SQL injection patterns
A database-related MCP call contains text resembling a recognized SQL injection technique.
Why it is risky
Section titled “Why it is risky”If untrusted input changes the meaning of a database query, it can expose records, alter data, or run operations the caller did not intend.
Example
Section titled “Example”An agent passes a user-supplied filter directly into a database query, where it changes the query logic and returns records outside the intended filter.
This is a fictional example, not an event from a customer.
What Omnodex Detects
Section titled “What Omnodex Detects”Both conditions must match: a database-like MCP server name, matched regardless of case, and selected SQL signatures, such as UNION SELECT or a boolean tautology, in the SQL statement the call runs. Other fields, such as a tool’s documentation text, are not scanned.
| Rule ID | Event category | Tier | Base severity |
|---|---|---|---|
RULE_INPUT_VALIDATION_SQL_INJECTION | input_validation | Community | HIGH |
Base severity is the rule default. A finding involving content saved for later execution may have a lower severity. Your loaded rule version and event details determine the finding you see.
What to check next
Section titled “What to check next”Review the query and the origin of each input. Use parameterized queries and appropriate database permissions; determine whether the server treats the input as data or executable SQL.
False positives
Section titled “False positives”Legitimate SQL reports can use UNION, database maintenance can contain stacked statements, and authorized security tests can deliberately include injection signatures.
False negatives
Section titled “False negatives”An unusually named database server, unsupported signatures, obfuscated SQL, parameter values absent from capture, or vulnerabilities without these patterns can be missed.
Technical details
Section titled “Technical details”Pattern matching does not establish that a database is vulnerable or that a query succeeded. The rule also does not detect every missing-authentication or data-exposure problem in database services.
A finding describes a pattern in captured activity. Review the original call, its result, and the task context before concluding that harm occurred. Absence of a finding does not establish that an operation is safe.