OWASP Top 10, A05: Injection
OWASP A05: Injection · facts checked 2026-10-04
The facts
- OWASP's 2025 Top 10 moves injection down two places, from third to fifth. (OWASP)
- OWASP describes injection as untrusted input reaching an interpreter, such as a database, browser or command line, which then runs part of that input as commands. (OWASP)
- OWASP says injection had more CVEs than any other category for the 2025 edition, ranging from cross-site scripting, CWE-79, to SQL injection, CWE-89. (OWASP)
- A July 2024 Secure by Design alert from CISA and the FBI says OS command injection has long been avoidable by strictly keeping user input apart from the command being built. (CISA)
The panel
Jordan Pike AI host
OWASP's 2025 Top 10 moves injection down two places, from third to fifth. OWASP describes injection as untrusted input reaching an interpreter, such as a database, browser or command line, which then runs part of that input as commands. OWASP says injection had more CVEs than any other category for the 2025 edition, ranging from cross-site scripting, CWE-79, to SQL injection, CWE-89.
Jordan Pike AI host
A July 2024 Secure by Design alert from CISA and the FBI says OS command injection has long been avoidable by strictly keeping user input apart from the command being built.
Jordan Pike AI host
Dex, walk us through how an attacker would actually use this kind of injection.
Dex Moreno AI analyst
I’d give OS command injection a nine out of ten for elegance—nothing says clever like slipping user data into a shell command.
Jordan Pike AI host
Dex, we are not rating it out of ten.
Dex Moreno AI analyst
An attacker starts with a user-controlled field—say a web form that builds a system command. By injecting a semicolon or back-ticks they break out of the intended string and append their own command. The result is code running on the server, often a reverse shell. Once they have a shell they can harvest credentials, pivot to databases, and reach other services that trust the compromised host. Weaponisation happens fast; public exploit kits appear within days, so you should assume the foothold may already exist and hunt for unexpected processes.
Maya Okafor AI analyst
Injection is a classic input-validation failure. Developers concatenate raw input into queries or command lines, trusting that the data is harmless. The fix is to stop building commands from strings altogether—use parameterised APIs, prepared statements, or safe libraries that treat input as data, not code. Validate, sanitise, and whitelist where possible. In practice you also need to scan for any lingering string-concatenation patterns, because the bug class shows up in many places.
Maya Okafor AI analyst
Injection joins my tally of bugs that should have died in two thousand four.
Dex Moreno AI analyst
From a threat perspective, you want this fixed yesterday—drop a patch or block the vulnerable endpoint now.
Maya Okafor AI analyst
A rushed patch that merely masks the string concatenation leaves the interpreter still reachable; you need a proper sanitisation layer and a full audit of all similar code paths.
Jordan Pike AI host
So: Monday morning? Apply robust input handling, audit your code for any concatenation that reaches an interpreter, and monitor for anomalous command executions.





