Deals with the particulars of installing and setting up a secure SUSE Linux Enterprise Server, and additional post-installation processes required to further secure and harden that installation. Supports the administrator with security-related choices and decisions.
- About This Guide
- 1 Common Criteria
- 2 Linux Security and Service Protection Methods
- 2.1 Physical Security
- 2.2 Locking Down the BIOS
- 2.3 Security via the Boot Loaders
- 2.4 Verifying Security Action with
seccheck
- 2.5 Retiring Linux Servers with Sensitive Data
- 2.6 Backups
- 2.7 Disk Partitions
- 2.8 Firewall (iptables)
- 2.9 Security Features in the Kernel
- 2.10 AppArmor
- 2.11 SELinux
- 2.12 FTP,
telnet
, andrlogin
(rsh) - 2.13 Removing Unnecessary Software Packages (RPMs)
- 2.14 Patching Linux Systems
- 2.15 Securing the Network—Open Network Ports Detection
- 2.16
xinetd
Services - Disabling - 2.17 Securing Postfix
- 2.18 File Systems: Securing NFS
- 2.19 Copying Files Using SSH Without Providing Login Prompts
- 2.20 Checking File Permissions and Ownership
- 2.21 Default umask
- 2.22 SUID/SGID Files
- 2.23 World-Writable Files
- 2.24 Orphaned or Unowned Files
- 2.25 Restricting Access to Removable Media
- 2.26 Various Account Checks
- 2.27 Enabling Password Aging
- 2.28 Stronger Password Enforcement
- 2.29 Leveraging an Effective PAM stack
- 2.30 Restricting
root
Logins - 2.31 Setting an Inactivity Timeout for Interactive Shell Sessions
- 2.32 Preventing Accidental Denial of Service
- 2.33 Displaying Login Banners
- 2.34 Miscellaneous
Copyright © 2006–2024 SUSE LLC and contributors. All rights reserved.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or (at your option) version 1.3; with the Invariant Section being this copyright notice and license. A copy of the license version 1.2 is included in the section entitled “GNU Free Documentation License”.
For SUSE trademarks, see https://www.suse.com/company/legal/. All third-party trademarks are the property of their respective owners. Trademark symbols (®, ™ etc.) denote trademarks of SUSE and its affiliates. Asterisks (*) denote third-party trademarks.
All information found in this book has been compiled with utmost attention to detail. However, this does not guarantee complete accuracy. Neither SUSE LLC, its affiliates, the authors nor the translators shall be held liable for possible errors or the consequences thereof.
About This Guide #
The SUSE Linux Enterprise Server Security and Hardening Guide deals with the particulars of installation and set up of a secure SUSE Linux Enterprise Server and additional post-install processes required to further secure and harden that installation. Security and hardening elements and procedures are best applied to a server both during installation and post-installation and aim to improve the fitness of the system for the purposes demanded by its administrator.
This guide supports administrator in making security related choices and decisions. The individual steps and procedures should be seen as proposals, not as strict rules. You will often need to evaluate the usefulness of measures for your organization yourself.
The objective is to improve the security value of the system. Definitions about the meaning of the term security vary, but we want to settle on one that is both simple and abstract:
A good system does what it is expected to do, and it does it well.
A secure system is a good system that does nothing else.
The focus of this guide lies on doing “nothing else”. The Linux system is constructed in such way that security policies are enforced. These policies consist of the following concepts (fairly generic and incomplete list):
DAC (Discretionary Access Control): File and directory permissions, as set by
chmod
andchown
.Privileged ports: TCP and UDP ports 0-1023 and raw sockets can only be used by
root
.Other privileged operations: Loading kernel modules, configuring network interfaces, all security relevant settings of the Linux kernel. These are operations that can only be done by the root user, that is the user with the user ID 0, or any other process with the necessary capabilities.
Attacking a system means to attempt to overcome privilege boundaries, for example by circumventing or breaking them. That means the administrator or programmer of the system has not anticipated this scenario.
A hardened system raises the bar by reducing the area that the system exposes to the attacker (often called attack surface). A hardened system can also provide measures to reduce the impact of vulnerabilities in the parts of the systems that must be exposed to a potential attacker.
Security is about decisions, and whenever security is in (apparent) opposition to function, these decisions become trade-offs. While it can be argued that all systems should be set up to be as securely as possible, some levels of security and hardening may very well be overkill in some cases. Each system's operational environment has its own security requirements derived from business drivers or regulatory compliance mandates. SUSE Linux Enterprise Server can, for example, be configured to comply with security standards, such as SOX, HIPAA and PCIDSS. It can also be set up to fulfill the requirements from the German Federal Office of Information Security (Bundesamt für Sicherheit in der Informationstechnik) as described in BSI TR-02102-1. An effective business requirements analysis should be performed to determine the right level of security and hardening to be applied to a server or defined as part of a baseline server build.
