Modern operating systems such as Linux provide measures to control access to their resources. One such way is the possibility to change the user of the process. Hence, the change of privilege level follows.
In this tutorial, we’ll learn about the setuid special file permission and three kinds of user IDs to support it.
2. The Setuid Programs
By default, the process’ user is the user who runs it. Thus, the process is privileged as much as the user is. However, Linux offers the setuid bit functionality, which allows changing the process ownership to the file owner. It is especially useful when the file belongs to root.
Let’s apply this feature to an executable file test_id:
$ ls -all test_id -rwxrwxr-x. 1 joe joe 23984 Apr 11 12:01 test_id
First, let’s change the owner of the file to root with chown:
$ sudo chown root test_id $ ls -all test_id -rwxrwxr-x. 1 root joe 23984 Apr 11 12:01 test_id
Then we should use chmod u+s to set the s bit:
$ sudo chmod u+s test_id $ ls -all test_id -rwsrwxr-x. 1 root joe 23984 Apr 11 12:01 test_id
With ls -all, let’s check that the s bit appears in the place of the regular x executable bit of the owner. Perhaps the best-known setuid programs are sudo and passwd.
3. Three Kinds of Linux User IDs
The operating system manages the process’ ownership with three kinds of user ID:
- RUID – the real user ID identifies the user who runs the program
- EUID – the effective ID tells the kernel about the privileges of the process
- SUID – the saved user ID used when the process changes its UID
Let’s explain that when the executable with the ‘x’ bit starts, the EUID is set to the real one. In contrast, an executable with the special ‘s’ bit has an initial EUID equal to the file’s owner ID.
SUID comes into play when the privileged process drops its privileges to do some ordinary user’s work. Subsequently, the process may ask the kernel to restore its initial level of privilege defined by SUID.
4. How to Check User IDs of Process
We’re going to show the real and effective user of the process with the ps command. Let’s assume that we run our setuid test_id program:
$ ps -C test_id -o pid,tty,ruser,user,cmd PID TT RUSER USER CMD 3250 pts/1 joe root ./test_id
Let’s notice that the real user, listed as RUSER, is different from the effective one shown in the USER column.
Now let’s query the proc pseudo filesystem with grep to obtain information about the process’ UIDs.
$ grep "^Uid" /proc/3250/status Uid: 1000 0 0 0
In detail, the meaning of IDs is, from left to right: real, effective, saved, and filesystem UID.
5. Bash Scripts Security Limitations
Modern Linux distribution doesn’t allow setting the setuid bit on Bash script. We should know that the scripts are prone to attack in many ways.
Because the system environmental variables may be implicitly accessed by the script, modifications to PATH or IFS may lead to running malicious code.
Moreover, an attacker may exploit the time gap between calling the Bash interpreter and executing the script to replace its content.
In this article, we talked about the setuid programs with the ability to do privileged work for ordinary users. Then, we learned how to set the s bit for the executable file. Subsequently, we looked through UIDs used by the kernel to handle setuid programs.
Finally, we listed security objections to turning Bash scripts into such programs.