• Injection

SQL Injection

SQL injection occurs when untrusted input is concatenated into database queries, letting attackers change the query’s meaning — often to read, modify, or delete data.

How attackers abuse it

Classic SQLi uses quote-breaking payloads in login forms, search filters, and URL parameters. Boolean and time-based blind techniques extract data when responses do not show raw SQL errors.

Second-order injection stores malicious input that executes later in a different query path. ORM misuse (raw SQL fragments, unsafe order-by) recreates the same class of bug in modern stacks.

Impact includes authentication bypass, full table exfiltration, and sometimes OS command execution via database features — depending on privileges and engine.

What proof looks like

Proof is a controlled demonstration that attacker input altered query logic: unexpected row sets, auth bypass under a test account, or a safe boolean/time differential documented under rules of engagement. Raw error text alone is a lead, not a shipped alert without validation.

In PandaONE, this class maps to the Injection finding category. Alerts ship only with evidence — never as unverified scanner noise.

How to prevent it

  • Use parameterized queries / prepared statements everywhere; never string-build SQL with user input.
  • Prefer ORM APIs that bind parameters; avoid raw query helpers for untrusted data.
  • Apply least-privilege database accounts per service.
  • Validate types and ranges server-side; do not rely on WAFs as the primary control.
  • Keep database engines patched and disable dangerous extended procedures where unused.

• FAQ

Frequently asked questions

← All vulnerability classes

Hunt this class with proof

Or scan your domain.

Featured on ScrollLaunchFeatured on CodeHype