Repository navigation
Description
Priority
High — the SCF Audit Bank asks for both pieces with the application, and Tranche 3 needs its audit
User story / Problem statement
The Audit Bank's participant eligibility says projects "must include a STRIDE threat model for the project as part of their application" and "must submit the results of one of the 'self-service tooling' options to include a list of all identified vulnerabilities and a remediation plan for fixing identified critical, high, and medium severity vulnerabilities prior to audit start". The readiness checklist adds a data flow diagram with trust boundaries and data entities, and says the panel reads nothing but the submission.
What the repo has at main (db90b95), and what #442 adds:
- STRIDE. docs: map the threat model to STRIDE #442 maps the 39 threats of
docs/THREAT_MODEL.mdto STRIDE inside that file. It does not follow the shape of Stellar's STRIDE template (four headings,Spoof.N…Elevation.Nids, a remediation table, the five closing questions), and it adds no data flow diagram. - Self-service scan. None. CI runs
cargo-deny(dependencies) and gitleaks (secrets), and neither reads the contracts' code. There is no list of findings and no remediation plan.
Expected outcome
Two documents under docs/, linked from the README, that a reviewer can read without opening anything else: the STRIDE threat model in the template's format, and the scan's findings with a remediation plan.
Acceptance criteria
A. STRIDE in the template's format
-
docs/STRIDE-threat-model.mdhas the template's four headings, the STRIDE reminders table as written, and answers the five closing questions - At least one Mermaid data flow diagram: actors, contracts, the off-chain parts (console, investor app, attester, keeper) and external contracts; every trust boundary a subgraph; every arrow numbered with its step and the call it carries
- Numbered happy-path flows for deposit, request, epoch close, attest, fulfil, claim and a timelocked admin change, so a threat can cite a step
- Each of S, T, R, I, D, E has at least one issue (
Spoof.1,Tamper.1, …), each with its attacker, the step it targets and what it breaks - Each issue points to the O/R/X threat in
docs/THREAT_MODEL.mdthat already treats it, instead of restating it; an issue with no threat there becomes a new threat or a follow-up issue - Each issue has at least one remediation (
Spoof.1.R.1, …) marked In place (with the code or test), Recommended or Accepted (with the reason and the risk that remains) - The Soroban checks are each answered with a link or become an issue or an investigation point:
overflow-checksin the release profile, upgrades throughupdate_current_contract_wasm,__constructorand a second init, TTL and archival,require_authon the right address - What cannot be confirmed from the repo is listed as investigation points (IP.N) in the last section
- The README links the document
B. Self-service scan and remediation plan
- The tool is one of those on Stellar's security tools page that reads code (Scout, Almanax or Certora Sunbeam; the Soroban Security Portal is a database of reports, not a scanner). The choice, the version, the command and the crates scanned are written down
- It runs on every contract in scope (
async-vault,share-token,nav-oracle,compliance,identity-verifier, and the shared crates they use); the raw report is committed -
docs/SECURITY_SCAN.mdlists every finding with its severity, its location and its triage: fixed, false positive (with why) or accepted (with why) - The remediation plan covers each critical, high and medium finding that stays open: the fix, who does it, the issue that tracks it and a date before the audit starts
- Findings already fixed link their PR; open ones have an issue
- The README links the document
Technical notes
- Do the STRIDE part after docs: map the threat model to STRIDE #442 is merged: it builds on that mapping and must not contradict it.
- Do not rewrite
docs/THREAT_MODEL.md. The template document cites its ids. - The rules ask for one tool. A second one is a bonus, not a criterion.
- The readiness checklist (updated a year ago) calls the scan optional; the rules (updated three months ago) require it, and the rules prevail. Treat it as required.
- Out of scope: KYC, the co-payment, whether the multisig authorities count as code in the audit's scope, and filing the application.
Activity