As a final note before we begin: You may encounter individual requirements in regulatory compliance frameworks that may not make sense from a technical perspective, or they do not serve the purpose of improving security. It may be a productive attitude to simply implement what is required, but whenever there is a contradiction to security, an informed discussion in the documentation serves the overall purpose of your regulative compliance framework much more than blindly obeying the specifications. Feel encouraged to dispute list items that you think are counterproductive.
1 Assumptions and Scope #
References in this document will usually be made to a single server target or host, however the scope can generally be applied to more than one machine. We generally assume that the security target can cover one or more systems running SUSE Linux Enterprise Server.
We explicitly do not make any assumptions about the hostility of the network that the systems are connected to, or the cooperative nature of the users that leverage the services provided by the systems.
In turn, this means that you partially define your context on your own when reading through this document. You will need to broaden the meaning of individual portions to adapt it to your environment. In some cases, such as the use case of a server that is exposed to the Internet, this document may even be insufficient or incomplete; however, it may still serve as a good starting point on your journey toward an increased level of confidence that your system will behave like you want it to.
About trust: Trust relationships exist among all systems that participate in networked transactions. In this way, the trust relationship between the people that use the systems is transported across these systems. The chain that is formed by your trust relationships is only as strong as the weakest link. It is good practice to graphically visualize the trust relationships with the services in a schematic overview or map of your network. Generally, it is up to the owner of a resource to enforce the policies imposed on that resource; this would usually be the server that provides the resource. The client that opens a connection to request the resource can only be made responsible for the actions that it performs. This refers to the action of opening the connection to start with, but to nothing else as such.
The case of hostile users is special and unique: The Human Resources department may be able to solve some security problems in your computing environment; in addition, some technical measures can be taken. Make sure that the necessary regulations in your environment fit your needs, and that they back your intentions instead of obstructing them if you need to work around a missing support from your HR department (and your management).
Persons that have administrative privileges on a system are automatically considered trusted.
A Linux system—without any additional security frameworks such as SELinux—is a single level security system: From a security policy perspective there is only the superuser (root) and non-privileged users. System users are non-root user IDs that have access to files specific to their purpose. The separation of administrative duties is complicated by this simplicity. Some tools help: Use sudo(8) for administrative tasks, but be aware that after the privilege boundary is crossed, a program running with root privileges does not enforce any file access policies for non-privileged users anymore. vi(1) that runs as root can read and write to any file in the system.
Another tool to mitigate the risk of abuse or accidental misuse of administrative privileges is NetIQ's Privileged User Manager product. More information is available here:
Physical security of the server is another assumption made here, where the server is protected from theft and manipulation by unauthorized persons. A common sobering thought among security professionals is the “ten-second Denial of Service”: Unplug the wires and reboot the server. Physical security must be ensured and physical access must be controlled. Otherwise, all assumptions about at least the availability of these systems are void.
The use of cryptography to protect the confidentiality of transactions with the services that your system provides is generally encouraged. The need to implement cryptographic enhancements is strongly dependent on the operational environments of all participating systems. Keep in mind that you need to verify all of the possible security benefits that cryptography can provide, for all of your services, and that these benefits are not delivered automatically by turning on the “encrypt” option of your service (if you can enjoy the idyllic situation where encryption is available as a button to check):
- Confidentiality
Protection against reading the content of a transaction
- Privacy
Protection against knowing that a transaction exists, and some properties that it may have, such as size, identities of involved parties, their presence, etc.
- Integrity
Protection against alteration of content. Be aware that cryptography does not automatically provide this kind of protection.
- Authenticity
Protection against identity fraud. Cryptography that does not know about identities of participating entities cannot deliver this value.
Keep in mind that encryption of data for confidentiality purposes can merely reduce the size of the data to protect from the actual size to the size of the key that is used to encrypt the data. This results in a key exchange problem for encrypted transactions, and in a key management problem for encrypted data storage. Since data is (typically, there are exceptions!) processed in clear, you need your vault unlocked while data within is being worked with. The encryption of such data on the file system or block device layer helps against the theft of the system, but it does not help the confidentiality of the data while the system is running.
If you want to implement a consistent security policy covering multiple hosts on a network then organizational procedures must ensure that all those hosts can be trusted and are configured with compatible security configurations enforcing an organization wide security policy. Isolation of groups of systems that maintain data of the same trust domain can provide an adequate means of control; ultimately, the access controls to these systems, both for end users and for other systems, need to be carefully designed, configured, inspected and monitored.
