What Is the OWASP Top 10? A Practical Guide to the Most Critical Web Application Security Risks
Understand the OWASP Top 10:2025 and turn its risk categories into practical security checks, with test examples and clear evidence requirements.
On this page
What is the OWASP Top 10?
The OWASP Top 10 is an awareness guide to major web application security risks. It gives developers and QA teams a shared starting point for discussing risk. A test suite covering ten headings is not proof that an application is secure.
This guide uses the official OWASP Top 10:2025 release, checked on 22 October 2026. It concerns web application security; OWASP's separate guidance for LLM applications has a different scope.
The 2025 categories and what to test
The table summarizes the 2025 categories in plain language. Use OWASP's linked release for the official category names and definitions. The test ideas are our editorial recommendations, not an OWASP certification checklist.
| 2025 ID | Risk area, summarized | A practical QA question |
|---|---|---|
| A01 | Authorization and access boundaries | Can account A read, change or delete account B's resources by changing an identifier? |
| A02 | Unsafe security configuration | Are debug endpoints, unnecessary services and permissive defaults absent from the deployed environment? |
| A03 | Failures across the software supply chain | Can the team identify and verify the dependencies and build inputs used for this release? |
| A04 | Weak protection of sensitive data | Is sensitive data protected in transit and storage, and kept out of diagnostic output? |
| A05 | Injection into interpreters | Does untrusted input remain data when it reaches a query, command or rendered page? |
| A06 | Security gaps in design | What prevents a user from abusing a valid workflow, even when every individual request is well formed? |
| A07 | Failures in authentication | Do expired sessions, password-reset flows and repeated sign-in attempts behave as intended? |
| A08 | Untrusted software or data changes | Are updates and externally supplied artifacts checked before the application trusts them? |
| A09 | Missing security logging and alerts | Does an important security event produce a useful, redacted event and reach the responsible team? |
| A10 | Unsafe handling of exceptional conditions | When a dependency fails or an input is unexpected, does the application deny unsafe actions and recover cleanly? |
Why a 2021 checklist needs review
Category identifiers depend on the edition. In the 2025 release, A03 covers supply-chain failures and A10 covers exceptional conditions. Do not carry old category numbers into a new report without checking their meaning against the 2025 release.
When updating an existing test plan, preserve the original scenario and evidence, then remap its risk category. A changed label is not a reason to delete a useful test. Record the edition in issue templates so a finding marked A03 is unambiguous.
Turn a category into an executable check
Choose a real business boundary first. For an authorization check, create two disposable accounts in an environment where you have permission to test. Give each account a resource, authenticate as account A, and attempt the resource operations owned by account B.
Write the expected outcome before execution: the unauthorized request is denied, no protected data is returned, and the resource remains unchanged. Check each relevant operation separately. An interface hiding a button does not establish that the server rejects the request.
Keep a receipt containing the build identifier, environment, actor roles, operation, expected result and observed result. Redact credentials and customer data. If the result fails, retain enough evidence to reproduce it without sharing session tokens.
For exceptional conditions, begin with a controlled staging failure such as an unavailable dependency. Check that the user receives a safe error, no partial privileged change is committed, and a retry cannot duplicate the operation. Agree the expected recovery behavior with the feature owner before treating the test as a release gate.
Fit the checks into a release workflow
Our recommended starting workflow is:
- Map the feature's trust boundaries and sensitive operations.
- Select a few high-consequence negative cases, with a named owner and expected result.
- Automate repeatable checks and retain evidence for manual investigation where needed.
- Run the relevant checks after changes to permissions, authentication, dependencies or deployment configuration.
- Reproduce each failure, fix its cause and rerun the original scenario.
Use SAST, DAST and IAST as complementary inputs to the investigation, rather than interpreting a clean tool report as complete coverage. A release decision should describe what was checked and what remains outside scope.
Sources and scope
The category mapping comes from the OWASP Top 10:2025. The testing workflow and examples above are editorial guidance. They are starting points to adapt to your architecture and threat model; this article makes no compliance or certification claim.
Editorial correction, 22 October 2026: replaced the previous mixture of 2021 and 2025 categories, removed unsupported coverage statistics and unfinished research notes, and added concrete test-planning examples.
Sources
Public pages this article was researched from.
- OWASP Top 10 | OWASP Foundationowasp.org
- What Is the OWASP Top 10? Top 10 Vulnerabilities Explainedwww.securityjourney.com
- What Are the OWASP Top 10 Vulnerabilities? | Veracodewww.veracode.com
- OWASP Top 10 (2025): All 10 Risks with Examplesblog.securelayer7.net
- OWASP Top 10 Vulnerabilities Explainedstrobes.co