[ previous ] [ Contents ] [ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ A ] [ B ] [ C ] [ D ] [ E ] [ F ] [ G ] [ H ] [ next ]
This chapter introduces some of the most common questions from the Debian security mailing list. You should read them before posting there or else people might tell you to RTFM.
A system is only as secure as its administrator is capable of making it. Debian's default installation of services aims to be secure, but may not be as paranoid as some other operating systems which install all services disabled by default. In any case, the system administrator needs to adapt the security of the system to the local security policy.
For a collection of data regarding security vulnerabilities for many operating
systems, see the
US-CERT stats or
generate stats using the
Database (formerly ICAT) Is this data useful? There are several
factors to consider when interpreting the data, and it is worth noticing that
the data cannot be used to compare the vulnerabilities of one operating system
versus another. Also, keep
in mind that some reported vulnerabilities regarding Debian apply only to the
unstable (i.e. unreleased) branch.
There are not really many differences between Linux distributions, with exceptions to the base installation and package management system. Most distributions share many of the same applications, with differences mainly in the versions of these applications that are shipped with the distribution's stable release. For example, the kernel, Bind, Apache, OpenSSH, Xorg, gcc, zlib, etc. are all common across Linux distributions.
For example, Red Hat was unlucky and shipped when foo 1.2.3 was current, which
was then later found to have a security hole. Debian, on the other hand, was
lucky enough to ship foo 1.2.4, which incorporated the bug fix. That was the
case in the big
problem from a couple years ago.
There is a lot of collaboration between the respective security teams for the
major Linux distributions. Known security updates are rarely, if ever, left
unfixed by a distribution vendor. Knowledge of a security vulnerability is
never kept from another distribution vendor, as fixes are usually coordinated
upstream, or by
CERT. As a
result, necessary security updates are usually released at the same time, and
the relative security of the different distributions is very similar.
One of Debian's main advantages with regards to security is the ease of system
updates through the use of
apt. Here are some other aspects of
security in Debian to consider:
Debian provides more security tools than other distributions, see Security tools in Debian, Chapter 8.
Debian's standard installation is smaller (less functionality), and thus more
secure. Other distributions, in the name of usability, tend to install many
services by default, and sometimes they are not properly configured (remember
Debian's installation is not as limited as OpenBSD (no daemons are active per
default), but it's a good compromise. 
Debian documents best security practices in documents like this one.
The Debian distribution boasts a large and growing number of software packages, probably more than provided by many proprietary operating systems. The more packages installed, the greater the potential for security issues in any given system.
More and more people are examining source code for flaws. There are many advisories related to source code audits of the major software components included in Debian. Whenever such source code audits turn up security flaws, they are fixed and an advisory is sent to lists such as Bugtraq.
Bugs that are present in the Debian distribution usually affect other vendors and distributions as well. Check the "Debian specific: yes/no" section at the top of each advisory (DSA).
Short answer: no.
Long answer: certification costs money (specially a serious security
certification), nobody has dedicated the resources in order to certify Debian
GNU/Linux to any level of, for example, the
Common Criteria. If you
are interested in having a security-certified GNU/Linux distribution, try to
provide the resources needed to make it possible.
There are currently at least two linux distributions certified at different
levels. Notice that some of the CC tests are being integrated into the
Linux Testing Project
which is available in Debian in the
Linux, originally oriented toward other Linux distributions (Red Hat
and Mandrake), it currently works also for Debian. Steps are being taken to
integrate the changes made to the upstream version into the Debian package,
Some people believe, however, that a hardening tool does not eliminate the need for good administration.
One of Debian's great strengths is the wide variety of choice available between packages that provide the same functionality (DNS servers, mail servers, ftp servers, web servers, etc.). This can be confusing to the novice administrator when trying to determine which package is right for you. The best match for a given situation depends on a balance between your feature and security needs. Here are some questions to ask yourself when deciding between similar packages:
Is the software maintained upstream? When was the last release?
Is the package mature? The version number really does not tell you about its maturity. Try to trace the software's history.
Is the software bug-ridden? Have there been security advisories related to it?
Does the software provide all the functionality you need? Does it provide more than you really need?