Data can only be trusted to the degree that is associated with the domain it comes from. If data leaves the domain in which security policies can be enforced, it should consequently be associated with the trust of the target domain.
For a review of industry best practices on security, the development of sound security processes, controls, development, reviews, audit practices and incident management, you can review a public RFC (request for comments). RFC 2196 is the ongoing work of the world-wide community and individual security and process experts. You can review it online here: http://www.faqs.org/rfcs/rfc2196.html. An RFC is an open and living document that invites comments and review. Enhancements and improvements are welcome; you will find instructions on where to send those suggestions within the document itself.
This guide provides initial guidance on how to set up and secure a SUSE Linux Enterprise Server installation but it is not intended to be the only information required for a system administrator to learn how to operate Linux securely. Assumptions are made within this guide that the reader has knowledge and understanding of operating security principles in general, and of Linux administrative commands and configuration options in particular.
2 Contents of this Book #
Chapter 1, Common Criteria contains a reference to Common Criteria and SUSE Linux Enterprise Server. Chapter 2, Linux Security and Service Protection Methods contains more general system security and service protection schemes.
3 Available documentation #
- Online documentation
Our documentation is available online at https://documentation.suse.com. Browse or download the documentation in various formats.
Note: Latest updatesThe latest updates are usually available in the English-language version of this documentation.
- SUSE Knowledgebase
If you have run into an issue, also check out the Technical Information Documents (TIDs) that are available online at https://www.suse.com/support/kb/. Search the SUSE Knowledgebase for known solutions driven by customer need.
- Release notes
For release notes, see https://www.suse.com/releasenotes/.
- In your system
For offline use, the release notes are also available under
/usr/share/doc/release-notes
on your system. The documentation for individual packages is available at/usr/share/doc/packages
.Many commands are also described in their manual pages. To view them, run
man
, followed by a specific command name. If theman
command is not installed on your system, install it withsudo zypper install man
.
4 Improving the documentation #
Your feedback and contributions to this documentation are welcome. The following channels for giving feedback are available:
- Service requests and support
For services and support options available for your product, see https://www.suse.com/support/.
To open a service request, you need a SUSE subscription registered at SUSE Customer Center. Go to https://scc.suse.com/support/requests, log in, and click .
- Bug reports
Report issues with the documentation at https://bugzilla.suse.com/.
To simplify this process, click the
icon next to a headline in the HTML version of this document. This preselects the right product and category in Bugzilla and adds a link to the current section. You can start typing your bug report right away.A Bugzilla account is required.
- Contributions
To contribute to this documentation, click the
icon next to a headline in the HTML version of this document. This will take you to the source code on GitHub, where you can open a pull request.A GitHub account is required.
Note:only available for EnglishThe
icons are only available for the English version of each document. For all other languages, use the icons instead.For more information about the documentation environment used for this documentation, see the repository's README.
You can also report errors and send feedback concerning the documentation to <doc-team@suse.com>. Include the document title, the product version, and the publication date of the document. Additionally, include the relevant section number and title (or provide the URL) and provide a concise description of the problem.
5 Documentation conventions #
The following notices and typographic conventions are used in this document:
/etc/passwd
: Directory names and file namesPLACEHOLDER: Replace PLACEHOLDER with the actual value
PATH
: An environment variablels
,--help
: Commands, options, and parametersuser
: The name of a user or grouppackage_name: The name of a software package
Alt, Alt–F1: A key to press or a key combination. Keys are shown in uppercase as on a keyboard.
AMD/Intel This paragraph is only relevant for the AMD64/Intel 64 architectures. The arrows mark the beginning and the end of the text block.
IBM Z, POWER This paragraph is only relevant for the architectures
IBM Z
andPOWER
. The arrows mark the beginning and the end of the text block.Chapter 1, “Example chapter”: A cross-reference to another chapter in this guide.
Commands that must be run with
root
privileges. You can also prefix these commands with thesudo
command to run them as a non-privileged user:root #
command
tux >
sudo
command
Commands that can be run by non-privileged users:
tux >
command
Commands can be split into two or multiple lines by a backslash character (
\
) at the end of a line. The backslash informs the shell that the command invocation will continue after the line's end:tux >
echo
a b \ c dA code block that shows both the command (preceded by a prompt) and the respective output returned by the shell:
tux >
command
outputNotices
Warning: Warning noticeVital information you must be aware of before proceeding. Warns you about security issues, potential loss of data, damage to hardware, or physical hazards.
Important: Important noticeImportant information you should be aware of before proceeding.
