Security misconfiguration: what a filter sees of a setting
Security misconfiguration covers systems set up insecurely: defaults left in place, features nobody needs left switched on, errors that reveal too much. It rose to A02 in the OWASP Top 10:2025. The weakness is in configuration, so the fix is too; a request filter sees only the probing that looks for it.
Quick facts
- Category
- A02:2025 Security Misconfiguration, up from fifth place in 2021
- How common
- Every application OWASP's contributors tested had some form of misconfiguration
- Mapped weaknesses
- 16 CWEs, among them CWE-16 configuration and CWE-611 XML external entities
- Where it lives
- Servers, frameworks, libraries, databases and cloud permissions
- Does a WAF stop it?
- Little. A filter can refuse probes for well-known exposed paths and some payloads, such as XML external entities; the setting itself has to change
Figures are from the A02:2025 page of the OWASP Top 10.
What the category covers
OWASP lists the forms it takes.
- Missing hardening. Anywhere in the stack, including cloud permissions set too wide.
- Unneeded features left on. Extra ports, services, pages, accounts, test frameworks or privileges.
- Default accounts still active. With their default passwords unchanged.
- Errors that say too much. Stack traces and detailed messages reach the visitor.
- Security features off after an upgrade. New protections exist but were never enabled.
- Insecure values kept for compatibility. Old behaviour preserved at the cost of safe settings.
- Missing security headers. The server sends no security directives, or sends them with insecure values.
OWASP's attack scenarios for A02:2025 give four concrete security misconfiguration examples.
- Sample applications left on the production server. One of them can be the admin console; if its default accounts were never changed, the attacker logs in with the default password.
- Directory listing not disabled. An attacker lists the directories, downloads the compiled Java classes, decompiles them and reads the code.
- Detailed error messages returned to users. A stack trace can expose sensitive information or component versions with known vulnerabilities.
- Cloud sharing open to the Internet by provider default. Sensitive data held in the cloud storage can then be accessed.
Each of these is a setting, so each is fixed by changing the setting.
What it looks like in a request
A misconfiguration is a state, and a state does not arrive in a request. What arrives is someone looking for it. A forgotten DNS record is one example: subdomain takeover.
OWASP's scenarios show the pattern: a sample application with an administrative console left on a production server, directory listing that lets an attacker download compiled classes, stack traces returned to the visitor. Finding each of these starts with requests for paths and files an ordinary visitor never asks for, which scanners send in bulk. Errors that reveal internals are a separate category, mishandling of exceptional conditions.
One weakness in the category does carry a payload. An XML external entity attack puts a document type declaration that references an outside resource into an XML body, and that structure is recognisable.
What a request filter can and cannot do
- It can slow down the search. Requests for well-known sensitive paths and bulk scanning can be matched and refused, which raises the cost of finding a mistake.
- It can refuse a recognisable payload. XML external entity declarations are one example.
- It cannot change what the server answers. A stack trace returned to one request is not something a filter in front of the request can see.
- It cannot reach what is outside the site. Storage opened to the internet by a cloud default is accessed directly, without passing through your request path.
Of all the categories, this is the one where a filter is most tempting as a substitute and least suited to be one, because the misconfiguration stays in place behind it.
The same limit applies to broken access control, where the flaw is in what the application allows, and in part to authentication failures, where repeated attempts are the visible part.
What fixes it
OWASP's measures are about process as much as settings.
- A repeatable hardening process. New environments come up locked down, and development, test and production are configured the same, each with its own credentials.
- A minimal platform. No unneeded features, components, documentation or samples.
- Configuration reviewed with patches. Including cloud storage permissions.
- Segmented architecture. Components and tenants separated.
- Security directives sent to clients. Security headers, for example, or a CORS policy that trusts any origin.
- Automated verification in every environment. And a central handler that stops detailed errors as a backstop.
- No static secrets in code or pipelines. Short-lived credentials or platform roles instead.
Configuration drift
Most misconfigurations are not chosen; they accumulate. A setting changed by hand during an incident stays changed, a test environment is copied to become production, a new server is built from an older image. Over months, systems that started identical end up different, and nobody can say which one is right.
Keeping configuration in version control and building systems from it, rather than editing them in place, turns that drift into a visible change with an author. Comparing live systems with the stored configuration on a schedule catches what was changed by hand anyway.
A misconfiguration review checklist
Defaults: no default accounts, passwords, sample applications or documentation left on any server. Errors: production shows generic error pages, and stack traces go only to logs. Features: debug modes, directory listing, unused HTTP methods and admin consoles switched off or kept off the internet. Headers: the browser-side protections on security headers in place.
Storage and cloud: no bucket or share readable by everyone unless that is its purpose. Cross-origin rules: no CORS policy that trusts any origin with credentials. Updates: configuration reviewed whenever software is patched, since new versions can add features that are on by default.
Rules that describe it
In the OWASP Core Rule Set, REQUEST-913 detects scanners, REQUEST-920 enforces the HTTP protocol, and REQUEST-930 covers local file inclusion, which includes requests for sensitive files. On the response side, the RESPONSE-95x files look for data leaks such as error output.
Questions
Can a WAF fix security misconfiguration?
No. A WAF can refuse probes for well-known exposed paths and some payloads such as XML external entities, but the misconfiguration stays in place until the setting is changed.
Why did security misconfiguration move up in 2025?
OWASP's 2025 data put it second, up from fifth in 2021; every application in the contributed testing data had some form of it.
Is XXE still in the OWASP Top 10?
Yes. XML external entity processing, CWE-611, is mapped to A02:2025 Security Misconfiguration.
Are missing security headers a misconfiguration?
Yes. OWASP lists servers that send no security headers, or send them with insecure values, as part of A02:2025.
What are examples of security misconfiguration?
Four come from OWASP's A02:2025 scenarios: sample applications left on a production server whose admin console still has its default accounts, directory listing left enabled, detailed error messages such as stack traces returned to users, and cloud storage whose provider default is sharing open to the Internet. A filter can refuse probes for well-known paths, but each of the four is fixed by changing the setting.