You will find information in this document to make some services (FTP, Bind) more secure in Debian GNU/Linux. For services not covered here, check the program's documentation, or general Linux information. Most of the security guidelines for Unix systems also apply to Debian. In most cases, securing service X in Debian is like securing that service in any other Linux distribution (or Un*x, for that matter).
If you do not like users connecting to your POP3 daemon, for example, and
retrieving information about your system, you might want to remove (or change)
the banner the service shows to users.  Doing so depends on the software you are running for a
given service. For example, in
postfix, you can set your SMTP
smtpd_banner = $myhostname ESMTP $mail_name (Debian/GNU)
Other software is not as easy to change.
ssh will need to be
recompiled in order to change the version that it prints. Take care not to
remove the first part (SSH-2.0) of the banner, which clients use
to identify which protocol(s) is supported by your package.
The Debian security team cannot possibly analyze all the packages included in Debian for potential security vulnerabilities, since there are just not enough resources to source code audit the whole project. However, Debian does benefit from the source code audits made by upstream developers.
As a matter of fact, a Debian developer could distribute a Trojan in a package, and there is no possible way to check it out. Even if introduced into a Debian branch, it would be impossible to cover all the possible situations in which the Trojan would execute. This is why Debian has a "no guarantees" license clause.
However, Debian users can take confidence in the fact that the stable code has a wide audience and most problems would be uncovered through use. Installing untested software is not recommended in a critical system (if you cannot provide the necessary code audit). In any case, if there were a security vulnerability introduced into the distribution, the process used to include packages (using digital signatures) ensures that the problem can be ultimately traced back to the developer. The Debian project has not taken this issue lightly.
Of course, you can change the default Debian permissions on your system. The current policy regarding log files and configuration files is that they are world readable unless they provide sensitive information.
Be careful if you do make changes since:
Processes might not be able to write to log files if you restrict their permissions.
Some applications may not work if the configuration file they depend on cannot
be read. For example, if you remove the world-readable permission from
smbclient program will not
work when run by a normal user.
FIXME: Check if this is written in the Policy. Some packages (i.e. ftp daemons) seem to enforce different permissions.
As a matter of fact, the same questions stand for any other user. Since Debian's installation does not place any file under that directory, there's no sensitive information to protect there. If you feel these permissions are too broad for your system, consider tightening them to 750. For users, read Limiting access to other user's information, Section 188.8.131.52.
This Debian security mailing list
has more on this issue.
If you are receiving console messages, and have configured
/etc/syslog.conf to redirect them to either files or a special
TTY, you might be seeing messages sent directly to the console.
The default console log level for any given kernel is 7, which means that any message with lower priority will appear in the console. Usually, firewalls (the LOG rule) and some other security tools log lower that this priority, and thus, are sent directly to the console.
To reduce messages sent to the console, you can use
(-n option, see
dmesg(8)), which examines and
controls the kernel ring buffer. To fix this after the next reboot,
Use a lower number for -c if you are still seeing them. A
description of the different log levels can be found in
#define LOG_EMERG 0 /* system is unusable */ #define LOG_ALERT 1 /* action must be taken immediately */ #define LOG_CRIT 2 /* critical conditions */ #define LOG_ERR 3 /* error conditions */ #define LOG_WARNING 4 /* warning conditions */ #define LOG_NOTICE 5 /* normal but significant condition */ #define LOG_INFO 6 /* informational */ #define LOG_DEBUG 7 /* debug-level messages */
Yes and no. Debian comes with some predefined users (user id (UID) < 99 as
/usr/share/doc/base-passwd/README) to ease
the installation of some services that require that they run under an
appropriate user/UID. If you do not intend to install new services, you can
safely remove those users who do not own any files in your system and do not
run any services. In any case, the default behavior is that UID's from 0 to 99
are reserved in Debian, and UID's from 100 to 999 are created by packages on
install (and deleted when the package is purged).
To easily find users who don't own any files, execute the following command (run it as root, since a common user might not have enough permissions to go through some sensitive directories):
cut -f 1 -d : /etc/passwd | \ while read i; do find / -user "$i" | grep -q . || echo "$i"; done
These users are provided by
base-passwd. Look in its
documentation for more information on how these users are handled in Debian.
