeuroNFT’s owner is fixed. The planned multisig migration moves these five contracts to quorum control.

Phase 5: Two-Party Bilateral Consent — Doctoral Research Contribution

Objective: A single wallet-signed consent architecture that simultaneously solves security (zero-trust, no custodial intermediaries), privacy (zero PHI by design), clinical governance (bilateral session consent), and blockchain infrastructure (on-chain attestation + token economics) — unified by one cryptographic primitive. This is the core contribution under U.S. Provisional Patent Application 64/063,037 (single-signer EIP-191 consent, zero-PHI constraint, Sovereign Guide attestation; non-provisional due by May 11, 2027) and 64/106,748 (bilateral consent protocol, filed July 2026). The consent primitive is your domain.

THE DOCTORAL CONTRIBUTION — THE ASK

The doctoral contribution: completing bilateral two-party wallet-signed consent with client-side session encryption. This is the central research problem, covered by a filed provisional patent application (64/106,748, July 2026).

Each party (Participant and Sovereign Guide) produces two signatures — an EIP-191 over a plain-language consent document and an EIP-712 over a structured session attestation. The Guide’s EIP-712 embeds the Participant’s EIP-712, binding both into one enforceable consent state; the session key derives via HKDF from both parties’ EIP-712 signatures, so a session can only be established with both parties’ consent, and the server stores ciphertext only. The on-chain contract engineering and key-derivation design are Dr. Boscovic’s domain; the bilateral consent-primitive formalization is Dr. Ahn’s. Your advisement is the gate — without formal verification that the bilateral consent primitive satisfies verifiability, attributability, revocability, and portability, the architecture cannot be claimed as cryptographically sound for mainnet.

Why this is hard + what I need from your lab

Why this is hard: EIP-191 is single-signer by design — no established pattern for cryptographically binding two independent wallet signatures into one mutually-authorized session-consent primitive. EIP-1271 validates contract-account signatures — wrong shape for two independent EOA signatures. Existing multi-sig wallets (Safe, etc.) solve treasury authorization, not bilateral session consent. The consent topology has no existing formalization.

What I need: Dr. Ahn’s authentication/authorization/access-control expertise (SEFCOM) to formalize the bilateral consent primitive so it is cryptographically sound and access-control-correct. The CLEARS framework’s collaboration-aware authorization model maps directly — bilateral consent is a two-party collaboration context with dynamic privilege expansion (consent grant) and contraction (revocation).

Where I am + deployment status

Single-party attestation is built and exercised: session completion → HNT mint → DB record → on-chain attestation via registerClaim on SovereignLedger. 7 sessions recorded, 3 with on-chain attestation hashes. Today’s attestation is deployer-signed (the server writes to the contract). The participant’s wallet address is recorded but not cryptographically verified on-chain via their own signature. SovereignLedger v4 (0xb24A…CB5C), deployed and source-verified on October 13, 2026, verifies the participant’s EIP-712 signature on-chain; its first claim came from a preview build, and production moves to v4 in a planned switch.

PhaseLabelStatus
1Contract DeploymentDeployed
2Bilateral Consent CaptureScaffolded
3EncryptHealth IntegrationSpecified
4Session AttestationSpecified
5Bilateral Consent with Client-Side Session EncryptionNot implemented

SovereignSession source-verified on Blockscout. Included in the planned non-provisional filing that claims provisionals 64/063,037 + 64/106,748.

Outside-HIPAA-by-Design

Architectural Scope Boundary

FSL is not a covered entity. No protected health information is written on-chain. On-chain attestations contain only wallet addresses, cryptographic hashes, and timestamps. The regulatory load of HIPAA and 42 CFR Part 2 is acknowledged — and the scope boundary is architectural, not compliance-driven.

Design Implication

By holding no PHI, FSL avoids the compliance burden while preserving the ability to operate in behavioral health contexts. Participants control their own records; FSL provides the sovereign infrastructure, not the data custody. This is a deliberate zero-trust design choice: FSL holds pseudonymous participant data in PostgreSQL (no PHI by design). On-chain records contain only wallet addresses, hashes, and timestamps — no personal health information.

Key Paper: CLEARS (CIC 2025)

Citation

Nath, Soneji, Baek, Rubio-Medrano, Ahn. “Towards Collaboration-Aware Resource Sharing in Research Computing Infrastructures.” CIC 2025: 43–53.

What CLEARS Does

CLEARS formalizes collaboration context and uses it to drive dynamic access control — specifically Privilege Expansion (granting access as collaboration deepens) and Privilege Contraction (revoking access as collaboration ends or roles change). It introduces a Collaboration Network construct to model who has access to what, under what collaboration context, across a project lifecycle.

FSL ↔ CLEARS Mapping

CLEARS ConstructFSL Equivalent
Collaboration NetworkSovereign Guide / Participant bilateral relationship
Privilege ExpansionEIP-191 consent grant at session start — participant authorizes Sovereign Guide access
Privilege ContractionParticipant-initiated revocation — IMPLEMENTED. The participant signs a plain-language EIP-191 revocation message (domain-separated, nonce-bound); the server ecrecovers the signer, validates the nonce, and writes revoked_at to the consent grant record. Revoking every grant at once runs from the participant’s signed-in session, without a separate signature. hasActiveGrant enforcement rejects subsequent cross-wallet data reads. On-chain Merkle root anchoring is deployed (MerkleAnchor 0x09f3…168c, msg.sender-keyed). The first participant-anchored root was recorded July 22 2026 (wallet 0x831f…c86e, tx 0xde3b1cb3…56ae7f88, block 11329792). After a leaf-construction fix, the participant re-anchored on August 15 2026 (tx 0x2553cbaa…4dad2a10, block 11497118); MerkleAnchor keeps one root per address, so roots(0x831f) returns the latest anchored root. The mapping is msg.sender-keyed — only the participant can write their own slot. Access is both session-scoped and explicitly revocable on the participant’s terms.
Dynamic lifecycle-aware authorizationSession-bound EIP-191 attestation — access is temporally scoped, not permanent

Where Phase 5 Extends Beyond CLEARS

CLEARS handles WHO can access WHAT under which collaboration context. It does not solve the encryption problem — it does not prevent the infrastructure operator from reading the data. Phase 5 extends the model by adding client-side session encryption so that a shared session can only be established with both parties’ consent, and the server stores ciphertext only. This encryption layer is the engineering contribution under Dr. Boscovic (primary advisor target, Blockchain Research Lab); the bilateral consent primitive that CLEARS formalizes is Dr. Ahn’s domain. Phase 5 is specified — it is the doctoral contribution.

Research Convergence

Why This Matters to the Center for Cybersecurity & Trusted Foundations

Dr. Ahn’s research in access control, formal models for computer security, and distributed-systems security is the formal theoretical foundation for this system. The consent architecture treats the wallet signature as the single source of identity truth, with EIP-191 providing non-repudiation without a certificate authority — a direct application of zero-trust principles to health data governance.

The bilateral consent primitive is the core doctoral contribution — formalizing how two independent wallet signatures bind into one enforceable consent state. Your lab’s work on collaboration-aware authorization (CLEARS) and privilege management provides the formal framework this primitive needs. The client-side session-encryption layer that sits on top of the consent primitive is under Dr. Boscovic (primary advisor target, Blockchain Research Lab). I am asking for your formalization of the consent primitive itself as co-advisor.

Known security gaps are catalogued in a private pre-mainnet checklist and summarized under Known Limitations on the home page; the system is built for security audit.