jump to navigation

Linux Advanced File Permissions – SUID,SGID and Sticky Bit December 8, 2010

Posted by Tournas Dimitrios in Linux.

My previous article made an introduction on the concepts of file permissions .
After you have worked for a while with Linux you discover probably
that there is much more to file permissions than just the “rwx” bits.
When you look around in your file system you will see “s” and “t”

$ ls -ld /tmp
drwxrwxrwt 29 root root 36864 Mar 21 19:49 /tmp

$ which passwd

$ ls -l /usr/bin/passwd
-rwsr-xr-x 1 root root 22984 Jan 6 2007 /usr/bin/passwd

What is this “s” and “t”  bit?  The vector of permission bits is really 4 * 3 bits long. Yes there are 12 permission bits,not just 9.The first three bits are special and are frequently zero. And you almost always learn about the trailing 9 bits first. Some people stop there and never learn those first three bits , but you are curious to learn all about the permission schema on Linux right ? OK then follow allong .
The fourth permission bit is used only when a special mode of a file needs to be set. It has the value 4 for SUID, 2 for SGID and 1 for the sticky bit. The other 3 bits have their usual significance.
Here we will discuss about the 3 special attributes other than the common read/write/execute:

  • 1.Set-User-Id (SUID)
  • 2.Set-Group-Id (SGID)
  • 3.Sticky Bit

1)Set-User_Id (SUID): Power for a Moment:

By default, when a user executes a file, the process which results in this execution has the same permissions as those of the user. In fact , the process inherits his default group and user identification.
If you set the SUID attribute on an executable file, the process resulting in its execution doesn’t use the user’s identification but the user identification of the file owner.
The SUID mechanism,invented by Dennis Ritchie,is a potential security hazard. It lets a user acquire hidden powers by running such a file owned by root.

$ ls -l /etc/passwd /etc/shadow /usr/bin/passwd
-rw-r–r– 1 root root 2232 Mar 15 00:26 /etc/passwd
-r——– 1 root root 1447 Mar 19 19:01 /etc/shadow

The listing shows that passwd is readable by all, but shadow is unreadable by group and others. When a user running the program belongs to one of these two categories (probably, others), so access fails in the read test on shadow. suppose normal user wants to change his password,
How can he do that? He can do that by running /usr/bin/passwd . Many UNIX/Linux programs have a special permission mode that lets users update sensitive system files –like /etc/shadow –something they can’t do directly with an editor. This is true of the passwd program.

$ ls -l /usr/bin/passwd
-rwsr-xr-x 1 root root 22984 Jan 6 2007 /usr/bin/passwd

The s letter in the user category of the permission field represents a special mode known as the set-user-id (SUID). This mode lets a process
have the privileges of the owner of the file during the instance of the program. Thus when a non privileged user executes passwd, the effective UID of the process is not the user’s, but of root’s – the owner of the program. This SUID privilege is then used by passwd to edit /etc/shadow.

What is effective user-id:

Every process really has two user IDs: the effective user ID and the real user ID. (Of course, there’s also an effective group ID and real group ID.Just about everything that’s true about user IDs is also true about group IDs) Most of the time,the kernel checks only the effective user ID. For example, if a process tries to open a file, the kernel checks the effective user ID when deciding whether to let the process access the file.
Save the following script under the name reids.pl and make it
executable (chmod 755 reids.pl).
# print real UID
print “Real UID: $<\n”;
# print real GID
print “Real GID: $(\n”;
# print effective UID
print “Effective UID: $>\n”;
# print effective GID
print “Effective GID: $)\n”;

check file permissions:
$ ls -l reids.pl
-rwxr-xr-x 1 venu venu 203 Mar 24 10:40 reids.pl

Note: For security reasons the s-bit works only when used on binaries (compiled code) and not on scripts (an exception are perl  scripts).Scripts,i.e. programs that cannot be executed by the kernel directory but need an interpreter such as the Bourne shell or Java,can have their setuid bit set, but it doesn’t have any effect. There are some platforms that honor the s bits even on scripts ( some system V variants, for example), but most systems don’t because it has proven such
a security headache – most interpreters simply aren’t written with much security in mind. Set the SUID bit on shell script is useless, that’s why I am using perl script here.