Note: Note noticeAdditional information, for example about differences in software versions.
Tip: Tip noticeHelpful information, like a guideline or a piece of practical advice.
Compact Notices
Additional information, for example about differences in software versions.
Helpful information, like a guideline or a piece of practical advice.
6 Support #
Find the support statement for SUSE Linux Enterprise Server and general information about technology previews below. For details about the product lifecycle, see https://www.suse.com/lifecycle.
If you are entitled to support, find details on how to collect information for a support ticket at https://documentation.suse.com/sles-15/html/SLES-all/cha-adm-support.html.
6.1 Support statement for SUSE Linux Enterprise Server #
To receive support, you need an appropriate subscription with SUSE. To view the specific support offers available to you, go to https://www.suse.com/support/ and select your product.
The support levels are defined as follows:
- L1
Problem determination, which means technical support designed to provide compatibility information, usage support, ongoing maintenance, information gathering and basic troubleshooting using available documentation.
- L2
Problem isolation, which means technical support designed to analyze data, reproduce customer problems, isolate a problem area and provide a resolution for problems not resolved by Level 1 or prepare for Level 3.
- L3
Problem resolution, which means technical support designed to resolve problems by engaging engineering to resolve product defects which have been identified by Level 2 Support.
For contracted customers and partners, SUSE Linux Enterprise Server is delivered with L3 support for all packages, except for the following:
Technology previews.
Sound, graphics, fonts, and artwork.
Packages that require an additional customer contract.
Some packages shipped as part of the module Workstation Extension are L2-supported only.
Packages with names ending in -devel (containing header files and similar developer resources) will only be supported together with their main packages.
SUSE will only support the usage of original packages. That is, packages that are unchanged and not recompiled.
6.2 Technology previews #
Technology previews are packages, stacks, or features delivered by SUSE to provide glimpses into upcoming innovations. Technology previews are included for your convenience to give you a chance to test new technologies within your environment. We would appreciate your feedback. If you test a technology preview, please contact your SUSE representative and let them know about your experience and use cases. Your input is helpful for future development.
Technology previews have the following limitations:
Technology previews are still in development. Therefore, they may be functionally incomplete, unstable, or otherwise not suitable for production use.
Technology previews are not supported.
Technology previews may only be available for specific hardware architectures.
Details and functionality of technology previews are subject to change. As a result, upgrading to subsequent releases of a technology preview may be impossible and require a fresh installation.
SUSE may discover that a preview does not meet customer or market needs, or does not comply with enterprise standards. Technology previews can be removed from a product at any time. SUSE does not commit to providing a supported version of such technologies in the future.
For an overview of technology previews shipped with your product, see the release notes at https://www.suse.com/releasenotes.
1 Common Criteria #
Common Criteria is the best known and most widely used methodology to evaluate and measure the security value of an IT product. The methodology aims to be independent, as an independent laboratory conducts the evaluation, which a certification body will certify afterward. Security Functional Requirements (SFR) are summarized in so-called Protection Profiles (PP). If the definition of a Security Target (ST) and the Evaluation Assurance Levels (EAL) are comparable, this allows the comparison of security functions of different products. (The definition of a Security Target typically references the PP—if one exists that fits the purpose of the product.)
1.1 Introduction #
A clear definition of security in IT products is challenging. Security should be considered a process that never ends, not a static condition that can be met or not. A Common Criteria certificate (below EAL7) does not make a clear statement about the error-proneness of the system, but it adds an important value to the product that cannot be described with the presence of technology alone: That someone has independently inspected the design of the system in such way that it corresponds to the claims that are made, and that explicit care has been taken in producing and maintaining the product.
The certificate states a degree of maturity of both the product with its security functions and the processes of the company that has designed, built and engineered the product, and that will maintain the product across its lifecycle. As such, Common Criteria aims to be fairly holistic with its approach to take everything into account that is relevant for the security of an IT product.
1.2 Evaluation Assurance Level (EAL) #
The Evaluation Assurance Level denotes the degree of confidence that the product fulfills the described claims. The levels are from 1 through 7:
EAL1: Functionally tested
EAL2: Structurally tested
EAL3: Methodically tested and checked
EAL4: Methodically designed, tested and reviewed
EAL5: Semi-formally designed and tested
EAL6: Semi-formally verified design and tested
EAL7: Formally verified design and tested
While EAL1 only provides basic assurance for products to meet security requirements, EAL2 to 4 are medium assurance levels. EAL5-EAL7 describe medium-to-high and high assurance. EAL4 is expected to be the highest level of assurance that a product can have if it has not been designed from the start to achieve a higher level of assurance.
