fix: harden auth, validation, and sandboxing (audit #12-#23)

[LKSNDRTMLKV]

Summary

Hardening pass across dpp-engine closing the open audit findings (#12–#23). Same posture as the core pass: adversarial/malformed input that previously failed open, panicked, or bypassed a check now fails closed with an explicit error, each with a regression test. The full just check gate is green.

This PR contains only the audit fixes. An in-progress webhooks feature that happened to be in the working tree is deliberately excluded — see "Scope" below.

Fixes by area

Scope enforcement, end-to-end (#23, #21, #20, #18) — the central finding.

• dpp-types (dpp-types: unrecognized API-key scope strings fail open to Admin; retention policy has no lower bound #21): unrecognized API-key scope strings now fail closed to Read (were failing open to Admin); retention policy gains a lower bound.
• dpp-dal (dpp-dal: LIKE-pattern injection exposes arbitrary passports; stale compliance-critical columns; fail-open scope default #18): fixed LIKE-pattern injection that exposed arbitrary passports; dropped stale compliance-critical columns from patch writes; made the scope default fail closed; guarded the resubmit path and rows_affected checks.
• dpp-vault (dpp-vault: Read-scoped API keys can perform full passport-lifecycle writes #20): every mutation handler (create/update/publish/suspend/archive/eol/transfer + operator) now enforces can_write(), so Read-scoped keys can no longer perform lifecycle writes.

dpp-common (#13) — the request-ID middleware now matches its own application/problem+json error content-type (it never did), and oversized error bodies pass through with the id in the header instead of being silently emptied.

dpp-integrator (#12) — steel/aluminium/tyre importers validate the GTIN mod-10 checksum like the battery importer already did; a malformed vault response is no longer recorded as a successful create.

dpp-identity (#14) — internal mTLS endpoints require a shared-secret proxy binding instead of trusting client-supplied headers alone; key rotation aborts on archive failure instead of failing open; removed a dead per-operator DID route.

dpp-resolver (#15) — gtin/dppId are validated at the edge (closes an SSRF / path-traversal surface); the linkset base URL is a fixed constant instead of a tamperable qrCodeUrl-derived value (closes an open redirect).

dpp-seal (#16) — capabilities() reports no sealing support when no QTSP is configured, instead of advertising modes the adapter cannot actually deliver.

dpp-plugin-host (#17) — added the missing Wasm table-growth limiter; sandbox-breach metrics are classified from the trap rather than a spoofable message; the in-force compliance gate is applied on the previously-missed output path; unsigned plugins are refused unless explicitly opted in.

dpp-node (#19) — an EU-registry payload that fails validation is no longer sent anyway; the boot-time operator-country lookup surfaces DB errors instead of baking an empty country in for the process lifetime; fixed a token-expiry underflow panic.

odal CLI (#22) — byte-index slicing on untrusted UTF-8 no longer panics (char-safe truncation); the API secret for odal key use is read from a flag, ODAL_API_SECRET, or a no-echo prompt instead of being accepted as a plaintext argv argument.

dpp-evidence verification — also fixes odal-node/dpp-core#25 (its fix lives here, in dpp-vault): unsigned-but-completed transfer records now fail verification, and a present-but-unresolvable kid returns an accurate diagnosis instead of silently falling back to the primary key. (GitHub won't auto-close a cross-repo issue on merge — close dpp-core#25 manually.)

Scope — webhooks excluded

An unrelated, in-progress webhooks feature (28 modified files + 11 new files) was present in the working tree and is not part of this PR. The commit was assembled to contain only the 42 audit files and verified in isolation — just check was re-run green with the webhooks WIP stashed, so the gate reflects exactly what is committed here. That work continues separately on main.

Deferred follow-ups (not blockers)

• The Docker test-integration tier needs DPP_ALLOW_UNSIGNED_PLUGINS=true in the plugin-host feature-gated tests/benches — a consequence of the dpp-plugin-host: Wasm sandbox limiter gap; spoofable sandbox alerts; missing in-force gate on one output path #17 fail-closed change. The fast just check gate is unaffected.
• dpp-node: EU registry payload validation failure sends anyway; transfer notifications always omit the prior operator #19's full fail-close on invalid registry payloads and the transfer-notification prior-operator both require threading the operator legal-name / from-operator through RegistrySyncPort + RegistrationRequest; left as a follow-up because it interacts with #27's stricter validation (which now rejects an empty operator name/country).

Dependency / merge order

Builds against dpp-core fix/core-audit via the local [patch.crates-io]. Land the core PR and publish dpp-core 0.7.0 first, then this branch can pin dpp-* = 0.7.0 from the registry and drop the patch.

Testing

just check green: fmt-check · clippy --workspace --all-targets -D warnings · debug-check · subjects-check · mod-rs-check · cargo nextest --workspace · audit.

Closes #12, #13, #14, #15, #16, #17, #18, #19, #20, #21, #22, #23

[codacy-production]

Pull Request Overview

This PR is currently incomplete or mislabeled. While the title and description claim to implement extensive security hardening for audit findings #12 through #23—including authentication fail-closed logic, SQL injection prevention, and Wasm sandboxing—the provided code changes only consist of a version update for the 'reqwest' library in Cargo.toml. None of the eight acceptance criteria are met, and the regression tests mentioned in the description are entirely absent from the diff. This PR should not be merged until the promised implementation is included.

About this PR

• The PR description mentions the inclusion of regression tests for each security fix, but no new test files or modifications to existing tests are present in the change set.
• The implementation is missing all security hardening logic described in the PR summary. No code changes related to audit findings #12-#23 (such as mTLS endpoints, Wasm limiters, or GTIN validation) were found in the provided diff.

Test suggestions

• Verify that an unrecognized API-key scope string defaults to 'Read' access.
• Verify that LIKE-pattern injection attempts in dpp-dal passport queries are neutralized.
• Confirm that Read-scoped keys are blocked from calling dpp-vault mutation handlers.
• Validate GTIN mod-10 checksum enforcement for tyre and steel importers.
• Verify Wasm table-growth limits prevent sandbox escapes or resource exhaustion.
• Test that internal mTLS endpoints reject requests lacking the shared-secret proxy binding.

Prompt proposal for missing tests

Consider implementing these tests if applicable:
1. Verify that an unrecognized API-key scope string defaults to 'Read' access.
2. Verify that LIKE-pattern injection attempts in dpp-dal passport queries are neutralized.
3. Confirm that Read-scoped keys are blocked from calling dpp-vault mutation handlers.
4. Validate GTIN mod-10 checksum enforcement for tyre and steel importers.
5. Verify Wasm table-growth limits prevent sandbox escapes or resource exhaustion.
6. Test that internal mTLS endpoints reject requests lacking the shared-secret proxy binding.

_{TIP Improve review quality by adding custom instructions
TIP How was this review? Give us feedback}

[codacy-production]

Not up to standards ⛔

🔴 Issues 4 high

Alerts:
⚠ 4 issues (≤ 0 issues of at least minor severity)

Results:
4 new issues

Category	Results
Security	4 high

View in Codacy

🟢 Metrics 16 duplication

Metric	Results
Duplication	16

View in Codacy

AI Reviewer: first review requested successfully. AI can make mistakes. Always validate suggestions.

_{TIP This summary will be updated as you push new changes.}