When you run the script you will see that the process that runs it gets your user-ID and your group-ID:
$ ./reids.pl
Real UID: 500
Real GID: 500 500
Effective UID: 500
Effective GID: 500 500
Note: If you get an error like this:
Can’t do setuid (cannot exec sperl)
In Debian install perl-suid using following command:
apt-get install perl-suid
In Centos install perl-suidperl using following command:
yum install perl-suidperl
Now change owner ship to another user (Do it as an administrator).
# chown king /home/venu/reids.pl
# ls -l /home/venu/reids.pl
-rwxr-xr-x 1 king venu 203 Mar 24 10:40 /home/venu/reids.pl
Now run the script again.
$ ./reids.pl
Real UID: 500
Real GID: 500 500
Effective UID: 500
Effective GID: 500 500
What you observed, the output of the program depends only on the user that runs it and not the one who owns the file.
How to assign SUID permission:
The SUID for any file can be set (mostly by the superuser) with a special syntax of the chmod command. This syntax uses the character s as the permission. Now add SUID permission to the script reids.pl :
# chmod u+s /home/venu/reids.pl (Do it from root account)
Now return from the super user mode to the usual non privileged mode.
$ ls -l reids.pl
-rwsr-xr-x 1 king venu 203 Mar 24 10:40 reids.pl
To assign SUID in an absolute manner, simply prefix 4 to whatever octal string you would otherwise use (like 4755 instead of 755).
The file reids.pl is owned by king and has the s-bit set where normally the x is for the owner of the file. This causes the file to be executed under the user-ID of the user that owns the file rather than the user that executes the file. If venu runs the program then this looks as follows:
$ perl reids.pl
Real UID: 500
Real GID: 500 500
Effective UID: 503
Effective GID: 500 500
Effective user id of process is 503, this is not the venu’s , but of king’s – the owner of the program. As you can see this is a very powerful feature especially if root owns the file with s-bit set. Any user can then do things that normally only root can do.
Caution: When you write a SUID program then you must make sure that it can only be used for the purpose that you intended it to be used.
As administrator, you must keep track of all SUID programs owned by root that a user may try to create or copy. The find command easily locate them:
# find /home -perm -4000 -print | mail root
The extra octal bit (4) signifies the SUID mode, but find treats the “–” before 4000 as representing any other permissions.

2)Set-Group_Id (SGID):
The set-group-id (SGID) is similar to SUID except that a program with SGID set allows the user to have the same power as the group which owns the program. The SGID bit is 2,and some typical examples could be   chmod g+s reids.pl or chmod 2755 reids.pl.
You can remove SGID bit using following commands:
$ chmod g-s reids.pl
$ chmod 755 reids.pl (Absolute manner)
It is really useful in case you have a real multi-user setup where users access each others files. As a single home user I haven’t really found a lot of use for SGID. But the basic concept is the same as the SUID,Similar to SUID, SGID also grants privileges and access rights to the process running the command, but instead of receiving those of the file’s owner it receives those of the file’s group. In other words , the process group owner will be set to the file’s group.
I explain it with an example. I have created two user accounts king and venu with same home directory project. king belongs to king and development groups, venu belongs to venu and development groups.

# groups king venu
king : king development
venu : venu development
venu’s default group is venu and king’s default group is king.
Login as king and create reids.pl file again and make it executable (using chmod 755 reids.pl) .

$ id
uid=503(king) gid=503(king) groups=501(development),503(king)
$ ls -l reids.pl
-rwxr-xr-x 1 king development 203 Mar 25 19:00 reids.pl
Now login as venu and run the program:
$ id
uid=501(venu) gid=504(venu) groups=501(development),504(venu)
$ perl reids.pl
Real UID: 501
Real GID: 504 504 501
Effective UID: 501
Effective GID: 504 504 501
The effective GID of the process is the venu’s,but not of the king’s -the owner of the program.
Now login as king and assign SGID bit to reids.pl program:
$ chmod 2755 reids.pl; ls -l reids.pl
-rwxr-sr-x 1 king development 203 Mar 25 19:00 reids.pl
Now login as venu and run the reids.pl program:
$ perl reids.pl
Real UID: 501
Real GID: 504 504 501
Effective UID: 501
Effective GID: 501 504 501
Real GID and Effective GID are different,here Effective GID is the king’s – the owner of the program.
Set SGID on a directory:
When SGID is set on a directory it has a special meaning. Files created in a directory with SGID set will inherit the same group ownership as the directory itself,not the group of the user who created the file.
If the SGID is not set the file’s group ownership corresponds to the user’s default group.
In order to set the SGID on a directory or to remove it, use the following commands:
$ chmod g+s directory or $ chmod 2755 directory
$ chmod g-s directory or $ chmod 755 directory