1.3 Generic Guiding Principles #
Much of the advice in this guide is based on the following guidelines. Consider them when defining your own security processes or deciding about configurations that are not explicitly covered here.
- Use Data Encryption Whenever Possible
Refer to the About This Guide section of this guide. In Section 1, “Assumptions and Scope”, the limitations of cryptography are briefly outlined.
Be aware that cryptography is certainly useful, but only for the specific purposes that it is good for. Using cryptography is not a generic recipe for better security in a system, its use may even impose additional risk on the system. Make informed decisions about the use of cryptography, and feel obliged to have a reason for your decisions. A false sense of security can be more harmful than the weakness itself.
SUSE Linux Enterprise Server supports encryption for:
Network connections (the
openssl
command,stunnel
), for remote login (openssh
,man ssh(1)
)Files (
gpg
)Entire file systems at block layer (
dm-crypt
,cryptsetup
)VPN (
ipsec
,openvpn
)
- Minimal Package Installation
It is useful to restrict the installed packages in your system to a minimum. Binaries not installed cannot be executed.
During installation of the system, you can limit the set of packages that is installed. For example, you can deselect all packages and select only those that you want to use. For example, the selection of the
apache2-mod_perl
package in YaST would automatically cause all packages to be selected for installation that are needed for the Apache package to operate. Dependencies have often been artificially cut down to handle the system's dependency tree more flexibly. You can chose the minimal system, and build the dependency tree from there with your (leaf) package selection.- Service Isolation—Run Different Services on Separate Systems
Whenever possible, a server should be dedicated to serving exactly one service or application. This limits the number of other services that could be compromised if an attacker can successfully exploit a software flaw in one service (assuming that flaw allows access to others).
The use of AppArmor for services that are provided on a system is an effective means of containment. For more information, see Book “Security and Hardening Guide” and the man page of
apparmor
.The use of virtualization technology is supported with SUSE Linux Enterprise Server. While virtualization is generally designed for server consolidation purposes, it is also usefulness for service isolation. However, virtualization technology cannot match or substitute the separation strength that is given by running services on different physical machines! Be aware that the capability of the hypervisor to separate virtual machines is not higher or stronger than the Linux kernel's capability to separate processes and their address spaces.
- System Fingerprinting and Backups
Doing regular backups and having a fingerprint of your system is vital, especially in the case of a successful attack against your system. Make it an integral part of your security routine to verify that your backups work.
A fast and directly accessible backup adds confidence about the integrity of your system. However, it is important that the backup mechanism/solution has adequate versioning support so that you can trace changes in the system. As an example: The installation times of packages (
rpm
-q
--queryformat='%{INSTALLTIME} %{NAME}\n'
PACKAGE NAME) must correspond to the changed files in the backup log files.Several tools exist on SUSE Linux Enterprise Server 12 SP5 which can be used for the detection of unknown, yet successful attacks. It does not take much effort to configure them.
In particular, we recommend using the file and directory integrity checker
AIDE
(Advanced Intrusion Detection Environment). When run for initialization, it creates a hash database of all files in the system that are listed in its configuration file. This allows verifying the integrity of all cataloged files at a later time.Warning: BackdoorsIf you use AIDE, copy the hash database to a place that is inaccessible for potential attackers. Otherwise, the attacker may modify the integrity database after planting a backdoor, thereby defeating the purpose of the integrity measurement.
An attacker may also have planted a backdoor in the kernel. Apart from being very hard to detect, the kernel-based backdoor can effectively remove all traces of the system compromise, so system alterations become almost invisible. Consequently, an integrity check needs to be done from a rescue system (or any other independent system with the target system's file systems mounted manually).
Be aware that the application of security updates invalidates the integrity database.
rpm
-qlv packagename
lists the files that are contained in a package. The RPM subsystem is very powerful with the data that it maintains. It is accessible with the--queryformat
command line option. A differential update of integrity database with the changed files becomes more manageable with some fine-grained usage of RPM.
1.4 For More Information #
The Common Criteria evaluations inspect a specific configuration of the product in an evaluated setup. How to install and configure the reference system that was used as baseline in the Common Criteria evaluation is documented in the “Administrator's Guide” part of the Common Criteria evaluation documentation.
However, it would be incorrect to understand the evaluated configuration as a hardened configuration. The removal of setuid bits and the prescription of administrative procedures after installation help to reach a specific configuration that is sane. But this is not sufficient for a hardening claim.
For more information about SUSE Linux Enterprise Server security certifications and features, see https://www.suse.com/support/security/certifications/.
Find a list of SUSE security resources at https://www.suse.com/support/security/.
