Published Updated

Test the document journey from upload to final delivery, including requests that should be denied and failures that interrupt processing. A document API security review should cover who can act, which record and version they can reach, which fields they can change, and what happens when a dependency fails.
For example, a caseworker starts an export while assigned to a case. Their access changes before the background job finishes. The download endpoint may pass its own tests while the worker still publishes a file for the old audience. The test needs to follow the job, its output, and its delivery credentials.
OWASP, the Open Worldwide Application Security Project, publishes security guidance for software teams. Its Top 10:2025 is a web-application risk awareness list. Use it to organise coverage, then turn each relevant risk into a concrete test with an expected outcome. The examples below are proposed tests, not reported FormKiQ vulnerabilities or a completed security assessment.
The separate OWASP API Security Top 10:2023 focuses on API risks, including object-, property-, and function-level authorisation, resource consumption, and unsafe use of downstream APIs. Those distinctions are useful for document systems: permission to read a record, change its classification, and run a bulk export are three separate decisions.
Keep the edition and category IDs in the test plan so “A01” and “API1” are not confused. Neither list supplies a complete set of acceptance tests for a deployment. The table maps the 2025 web risks to document-specific examples that a team can adapt.
| 2025 risk | Document test and acceptance condition |
|---|---|
| A01 · Broken Access Control | Use an authorised test account from another case or site to request content, versions, search results, and exports. Verify denied data and metadata stay outside that audience. |
| A02 · Security Misconfiguration | Review exposed endpoints, storage policies, debug output, and production permissions. Confirm the deployed settings match the approved environment configuration. |
| A03 · Software Supply Chain Failures | Trace the running parser, renderer, worker image, and build dependencies to reviewed versions. Exercise the process for identifying an affected component and deploying its fix. |
| A04 · Cryptographic Failures | Check transport and encryption controls for originals, exports, temporary files, and backups. Test intended and unintended identities against the relevant data and key permissions. |
| A05 · Injection | Use controlled fixtures containing unexpected filename, metadata, query, and template input. Verify each consumer treats the input under its own validation and encoding rules. |
| A06 · Insecure Design | Try skipping review, changing an approved version, or requesting excessive processing. Verify workflow preconditions and agreed workload limits hold across interactive and background paths. |
| A07 · Authentication Failures | Test expired, invalid, and inappropriate credentials on each supported route. Confirm user and integration credentials cannot be substituted outside their intended trust and permission scope. |
| A08 · Software or Data Integrity Failures | Alter a test callback or processing result, including its destination reference. Verify the receiving integration authenticates its source and validates the result before accepting or publishing it. |
| A09 · Security Logging and Alerting Failures | Trigger a selected denial and integrity failure. Check correlation, alert delivery, ownership, and restricted log access, while excluding document contents and credentials from routine diagnostics. |
| A10 · Mishandling of Exceptional Conditions | Interrupt an upload, time out a dependency, and return malformed worker output. Verify the job enters an explicit recoverable or failed state without releasing an unauthorised or incomplete result. |
Run these checks in an authorised test environment with controlled data and bounded workloads. Record any excluded route, module, or delivery channel as a coverage gap with an owner. A category marked “covered” should point to an actual test result.
Start with two isolated sites, two cases in one site, ordinary users with different assignments, a reviewer, and the required service identities. Include a normal document, restricted evidence, an earlier version, and a derived preview or export where the workflow uses one.
Record the caller, site, record, version, workflow state, and expected delivery scope.
Change assignment, document reference, governed field, credential, or dependency response.
Inspect the response, stored state, output audience, job status, and operational evidence.
Pair denied cases with allowed controls so a broken environment cannot pass simply by rejecting everything. Check response bodies, search counts and snippets, preview endpoints, exports, and notification payloads as relevant to the application. Also verify that rejected mutations leave the governed record unchanged.
Retain the build and configuration reference, fixture identities, policy revision where applicable, expected result, actual result, and correlation ID. Keep credentials out of the report. The OPA authorisation guide covers policy inputs and enforcement in more detail.
This illustrative test uses an application that checks the requester’s current permission again before releasing an export. That is its configured release rule; it does not assume that every queued operation should be cancelled when an initiating user leaves a team.
A caseworker assigned to Case A requests an export of selected evidence. Check that an unchanged authorised request completes with the intended records and audience.
Pause a second job before release, remove the caseworker’s assignment, and confirm the application’s current permission view has changed. The processing service may still have authority to finish internal work.
Return malformed output or time out the processing dependency. Check that failure handling cannot turn the job into a successful release or include raw document data in an error response.
Resume processing and attempt delivery. The application re-evaluates the audience and withholds release to the removed caseworker. A repeated completion callback must not create an additional unreviewed publication.
Check job status, output storage permissions, search visibility, notifications, and direct retrieval. Separately test any previously issued download credentials against their actual expiry and revocation rules. Record the denial, recovery outcome, and responsible owner.
Changing an application permission and invalidating an issued download credential are different operations to test. The presigned-URL guide explains that delivery boundary. For document processing, test both the scan-to-release rule and preview handling used by the deployed workflow.
OWASP’s A10 exceptional-conditions category addresses unsafe error handling and failure states. For a document pipeline, write down the permitted state after a timeout, missing policy response, unavailable scanner, or invalid downstream result. Bound retries and resource use; assign unresolved work to an observable recovery path.
Test the alert and repair process as well as the initial error. An operator should be able to identify the affected record and job without receiving unnecessary sensitive content. After a fix, rerun the denied case, its allowed control, and the related delivery path.
A useful security test names the actor, action, record, state, and expected result, then follows the effect through processing and delivery. Use the OWASP lists to organise that work and retain evidence of what the deployed application actually did.
The open-source foundation — API-first, deployable into your own AWS account, and free to use. Right for architecture validation and early implementation.
Production-ready editions for departments and complex workflows. Start with a Proof-of-Value deployment or go straight to production.
For governance-heavy environments with residency, sovereignty, assurance, and multi-jurisdiction requirements. Talk to us about the right deployment model.