April 21, 2004

Securing a fresh Linux install, part 2

Author: Mike Peters

In the first article in this series we began looking at ways to secure a new Linux server, starting with locking down services. Next, let's look at securing files and monitoring system logs.SUID and SGID files

SUID and SGID files are executables which, when run by a normal user, may have access to resources not normally available to the user running the program. For example, an SUID program could have the permissions:

-r-sr-xr-x 1 root root 11267 Jan 21 00:28 /usr/sbin/foo

The s in the owner's permission field in place of the usual x indicates that /usr/sbin/foo is SUID. If run by a normal user, the executable will run with the privileges of the owner of the file, in this case root. In this case the program will have access to the same system resources as root.

Below is an example of an SGID file:

-r-xr-sr-x 1 root foo 11267 Jan 21 00:28 /usr/sbin/foo

Here there is an s in the place of the group's executable bit, meaning the file is SGID and will be executed with the group permissions of the foo group.

SGID and SUID programs may be used by a cracker to gain elevated permissions on a system, so you should keep track of such files. You can find SUID and SGID files using find:

# find / -perm -4000 -o -perm -2000 -exec ls -ldb {} \; >> SUID_files.txt

This command finds all SUID or SGID files and lists them in a file called SUID_files.txt. You can unset SUID or SGID privileges with the command chmod -s /usr/sbin/foo, but be warned, unsetting the SUID or SGID bit on some programs may mean that they will no longer run. Periodically check for new files.

There should be no reason for users to have SUID files in their home directories
so you should use the nosuid option in /etc/fstab for the partition containing
users $HOME directories. For example:

/dev/hda3 /home ext3 defaults,nosuid 1 1

World readable/writable files

Files should be world readable or writable only for very good reasons. You should check for such files the way we did above for SGID and SUID files:

# find / \( -perm -a+r -o -perm -a+w \) ! -type l >> world_readwrite.txt

Again, check through the list and remove permissions from files that do not
need to be world readable or writable, and run checks regularly for new world readable or writable files.

Files with no owner or group

Ownerless files can be an indication that someone has gained access to your system. You should check regularly using the command # find / -nouser -o -nogroup. If you find any ownerless files, either delete them, or, if you know what they are and wish to keep them, assign them to an appropriate user and group. For example, assign myfile to the user foo and the group bar you would issue the command # chown foo.bar myfile

Using umask

The umask command can be used to determine the permissions given to newly created files on your system. Addng the line umask 022 to the /etc/profile file tells the system that any files created by users should have the permissions 0644, or -rw-r--r--. This means users must explicitly make a file executable by using chmod in their $HOME directory.

The immutable and append-only bits

With chattr, root can set files to be read-only or append-only. Setting the immutable bit (making a file read-only) ensures that a file cannot be altered, even by root (of course root can remove the immutable bit and then alter the file, so it's not watertight). Setting the append-only bit ensures that the existing contents of a file cannot be changed, only added to. It is a good idea to set the append-only bit on log files: # chattr +a /var/log/messages.

You can set the immutable bit to make it more difficult to replace important executables or change critical configuration files:

# chattr +i /usr/bin/ps
# chattr +i /etc/services

The attributes of files set by chattr can be displayed using the lsattr utility.

System logs

In order to trace any unwanted activity on your computer, you should keep complete and accurate logs. On Linux machines, logging is handled by the syslog daemon, syslogd. Syslogd reads its configuration from the /etc/syslog.conf file. You can set the facilities to be logged, the log priority, and the files in which to log information here. The default values in most distributions do not give you enough information.

A sensible log policy is to log almost everything in /var/log/messages and /var/log/syslog and then to have each individual facility log to its own separate file, as shown in the example below:

--- Begin Example syslog.conf-----

# Log anything 'info' or higher, but lower than 'warn'.
# Exclude mail.  This is logged elsewhere.
*.info;*.!warn;mail.none                                -/var/log/messages

# Log anything 'warn' or higher.
# Exclude mail.  This is logged elsewhere.
*.warn;                                                 -/var/log/syslog

# Debugging information is logged here.
*.=debug                                    	        -/var/log/debug

# Kernel related logs:
kern.*  						-/var/log/kernel

# Private authentication message logging:
authpriv.*                                              -/var/log/secure

# Cron related logs:
cron.*                                                  -/var/log/cron

# Mail related logs:
mail.*                                                  -/var/log/maillog

# Daemon related logs:
daemon.*                                                -/var/log/daemonlog

# User related logs:
user.*                                                  -/var/log/userlog

# Mark logs:
mark.*                                                  -/var/log/marklog

# Emergency level messages go to all users consoles:
*.emerg                                                 *

--- End Example syslog.conf-----

Note the dash before the log files' names. This tells syslogd not to sync after every log. The disadvantage of this is that log information may be lost in the event of a system crash. Removing the dash, however, can cause a performance loss with heavy logging.

If you want to be able to track logs in real time, you can open a log file
using the command tail -f /var/log/messages. Alternatively you can have a permanent log console by adding the line

*.* /dev/tty8

to the end of your syslog.conf file. This displays logs in real time on
/dev/tty8. (Be sure that tty8 exists, of course!)

In order to keep accurate logs, ensure that your system clock is accurate at all times. You should look to using Network Time Protocol (NTP) to maintain your system clock's accuracy. The easiest way to do this is to regularly run ntpdate some.time.server from a cron job.

Finally, although it's not really related to your system logs, make sure you redirect root's mail to your normal user account so you don't miss any important warning mail messages sent to root. You should do this either by placing the line:

root: foo@mydomain.com

in /etc/aliases and running the command newaliases, or, alternatively, create a file named .forward in /root containing the address you want mail to be forwarded to.

In the third and final part of this series we'll look at security considerations for some important networking tools.

Mike Peters is a freelance consultant and programmer and long-time Linux user.

Click Here!