Field guide

OWASP Top 10:2025: what changed from 2021, and what it means for testing.

Two categories are new in the OWASP Top 10:2025: Software Supply Chain Failures (A03) and Mishandling of Exceptional Conditions (A10). SSRF no longer has its own entry and now sits inside Broken Access Control, and Security Misconfiguration jumps from fifth place to second. This is what moved, what it says about how applications are broken, and what to test for each category.

What changed in OWASP Top 10:2025

The edition is published on the OWASP Top 10:2025 site, which OWASP's project page describes as the most current released version. Neither page gives a release date, so we do not quote one. The ten categories, in order, are: A01 Broken Access Control; A02 Security Misconfiguration; A03 Software Supply Chain Failures; A04 Cryptographic Failures; A05 Injection; A06 Insecure Design; A07 Authentication Failures; A08 Software or Data Integrity Failures; A09 Security Logging and Alerting Failures; and A10 Mishandling of Exceptional Conditions.

The introduction sets out the differences from the 2021 list. Two entries are new in substance, one is absorbed, two are renamed, and four others change position.

2021 versus 2025, side by side

Broken Access Control stays at number one. Security Misconfiguration rises from fifth to second. Cryptographic Failures falls from second to fourth, Injection from third to fifth, and Insecure Design from fourth to sixth. Authentication, integrity and logging keep their places at seven, eight and nine.

The structural changes matter more than the shuffling. A06:2021 Vulnerable and Outdated Components becomes A03:2025 Software Supply Chain Failures, a wider category. A10:2021 Server-Side Request Forgery (SSRF) is gone as a separate entry, folded into A01 Broken Access Control. A10:2025 Mishandling of Exceptional Conditions did not exist in 2021. Two names changed: A07 was Identification and Authentication Failures and is now Authentication Failures, and A09 was Security Logging and Monitoring Failures and is now Security Logging and Alerting Failures. The introduction also notes that OWASP built the list from contributed testing data on over 2.8 million applications, with eight categories chosen from the data and two from a practitioner survey. The 2025 list also writes A08 as "Software or Data Integrity Failures" (2021: "and"); the introduction does not comment on the wording.

What the changes say about how applications are being broken

Three things stand out. First, access control is still where applications fail, and the list now treats server-side request forgery as part of the same problem: a server making a request on a user's behalf is an authorisation question about what that user should be able to reach. OWASP gives Broken Access Control an incidence of 3.73% of tested applications in the introduction.

Second, the weight has moved from "your code" to "everything around your code". Misconfiguration rising to second place and the supply chain entry replacing a components entry both say that the build pipeline, the cloud settings and the dependencies are now as likely to be the weak point as the application logic. The A03 page explicitly covers build tools, CI/CD pipelines, registries and developer workstations, not just known-vulnerable libraries.

Third, failure handling is now a security topic in its own right. The A10 page describes programs that fail to prevent, detect or respond well to unusual conditions, leading to crashes and unexpected behaviour; its example scenarios include sensitive error messages, resource exhaustion and corrupted multi-step transactions. Testers have always found these bugs; the list now names them.

What to test for each of the ten categories

This is a tester's reading, not OWASP's text. For each category, what we would probe and what a genuine finding looks like.

A01 Broken Access Control. Test with two accounts of different privilege and compare. A real finding is a request, replayed with the other identity, that returns data or performs an action it should not: another user's invoice, an admin endpoint reachable by a normal user. The same habit covers SSRF: any feature that fetches a URL on your behalf (webhooks, importers, preview generators) is worth pointing at an address you should not be able to reach. For the API variant of this problem, see our BOLA write-up.

A02 Security Misconfiguration. Test what is exposed by default: debug endpoints, directory listings, permissive cross-origin rules, missing security headers, verbose server banners, storage open to the world. A real finding is something reachable that should not be, with the exact URL or setting that proves it. This is the category where passive checks help most; our security headers tool inspects the response-header part of it, though it is a passive check, not a penetration test.

A03 Software Supply Chain Failures. From outside, you can see a lot less. What you can test is version exposure: components with known vulnerabilities that are reachable, exposed build artefacts, and secrets left in public repositories or client bundles. A real finding names the component, the version, the reachable path and the vulnerability. Pipeline and registry weaknesses usually need a code or configuration review rather than a black-box test.

A04 Cryptographic Failures. Test transport and storage: weak or deprecated TLS versions and ciphers, missing HSTS, sensitive data sent or stored without encryption, sensitive values in URLs or cached responses, passwords hashed with fast algorithms, hard-coded keys in client bundles. A real finding shows the sensitive value in a place or form it should not be in.

A05 Injection. Test every input that reaches an interpreter: SQL, NoSQL, operating-system commands, template engines, LDAP and, in the browser, cross-site scripting. A real finding is a payload that changes behaviour in a way you can reproduce, such as a response that differs under a boolean condition, or script that executes in another user's session. A reflected string is not yet a finding.