As I mentioned earlier venu and king’s home directory is same that is /home/project. I changed group ownership of /home/project directory to development.
# ls -ld /home/project/
drwxrwxr-x 16 root development 4096 Mar 26 00:22 /home/project/
Now login as king and create a temp file.
$ whoami
$ pwd
$ touch temp; ls -l temp
-rw-r–r– 1 king king 0 Mar 26 12:34 temp
You can see from the ls output that the group owner for project is development, and that the SGID bit has not been set on the directory yet. When king creates a file in project, the group for the file is king (king’s primary gid).
Set SGID bit on project directory. For that login as administrator and set SGID bit using following command:
# chmod g+s /home/project/
# ls -ld /home/project/
drwxrwsr-x 15 root development 4096 Mar 26 12:34 /home/project/

From the ls output above, you know the SGID bit is set because of the s in the third position of the group permission set,which replaces the x in the group permissions.
Now login as king and create temp2 file.

$ whoami
$ touch temp2; ls -l temp2
-rw-r–r– 1 king development 0 Mar 26 13:49 temp2

Notice the group ownership for temp2 file. It inherits group permission from the parent directory.
Enabling SGID on a directory is extremely useful when you have a group of users with different primary groups working on the same set of files.
For system security reasons it is not a good idea to set many program’s set user or group ID bits any more than necessary,since this can allow an unauthorized user privileges in sensitive system areas.If the program has a flaw that allows the user to break out of the intended use of the program, then the system can be compromised.

3) Sticky bit:
The sticky bit(also called the saved text bit) is the last permission bit remaining to be discussed. It applies to both regular files and directories. When applied to a regular file, it ensures that the text image of a program with the bit set is permanently kept in the swap area so that it can be reloaded quickly when the program’s turn to use the CPU arrives. Previously, it made sense to have this bit set for programs like vi and emacs. Today,machines with ultra-fast disk drives and lots of cheap memory don’t need this bit for ordinary files and that is also useless.
However, the sticky bit become a useful security feature when used with a directory. The UNIX/Linux system allows users to create files in /tmp, but none can delete files not owned by him. That’s possible because sticky bit set for /tmp directory.
The /tmp directory is typically world-writable and looks like this in a listing:

# ls -ld /tmp
drwxrwxrwt 32 root root 36864 Mar 27 12:38 /tmp

Everyone can read,write and access the directory.The t indicates that only the user (root and owner of the directory,of course) that created a file in this directory can delete that file.
In order to set or to remove the sticky bit, use the following commands:

$ chmod +t directory or $ chmod 1754 directory
$ chmod -t directory or $ chmod 754 directory
Note: 754 permissions for a directory are powerful enough to guard your directories from intruders with malicious intentions, that’s why I used 754 as default,if yow want you can change it.
I logged in as king and created a temp file.
$ whoami
$ pwd
$ touch temp; ls -l
-rw-r–r– 1 king king 0 Mar 27 13:44 temp
Now logged in as venu and try to delete temp file.

$ whoami
$ rm temp
rm: remove write-protected regular empty file `temp’? Y
$ ls temp
ls: temp: No such file or directory

So what happened? venu deleted file owned by king.
Assign sticky bit to the project directory.As a owner of the directory or administrator.
# chmod +t /home/project
# ls -ld /home/project/
drwxrwxr-t 15 root development 4096 Mar 27 13:46 /home/project/
From the ls output above, you know the sticky bit is set because of the t in the third position of the other permission set,which replaces the x in the other permissions.
Now repeat same steps again,then you get the following message:

$ whoami
$ ls -l temp
-rw-r–r– 1 king king 0 Mar 27 17:36 temp
$ rm temp
rm: remove write-protected regular empty file `temp’? y
rm: cannot remove `temp’: Operation not permitted

What programs or scripts with setgid bit set to on are on your Linux box:

# Shell script to find all programs and scripts with setgid bit set on.
# If your system ever cracked (aka hacked) then system has this kind of binary
# installed; besides the normal setuuid scripts/programs
# -------------------------------------------------------------------------
# -------------------------------------------------------------------------
#[ ! -f $INITBMS ] && exit 1 || . $INITBMS

[ $# -eq 1 ] && : || die "Usage: $($BASENAME $0) directory" 1

$FIND $DIRNAME -xdev -type f -perm +g=s -print

Read also this related article ” access control lists ” (acl)


No comments yet — be the first.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s