The list of default users (with a corresponding group) follows:
root: Root is (typically) the superuser.
daemon: Some unprivileged daemons that need to write to files on disk run as
atd, probably others).
Daemons that don't need to own any files can run as nobody.nogroup instead, and
more complex or security conscious daemons run as dedicated users. The daemon
user is also handy for locally installed daemons.
bin: maintained for historic reasons.
sys: same as with bin. However, /dev/vcs* and
owned by group sys.
sync: The shell of user sync is
/bin/sync. Thus, if its password
is set to something easy to guess (such as ""), anyone can sync the
system at the console even if they have don't have an account.
games: Many games are SETGID to games so they can write their high score files. This is explained in policy.
man: The man program (sometimes) runs as user man, so it can write cat pages to
lp: Used by printer daemons.
mail: Mailboxes in
/var/mail are owned by group mail, as explained
in policy. The user and group are used for other purposes by various MTA's as
news: Various news servers and other associated programs (such as
suck) use user and group news in various ways. Files in the news
spool are often owned by user and group news. Programs such as
inews that can be used to post news are typically SETGID news.
uucp: The uucp user and group is used by the UUCP subsystem. It owns spool and configuration files. Users in the uucp group may run uucico.
proxy: Like daemon, this user and group is used by some daemons (specifically,
proxy daemons) that don't have dedicated user id's and that need to own files.
For example, group proxy is used by
runs as user proxy.
Majordomo has a statically allocated UID on Debian
systems for historical reasons. It is not installed on new systems.
Postgresql databases are owned by this user and group.
All files in
/var/lib/postgresql are owned by this user to enforce
www-data: Some web servers run as www-data. Web content should not be owned by this user, or a compromised web server would be able to rewrite a web site. Data written out by web servers, including log files, will be owned by www-data.
backup: So backup/restore responsibilities can be locally delegated to someone without full root permissions.
operator: Operator is historically (and practically) the only 'user' account that can login remotely, and doesn't depend on NIS/NFS.
list: Mailing list archives and data are owned by this user and group. Some mailing list programs may run as this user as well.
irc: Used by irc daemons. A statically allocated user is needed only because
of a bug in
ircd, which SETUID()s itself to a given UID on
nobody, nogroup: Daemons that need not own any files run as user nobody and group nogroup. Thus, no files on a system should be owned by this user or group.
Other groups which have no associated user:
adm: Group adm is used for system monitoring tasks. Members of this group can
read many log files in
/var/log, and can use xconsole.
/usr/adm (and later
/var/adm), thus the name of the group.
tty: TTY devices are owned by this group. This is used by write and wall to enable them to write to other people's TTYs.
disk: Raw access to disks. Mostly equivalent to root access.
kmem: /dev/kmem and similar files are readable by this group. This is mostly a BSD relic, but any programs that need direct read access to the system's memory can thus be made SETGID kmem.
dialout: Full and direct access to serial ports. Members of this group can reconfigure the modem, dial anywhere, etc.
dip: The group's name stands for "Dial-up IP", and membership in dip
allows you to use tools like
wvdial, etc. to dial up a connection. The users in this group
cannot configure the modem, but may run the programs that make use of it.
fax: Allows members to use fax software to send / receive faxes.
voice: Voicemail, useful for systems that use modems as answering machines.
cdrom: This group can be used locally to give a set of users access to a CDROM drive.
floppy: This group can be used locally to give a set of users access to a floppy drive.
tape: This group can be used locally to give a set of users access to a tape drive.
sudo: Members of this group don't need to type their password when using
audio: This group can be used locally to give a set of users access to an audio device.
src: This group owns source code, including files in
can be used locally to give a user the ability to manage system source code.
/etc/shadow is readable by this group. Some programs that
need to be able to access the file are SETGID shadow.
utmp: This group can write to
/var/run/utmp and similar files.
Programs that need to be able to write to it are SETGID utmp.
video: This group can be used locally to give a set of users access to a video device.
staff: Allows users to add local modifications to the system
/home) without needing root privileges.
Compare with group "adm", which is more related to
users: While Debian systems use the private user group system by default (each user has their own group), some prefer to use a more traditional group system, in which each user is a member of this group.