Apart from the documentation that comes with the Common Criteria effort, see also the following manual pages:
pam(8), pam(5) apparmor(7) and referred man pages rsyslogd(8), syslog(8), syslogd(8) fstab(5), mount(8), losetup(8), cryptsetup(8) haveged(8), random(4) ssh(1), sshd(8), ssh_config(5), sshd_config(5), ssh-agent(1), ssh-add(1), ssh-keygen(1) cron(1), crontab(5), at(1), atd(8) systemctl(1), daemon(7), systemd.unit(5), systemd.special(5), kernel-command-line(7), bootup(7), systemd.directives
2 Linux Security and Service Protection Methods #
In Chapter 1, Common Criteria we mentioned the Common Criteria EAL 4+ certified installation and setup that was sponsored by IBM for a select subset of hardware. This certified build is a great first step for customers wanting to build a secure and hardened base system, yet might not address all of the services and software specifics that many customers would be interested in.
This next part will present a more general view and give recommendations and guidance for SUSE Linux Enterprise Server system security. Some topics may seem repeated here (from the previous part) yet the context is very different. More detail will be provided in some sections and certainly some more general examples for a greater number of services.
This portion of the guide will only give basic recommendations instead of strict rules. The procedures and examples here should give you the ability to apply security enhancement techniques to a wider variety of server-based services and programs.
Some subjects of this chapter have been discussed before. However, you will find more details and explanations in this chapter. Selected general topics are:
Physical Security – Protection of the server from environmental threats (people, places, things).
Security Policies and Procedures – Server lifecycle management, disk/media reclamation, backup and archive security.
Systems Monitoring – Procedures around event notification/management.
Systems Automation – Mechanisms and/or procedures for automatic security measures. Heuristics, account control, security reporting and remediation, automated shutdown, etc.
Systems Management – Methods to obtaining packages, verification and signing keys, patching procedures and recommendations.
Securing Network – Addition programs, ports and service wrappers – iptables, tcpwrappers, services.
Remote Access – extra SSH information and key federation. CA integration.
Common Services – mail, NFS and automount.
Securing the kernel and init Process – parameters, systemd targets, and boot scripts.
Access Control – user/groups/permissions.
Password Security and Warnings – Proper setup of passwords, banners and
xinetd
.Miscellaneous Security – Assorted security settings and miscellany.
Resources – Web links, documentation and example references, howtos and general information, product links.
The sections will again be organized by a topical hierarchy for continuity-sake. Refer to the main table of contents for easy reference.
2.1 Physical Security #
Physical security should be one of the utmost concerns. Linux production servers should be in locked data centers where only people have access that have passed security checks. Depending on the environment and circumstances, you can also consider boot loader passwords.
Additionally, consider questions like:
Who has direct physical access to the host?
Of those that do, should they?
Can the host be protected from tampering and should it be?
The amount of physical security needed on a particular system depends on the situation, and can also vary widely by available funds.
2.1.1 System Locks #
Most server racks in data centers include a locking feature. Usually this will be a hasp/cylinder lock on the front of the rack that allows you to turn an included key to a locked or unlocked position – granting or denying entry. Cage locks can help prevent someone from tampering or stealing devices/media from the servers, or opening the cases and directly manipulating/sabotaging the hardware. Preventing system reboots or the booting from alternate devices is also important (for example CD/DVDs/USB drives/etc.).
Some servers also have case locks. These locks can do different things according to the designs of the system vendor and construction. Many systems are designed to self-disable if attempts are made to open the system without unlocking. Others have device covers that will not let you plug in or unplug keyboards or mice. While locks are sometimes a useful feature, they are usually lower quality and easily defeated by attackers with ill intent.
2.2 Locking Down the BIOS #
This section describes only basic methods to secure the boot process. To find out more advanced UEFI and the secure boot feature, see Book “Administration Guide”, Chapter 12 “UEFI (Unified Extensible Firmware Interface)”, Section 12.1 “Secure Boot”.
The BIOS (Basic Input/Output System) or its successor UEFI (Unified Extensible Firmware Interface) is the lowest level of software/firmware that dictates system configuration and low-level hardware. When this document references the BIOS, it usually means BIOS and/or UEFI. GRUB 2 and other Linux boot loaders access the BIOS to determine how to boot the host. Other hardware types (IBM POWER/IBM Z) that run Linux also have low-level software/firmware. Typically the BIOS can be configured to help prevent attackers from being able to reboot the host and manipulate the system.
Most BIOS varieties allow the setting of a boot password. While this does not provide a high level of security (a BIOS can be reset, removed or modified – assuming case access), it can be another deterrent.