A06 Insecure Design. This is flaws in the logic rather than the code: a business flow with no rate or quantity limit, a coupon that can be applied twice, a workflow step that can be skipped, tenants that are not separated by design. Testing means reading the intended flow and then breaking its assumptions. A real finding is a sequence of requests that achieves something the business never intended.

A07 Authentication Failures. Test login, registration, reset and session handling: weak lockout, credential stuffing exposure, predictable session identifiers, sessions that survive logout, a password reset that leaks whether an account exists, multi-factor flows that can be bypassed. A real finding is account access gained without the intended credential or factor.

A08 Software or Data Integrity Failures. Test whether the application trusts data or code it should verify: unsigned update mechanisms, insecure deserialisation, third-party scripts loaded without integrity checks, tamperable client-side state. A real finding is modified data or code being accepted and acted on.

A09 Security Logging and Alerting Failures. Hard to test without cooperation. The useful question is whether your own attack was noticed: did a burst of failed logins or an obvious injection attempt produce a record, and did anyone get alerted? A real finding is an attack that left no usable trace. This usually needs the client's side of the logs to confirm.

A10 Mishandling of Exceptional Conditions. Test what happens when things go wrong: malformed and oversized input, interrupted multi-step operations, concurrent requests, missing parameters. A real finding is a stack trace or database error that leaks internals, a failure that leaves an operation half-complete, or a race that lets a single-use action run twice.

What the OWASP Top 10 is not

It is an awareness document. It is not a standard you can comply with, so "OWASP Top 10 compliant" is not a meaningful claim; nobody certifies it. It is also not a complete test plan. Ten categories cannot enumerate everything that can go wrong in an application, and each category is a bucket of many weaknesses: the 2025 introduction reports 248 mapped CWEs across the ten.

If you need requirements to build against or to verify, use the Application Security Verification Standard, a framework of security requirements for building and testing web applications and services. If you need the detailed method for testing, the Web Security Testing Guide is the reference. The Top 10 is the summary that tells you where to start.

How to use the Top 10 to scope a penetration test

Use it as a coverage checklist, then add what is specific to your application. Ask the tester to show, per category, what was tested, how, and what was out of scope. Be explicit about what is outside the reach of an external test. Supply chain pipelines, logging effectiveness and much of insecure design depend on access to code, configuration and people, so they need a different engagement or extra inputs.

Give testers what makes access control testing possible: at least two accounts per role. Without them, the highest-ranked category cannot be properly tested. Then agree what happens to anything that cannot be proven. For background on the difference between a tester and a scanner, see penetration testing versus vulnerability scanning, and for the wider picture, our web application security guide.

How to read a report that maps findings to the Top 10

A category label is a filing system, not a severity. A finding under A05 can be critical or trivial; the label says what kind of weakness it is, not how much it matters. Read the evidence first: the request, the response, the impact in your environment.

Check which edition the report maps to. Many tools and reports still use the 2021 numbering, so "A03 Injection" in one report and "A03 Software Supply Chain Failures" in another are both correct for their edition. Look at the coverage statement as well. A report with no findings under a category is not evidence that the category was clean unless it also says the category was tested.

Finally, ask whether each finding is proven. A reproducible request is stronger than a version banner. A well-run assessment separates confirmed findings from suspected ones and says which is which.

Where AssurePort fits

Our Web Pentest engine ($399 per scan) uses AI agents to test for OWASP Top 10 classes alongside configuration, exposure and TLS issues, and attaches a reproducible proof of concept to findings it confirms. Findings without proof are labelled unconfirmed. It does not claim to find everything, and our own reports currently map findings to the 2021 numbering. For APIs, our API Pentest engine ($399 per scan) works against the OWASP API Security Top 10, which is a separate list; see our API testing guide.

Frequently Asked Questions

What is new in OWASP Top 10 2025?

Two categories are new: A03 Software Supply Chain Failures, which widens the old vulnerable-components entry, and A10 Mishandling of Exceptional Conditions. SSRF no longer has its own entry and is folded into A01 Broken Access Control, and Security Misconfiguration rises from fifth to second.

Is OWASP Top 10 2025 final?

OWASP's project page calls it the most current released version. Neither that page nor the Top 10:2025 site gives a release date or labels it a candidate, so we describe it as released and do not quote a date.

What are the OWASP Top 10 2025 categories?

In order: Broken Access Control, Security Misconfiguration, Software Supply Chain Failures, Cryptographic Failures, Injection, Insecure Design, Authentication Failures, Software or Data Integrity Failures, Security Logging and Alerting Failures, and Mishandling of Exceptional Conditions.

What happened to SSRF in the OWASP Top 10 2025?

It was A10 in 2021 and is no longer a separate category. According to the 2025 introduction it is now part of A01 Broken Access Control. It is still worth testing for; it has only changed where it is filed.

Can you be compliant with the OWASP Top 10?

No. It is an awareness document, not a standard, and nobody certifies against it. To verify an application against defined requirements use the OWASP Application Security Verification Standard, and for testing method use the Web Security Testing Guide.