If you have removed a system user and have not made a backup of your
group files you can try recovering from
this issue using
The 'adm' group are usually administrators, and this group permission allows
them to read log files without having to
su. The 'staff' group
are usually help-desk/junior sysadmins, allowing them to work in
/usr/local and create directories in
The default behavior in Debian is that each user has its own, private group. The traditional UN*X scheme assigned all users to the users group. Additional groups were created and used to restrict access to shared files associated with different project directories. Managing files became difficult when a single user worked on multiple projects because when someone created a file, it was associated with the primary group to which they belong (e.g. 'users').
Debian's scheme solves this problem by assigning each user to their own group; so that with a proper umask (0002) and the SETGID bit set on a given project directory, the correct group is automatically assigned to files created in that directory. This makes it easier for people who work on multiple projects, because they will not have to change groups or umasks when working on shared files.
You can, however, change this behavior by modifying
/etc/adduser.conf. Change the USERGROUPS variable to
'no', so that a new group is not created when a new user is created. Also, set
USERS_GID to the GID of the users group which all users will belong
That's just an approach to the problem of being, on one side, security conscious and on the other side user friendly. Unlike OpenBSD, which disables all services unless activated by the administrator, Debian GNU/Linux will activate all installed services unless deactivated (see Disabling daemon services, Section 3.5.1 for more information). After all you installed the service, didn't you?
There has been much discussion on Debian mailing lists (both at debian-devel and at debian-security) regarding which is the better approach for a standard installation. However, as of this writing (March 2002), there still isn't a consensus.
Inetd is not easy to remove since
netbase depends on
the package that provides it (
netkit-inetd). If you want to
remove it, you can either disable it (see Disabling daemon services, Section 3.5.1)
or remove the package by using the
Port 111 is sunrpc's portmapper, and it is installed by default as part of Debian's base installation since there is no need to know when a user's program might need RPC to work correctly. In any case, it is used mostly for NFS. If you do not need it, remove it as explained in Securing RPC services, Section 5.13.
In versions of the
portmap package later than 5-5 you can actually
have the portmapper installed but listening only on localhost (by modifying
identd(port 113) for?
Identd service is an authentication service that identifies the owner of a
specific TCP/IP connection to the remote server accepting the connection.
Typically, when a user connects to a remote host,
inetd on the
remote host sends back a query to port 113 to find the owner information. It
is often used by mail, FTP and IRC servers, and can also be used to track down
which user in your local system is attacking a remote system.
There has been extensive discussion on the security of
list archives). In general,
identd is more helpful on
a multi-user system than on a single user workstation. If you don't have a use
for it, disable it, so that you are not leaving a service open to the outside
world. If you decide to firewall the identd port, please use a reject
policy and not a deny policy, otherwise a connection to a server utilizing
identd will hang until a timeout expires (see
reject or deny
If you have run the command netstat -an and receive:
Active Internet connections (servers and established) Proto Recv-Q Send-Q Local Address Foreign Address State PID/Program name raw 0 0 0.0.0.0:1 0.0.0.0:* 7 - raw 0 0 0.0.0.0:6 0.0.0.0:* 7 -
You are not seeing processes listening on TCP/UDP port 1 and 6. In
fact, you are seeing a process listening on a raw socket for protocols
1 (ICMP) and 6 (TCP). Such behavior is common to both Trojans and some
intrusion detection systems such as
portsentry. If you have these packages simply remove them. If
you do not, try netstat's -p (process) option to see which process
is running these listeners.
Yes, of course. The ports you are leaving open should adhere to your
individual site's policy regarding public services available to other networks.
Check if they are being opened by
inetd (see Disabling
inetd or its services,
Section 3.5.2), or by other installed packages and take the appropriate
measures (i.e, configure inetd, remove the package, avoid it running on
/etc/serviceshelp secure my box?
/etc/services only provides a mapping between a
virtual name and a given port number. Removing names from this file will not
(usually) prevent services from being started. Some daemons may not run if
/etc/services is modified, but that's not the norm. To properly
disable the service, see Disabling daemon
services, Section 3.5.1.
The steps you need to take in order to recover from this depend on whether or
not you have applied the suggested procedure for limiting access to
lilo and your system's BIOS.