Many BIOS capabilities have other various security settings – checking with the system vendor, the system documentation or examine the BIOS during a system boot.
If a system host has been set up with a boot password, the host will not boot up unattended (for example a system reboot, power failure, etc.). This is a trade-off.
2.3 Security via the Boot Loaders #
The Linux boot loader GRUB 2, which is used by default in SUSE Linux Enterprise Server, can have a boot passwords set. It also provides a password feature, so that only administrators can start the interactive operations (for example editing menu entries and entering the command line interface). If a password is specified, GRUB 2 will disallow any interactive control until you press the key C and E and enter a correct password.
You can refer to the GRUB 2 man page for examples.
It is very important to keep in mind that when setting these passwords they will need to be remembered! Also, enabling these passwords might merely slow an intrusion, not necessarily prevent it. Again, someone could boot from a removable device, and mount your root partition. If you are using BIOS-level security and a boot loader, it is a good practice to disable the ability to boot from removable devices in your computer's BIOS, and then also password-protecting the BIOS itself.
Also keep in mind that the boot loader configuration files will need to
be protected by changing their mode to 600
(read/write for root
only), or others will be able to read
your passwords or hashes!
2.4 Verifying Security Action with seccheck
#
It is highly recommended to have scripts in place which can verify that
security actions or procedures have been run. Even the best systems
administrators can make errors or forget something. If you have a small
or large Linux installation or environment, you should consider the use
of the seccheck
scripts.
seccheck
is the SUSE Security Checker. It is a
set of several shell scripts designed to check the local security of
the system on a regular basis. There are three main scripts that are
executed at different time intervals. They are
security-daily,
security-weekly
and security-monthly
. If seccheck
is not installed on your system, install it with sudo zypper
in seccheck
. These scripts all have schedule entries that get
placed in cron that determine when they run. Although cron scheduling
is the default behavior, this can be controlled via configuration
settings (see next section). The daily script runs at midnight, and if
changes are detected since the last run (the night before), an e-mail
noting the differences will be sent. The weekly script runs every
Monday at 1:00 am, and only if changes to the last run (the week
before) are found, a mail with the differences will be sent. The
monthly script runs every on every 1st of the month and sends the full
last daily and weekly report via e-mail.
2.4.1 Seccheck Configuration #
Note that you can change the receiver of the seccheck mails from root
to anyone else if you add an entry like this one to
/etc/sysconfig/seccheck
:
SECCHK_USER="firewall" # exchange firewall is an admin user's account name
Also note that the START_SECCHK
variable from
/etc/sysconfig/seccheck
controls whether the
security check will be run from cron. (It is ignored if you call
security-control
manually.)
The following daily checks are done:
|
length/number/contents of fields, accounts with same UID accounts with UID/GID of 0 or 1 beside root and bin |
|
length/number/contents of fields, accounts with no password |
|
length/number/contents of fields |
user root checks |
secure umask and |
|
checks if important system users are put there |
|
checks for mail aliases which execute programs |
|
checks if users' |
home directory |
checks if home directories are writable or owned by someone else |
dot-files check |
checks many dot-files in the home directories if they are writable or owned by someone else |
mailbox check |
checks if user mailboxes are owned by user and are readable |
NFS export check |
exports should not be exported globally |
NFS import check |
NFS mounts should have the |
promisc check |
checks if network cards are in promiscuous mode |
list modules |
lists loaded modules |
list sockets |
lists open ports |
Weekly Checks are as follows: | |
password check |
runs |
RPM md5 check |
checks for changed files via RPM's MD5 checksum feature |
suid/sgid check |
lists all suid and sgid files |
exec group write |
lists all executables which are group/world-writable |
writable check |
lists all files which are world-writable (including executables) |
device check |
lists all devices |
Additional monthly checks are also run, however the key difference is mainly that the monthly file is not a diff like the daily/weekly ones but the full reports in one file.
2.4.2 Automatic Logout #
The seccheck
package provides an automatic
logout feature. It is a script that runs every 10 minutes and checks
both remote or local terminal sessions for inactivity, and terminates
them if an idle time is exceeded.
You can configure the functionality in the
/etc/security/autologout.conf
file. Parameters
include default idle and logout delay times, and the configuration for
limiting maximum idle times specific to users, groups, TTY devices and
SSH sessions. /etc/security/autologout.conf
also
includes several configuration examples.
The automatic logout feature is not enabled by default. To enable it,
edit /etc/cron.d/autologout
and uncomment the
example cron line.
