#069Technical4 min

Identify security misconfigurations

Recognize settings that widen access, separate context from evidence, and organize a safe configuration review.

Hackaiz team · updated 21/11/2026

Illustration: an engineer works next to servers.
Illustration: an engineer works next to servers. Illustrative photo, not a screenshot.

01Recognize a setting problem rather than a code defect

A security misconfiguration is a setting or permission that gives a system more exposure or power than intended. The application may work perfectly from a user’s perspective while a database role can read too many tables or a connection does not require encryption. Finding the setting is only the beginning: you still need to establish the intended behavior and the environment where it matters.

Use concrete questions rather than a general instruction to “harden everything”. Which role can connect? What objects can it read? Does the account need file capabilities? Are query values written into logs? Hackaiz has specific connected-database checks for such questions. It does not prove that every setting in your infrastructure has been inspected, nor does a clean report certify the deployment.

02Choose a source that can answer the question

For database accounts and server settings, use an authorized database connection in Connect. In “Choose a database”, “Live Postgres and MySQL” provides PostgreSQL and MySQL / MariaDB choices. Supply a dedicated read-only account and confirm ownership. This check queries metadata without reading business table contents or changing records. A private-network host is not reachable through this direct connection flow.

A website address alone does not reveal database grants or internal roles. If your separate review concerns a website or API, domain ownership must first be proved with a DNS TXT record at _hackaiz, an HTTPS file at /.well-known/hackaiz-verification.txt, or by asking the project’s AI to create that file. Do not confuse authorization to connect one database with authorization to analyze another domain.

  1. 1Identify the setting you want to check and the system that owns it.
  2. 2Select the matching database engine and save the authorized read-only connection.
  3. 3Review the selected environment before launching a credit-consuming check. Leave unwanted checks unticked: they are not run or billed.

03Read configuration examples with their limits

On PostgreSQL, a grant to PUBLIC can extend a permission to every database role. A SECURITY DEFINER function executable by PUBLIC deserves a separate review because it runs with its owner’s permissions. Neither result by itself describes what an external visitor can do: the application’s connection identity and reachable paths matter. Record the object name and ask why the broad permission exists.

Logging every PostgreSQL statement is another configuration concern. It may help debugging while also placing query values in logs, so the question includes retention and who can read them. The live audit flags the log_statement setting when it is all; it is not an inspection of every existing log file.

For MySQL, local_infile, file import/export restrictions, broad account hosts and powerful global privileges are distinct checks. A host wildcard is an account setting, not a network scan. A demo database named test, sakila or world is a reason to review deployment hygiene, not proof of a breach. Keep the finding’s actual condition separate from assumptions about consequences.

04Assign, correct and verify without automatic changes

Read the object and explanation alongside the report’s Danger, To fix and To watch levels. Assign the question to the person who owns the configuration, not automatically to whoever wrote the last feature. Capture the intended restriction, the current evidence and a test that would show legitimate use still works after the change. For example, reducing an account’s privileges should be checked against the application tasks that account performs.

Look for checks not run before relying on an absence of findings. Restricted metadata or an unsupported query can leave part of the review unanswered. Ask the administrator for an appropriate review rather than broadening the audit account’s rights indiscriminately. Hackaiz does not apply changes to the connected database. Code corrections offered through GitHub are draft pull requests requiring human review and approval, never automatic merges.

After a tested, approved adjustment, rerun the relevant analysis and compare the same object or setting. A changed result is evidence about that control at that time, not a permanent guarantee. Monitoring is optional and only applies to connected systems; it does not replace ownership of configuration changes. Plan follow-up checks as credit-consuming actions rather than assuming that verification is free.

Frequently asked questions

Was this article helpful?

Related articles