If you have limited both, you need to disable the BIOS setting that only allows booting from the hard disk before proceeding. If you have also forgotten your BIOS password, you will have to reset your BIOS by opening the system and manually removing the BIOS battery.
Once you have enabled booting from a CD-ROM or diskette enable, try the following:
Boot-up from a rescue disk and start the kernel
Go to the virtual console (Alt+F2)
Mount the hard disk where your /root is
Edit (Debian 2.2 rescue disk comes with the editor
ae, and Debian
3.0 comes with
nano-tiny which is similar to
/etc/shadow and change the line:
root:asdfjl290341274075:XXXX:X:XXXX:X::: (X=any number)
This will remove the forgotten root password, contained in the first colon separated field after the user name. Save the file, reboot the system and login with root using an empty password. Remember to reset the password. This will work unless you have configured the system more tightly, i.e. if you have not allowed users to have null passwords or not allowed root to login from the console.
If you have introduced these features, you will need to enter into single user
mode. If LILO has been restricted, you will need to rerun
just after the root reset above. This is quite tricky since your
/etc/lilo.conf will need to be tweaked due to the root (/) file
system being a ramdisk and not the real hard disk.
Once LILO is unrestricted, try the following:
Press the Alt, shift or Control key just before the system BIOS finishes, and you should get the LILO prompt.
Type linux single, linux init=/bin/sh or linux 1 at the prompt.
This will give you a shell prompt in single-user mode (it will ask for a password, but you already know it)
Re-mount read/write the root (/) partition, using the mount command.
# mount -o remount,rw /
Change the superuser password with
passwd (since you are superuser
it will not ask for the previous password).
For example, if you want to set up a POP service, you don't need to set up a
user account for each user accessing it. It's best to set up directory-based
authentication through an external service (like Radius, LDAP or an SQL
database). Just install the appropriate PAM library
libpam-mysql), read the documentation
(for starters, see User authentication: PAM,
Section 4.11.1) and configure the PAM-enabled service to use the back end
you have chosen. This is done by editing the files under
/etc/pam.d/ for your service and modifying the
auth required pam_unix_auth.so shadow nullok use_first_pass
to, for example, ldap:
auth required pam_ldap.so
In the case of LDAP directories, some services provide LDAP schemas to be included in your directory that are required in order to use LDAP authentication. If you are using a relational database, a useful trick is to use the where clause when configuring the PAM modules. For example, if you have a database with the following table attributes:
(user_id, user_name, realname, shell, password, UID, GID, homedir, sys, pop, imap, ftp)
By making the services attributes boolean fields, you can use them to enable or disable access to the different services just by inserting the appropriate lines in the following files:
/etc/nss-mysql*.conf:users.where_clause = user.sys =
Many vulnerability assessment scanners give false positives when used on Debian systems, since they only use version checks to determine if a given software package is vulnerable, but do not really test the security vulnerability itself. Since Debian does not change software versions when fixing a package (many times the fix made for newer releases is back ported), some tools tend to think that an updated Debian system is vulnerable when it is not.
If you think your system is up to date with security patches, you might want to use the cross references to security vulnerability databases published with the DSAs (see Debian Security Advisories, Section 7.2) to weed out false positives, if the tool you are using includes CVE references.
A trace of an attack does not always mean that your system has been compromised, and you should take the usual steps to determine if the system is indeed compromised (see After the compromise (incident response), Chapter 11). Even if your system was not vulnerable to the attack that was logged, a determined attacker might have used some other vulnerability besides the ones you have detected.
You might find the following lines in your system logs:
Dec 30 07:33:36 debian -- MARK -- Dec 30 07:53:36 debian -- MARK -- Dec 30 08:13:36 debian -- MARK --
This does not indicate any kind of compromise, and users changing between
Debian releases might find it strange. If your system does not have high loads
(or many active services), these lines might appear throughout your logs. This
is an indication that your
syslogd daemon is running properly.
-m interval The syslogd logs a mark timestamp regularly. The default interval between two -- MARK -- lines is 20 minutes. This can be changed with this option. Setting the interval to zero turns it off entirely.
