Northstar Custody research programme.
Programme overview
- Protocol category
- Custody
- Ecosystems in scope
- EthereumSolana
- In-scope components
- 4 (0 contracts, 4 other)
- Languages
- Rust, Go, TypeScript
Focus areas
Where research is most wanted.
- 01
Policy-engine bypass
- 02
Approval-quorum integrity
- 03
API authentication and tenancy isolation
- 04
Session handling in the approvals console
Scope
In scope.
| Component | Type | Identifier | Notes |
|---|---|---|---|
| Client API | Web / API | api.northstar-custody.example | REST API for transaction requests (test tenant) |
| Approvals console | Web / API | console.northstar-custody.example | Web application with hardware-key approvals |
| Policy engine | Infrastructure | policy.northstar-custody.example | Reachable only through the client API |
| Client SDK | Wallet / SDK | sdk.northstar-custody.example | TypeScript SDK for request signing (programme release) |
Out of scope
- MPC node hosts and physical security
- Other tenants’ data
- Phishing of Northstar staff
- Issues already reported in the research assessment
- Findings that require compromised keys or privileged access the attacker could not obtain
Severity
How findings in this programme are rated.
- Critical
- Signing a transaction that violates policy
- Accessing another tenant’s wallets
- High
- Approving with fewer approvers than the quorum
- Session takeover in the approvals console
- Medium
- Information disclosure of policy configuration
- Low
- Verbose error messages without sensitive data
Rules of engagement
Research within these rules.
- 01Test only against a local fork or the programme’s test deployment. Never interact with production systems or other users’ funds.
- 02Do not run denial-of-service, spam or social-engineering tests, and do not access data that is not yours.
- 03Report privately through the submission form. Keep details confidential until a disclosure date is agreed.
- 04One finding per submission, with a reproducible proof of concept where code is involved.
- 05Use only the test tenant credentials issued for the programme; never test with real customer accounts.
These rules, together with the disclosure policy, define what the programme authorises. Testing outside them is not authorised.
Recognition
Validated findings receive recognition. This programme offers no monetary reward.
Acknowledgement in the report
Validated findings are credited in the protocol’s research assessment or advisory, by name or handle, with your consent.
Recognition ledger
Each validated finding becomes a ledger entry recording programme, vulnerability class, severity and validation date. Entries are published only with the researcher’s consent.
Researcher standing
Standing reflects validated findings over time: Contributor after the first, Recognised after three, Distinguished after a validated critical or high finding with a published write-up.
Research collaboration
Researchers with standing are invited to closed challenge rounds and to co-author research on the vulnerability classes they reported.
Report a finding in this programme.
One finding per submission, with steps to reproduce. Keep details private until a disclosure date is agreed.