You might find lines in your logs like:
Apr 1 09:25:01 server su: + ??? root-nobody Apr 1 09:25:01 server PAM_unix: (su) session opened for user nobody by (UID=0)
Don't worry too much. Check to see if these entries are due to
cron jobs (usually
$ grep 25 /etc/crontab 25 9 * * * root test -e /usr/sbin/anacron || run-parts --report /etc/cron.daily $ grep nobody /etc/cron.daily/* find:cd / && updatedb --localuser=nobody 2>/dev/null
If you see entries like these in your logs:
May 1 12:35:25 linux kernel: possible SYN flooding on port X. Sending cookies. May 1 12:36:25 linux kernel: possible SYN flooding on port X. Sending cookies. May 1 12:37:25 linux kernel: possible SYN flooding on port X. Sending cookies. May 1 13:43:11 linux kernel: possible SYN flooding on port X. Sending cookies.
Check if there is a high number of connections to the server using
netstat, for example:
linux:~# netstat -ant | grep SYN_RECV | wc -l 9000
This is an indication of a denial of service (DoS) attack against your system's X port (most likely against a public service such as a web server or mail server). You should activate TCP syncookies in your kernel, see Configuring syncookies, Section 4.18.2. Note, however, that a DoS attack might flood your network even if you can stop it from crashing your systems (due to file descriptors being depleted, the system might become unresponsive until the TCP connections timeout). The only effective way to stop this attack is to contact your network provider.
You might see these kind of entries in your
May 2 11:55:02 linux PAM_unix: (cron) session closed for user root May 2 11:55:02 linux PAM_unix: (cron) session closed for user root May 2 12:00:01 linux PAM_unix: (cron) session opened for user root by (UID=0) May 2 12:00:02 linux PAM_unix: (cron) session closed for user root
These are due to a
cron job being executed (in this example, every
five minutes). To determine which program is responsible for these jobs, check
/etc/crond.daily and root's
There are several steps you might want to take in case of a break-in:
Check if your system is up to date with security patches for published
vulnerabilities. If your system is vulnerable, the chances that the system is
in fact compromised are increased. The chances increase further if the
vulnerability has been known for a while, since there is usually more activity
related to older vulnerabilities. Here is a link to
SANS Top 20 Security
Read this document, especially the After the compromise (incident response), Chapter 11 section.
Ask for assistance. You might use the debian-security mailing list and ask for advice on how to recover/patch your system.
Notify your local
CERT (if it
exists, otherwise you may want to consider contacting CERT directly). This
might or might not help you, but, at the very least, it will inform CERT of
ongoing attacks. This information is very valuable in determining which tools
and attacks are being used by the blackhat community.
By watching the logs (if they have not been tampered with), using intrusion
detection systems (see Set up
Intrusion Detection, Section 10.3),
whois and similar tools (including forensic analysis), you may be
able to trace an attack to the source. The way you should react to this
information depends solely on your security policy, and what you
consider is an attack. Is a remote scan an attack? Is a vulnerability probe
First, take a moment to see if the vulnerability has been announced in public
security mailing lists (like Bugtraq) or other forums. The Debian Security
Team keeps up to date with these lists, so they may also be aware of the
problem. Do not take any further actions if you see an announcement at
If no information seems to be published, please send e-mail about the affected
package(s), as well as a detailed description of the vulnerability (proof of
concept code is also OK), to
This will get you in touch with Debian's security team.
Instead of upgrading to a new release, Debian backports security fixes to the version that was shipped in the stable release. The reason for this is to make sure that the stable release changes as little as possible, so that things will not change or break unexpectedly as a result of a security fix. You can check if you are running a secure version of a package by looking at the package changelog, or comparing its exact (upstream version -slash- debian release) version number with the version indicated in the Debian Security Advisory.
proftpdis vulnerable to a Denial of Service attack.
Add DenyFilter \*.*/ to your configuration file, and for more
portsentry, there are a lot of ports open.
That's just the way
portsentry works. It opens about twenty
unused ports to try to detect port scans.
The security team keeps its list of Frequently Asked Questions at the
Debian Security FAQ.
Please refer to that web page for up to date information.
[ previous ] [ Contents ] [ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ A ] [ B ] [ C ] [ D ] [ E ] [ F ] [ G ] [ H ] [ next ]
Securing Debian ManualVersion: 3.17, built on Sun, 08 Apr 2012 02:48:09 +0000