2.5 Retiring Linux Servers with Sensitive Data #
Security policies usually contain some procedures for the treatment of
storage media that is going to be retired or disposed of. Disk and
media wipe procedures are frequently prescribed as is complete
destruction of the media. You can find several free tools on the
Internet. A search of “dod disk wipe utility” will yield
several variants. To retire servers with sensitive data, it is
important to ensure that data cannot be recovered from the hard disks.
To ensure that all traces of data are removed, a wipe
utility—such as scrub
—can be used.
Many wipe utilities overwrite the data several times. This
assures that even sophisticated methods are not able to retrieve
any parts of the wiped data. Some tools can even be operated from a
bootable removable device and remove data according to the U.S.
Department of Defense (DoD) standards. Note that many government
agencies specify their own standards for data security. Some
standards are stronger than others, yet may require more time to
implement.
Some devices, like SSDs, use wear leveling and do not necessarily write new data in the same physical locations. Such devices usually provide their own erasing functionality.
2.5.1 scrub: Disk Overwrite Utility #
scrub
overwrites hard disks, files, and other
devices with repeating patterns intended to make recovering data from
these devices more difficult. It operates in three basic modes: on a
character or block device, on a file, or on a directory specified. For
more information, set the manual page man 1 scrub
.
- nnsa
4-pass NNSA Policy Letter NAP-14.1-C (XVI-8) for sanitizing removable and non-removable hard disks, which requires overwriting all locations with a pseudo random pattern twice and then with a known pattern: ran- dom(x2), 0x00, verify.
- dod
4-pass DoD 5220.22-M section 8-306 procedure (d) for sanitizing removable and non-removable rigid disks which requires overwriting all addressable locations with a character, its complement, a random character, then verify. Note: scrub performs the random pass first to make verification easier: random, 0x00, 0xff, verify.
- bsi
9-pass method recommended by the German Center of Security in Information Technologies (http://www.bsi.bund.de): 0xff, 0xfe, 0xfd, 0xfb, 0xf7, 0xef, 0xdf, 0xbf, 0x7f.
- gutmann
The canonical 35-pass sequence described in Gutmann's paper cited below.
- schneier
7-pass method described by Bruce Schneier in "Applied Cryptography" (1996): 0x00, 0xff, random(x5)
- pfitzner7
Roy Pfitzner's 7-random-pass method: random(x7).
- pfitzner33
Roy Pfitzner's 33-random-pass method: random(x33).
- usarmy
US Army AR380-19 method: 0x00, 0xff, random. (Note: identical to DoD 522.22-M section 8-306 procedure (e) for sanitizing magnetic core memory).
- fillzero
1-pass pattern: 0x00.
- fillff
1-pass pattern: 0xff.
- random
1-pass pattern: random(x1).
- random2
2-pass pattern: random(x2).
- old
6-pass pre-version 1.7 scrub method: 0x00, 0xff, 0xaa, 0x00, 0x55, verify.
- fastold
5-pass pattern: 0x00, 0xff, 0xaa, 0x55, verify.
- custom=string
1-pass custom pattern. String may contain C-style numerical escapes: \nnn (octal) or \xnn (hex).
2.6 Backups #
If your system is compromised, backups can be used to restore a prior system state. When bugs or accidents occur, backups can also be used to compare the current system against an older version. For production systems, it is very important to take some backups off-site for cases like disasters (for example off-site storage of tapes/recordable media, or off-site initiated).
For legal reasons, some firms and organizations must be careful about backing up too much information and holding it too long. If your environment has a policy regarding the destruction of old paper files, you might need to extend this policy to Linux backup tapes as well.
The rules about physical security of servers apply on backups as well.
2.7 Disk Partitions #
Servers should have separate file systems for at least
/
, /boot
,
/usr
, /var
,
/tmp
, and /home
. You do not
want that for example logging and temporary space under
/var
and /tmp
fill up the
root partition. Third-party applications should be on separate file
systems as well, for example under /opt
.
You are advised to review Chapter 1, Common Criteria. It is
important to understand the need to separate the partitions that could
impact a running system (for example, log files filling up
/var/log
are a good reason to separate
/var
from the /
partition).
Another thing to keep in mind is that you will likely need to leverage
LVM or another volume manager or at the very least the extended
partition type to work around the limitations of primary partitions (4
partitions).
Another capability in SUSE Linux Enterprise Server is encrypting a partition or even a single directory or file as a container. Refer to Book “Security and Hardening Guide”, Chapter 12 “Encrypting Partitions and Files” for details.
2.8 Firewall (iptables) #
iptables will not be covered in detail in this guide. Most companies use dedicated firewalls or appliances to protect their servers in a production network. This is strongly recommended for secure environments. SUSE Linux Enterprise Server also includes SuSEfirewall2 which is a wrapper for iptables and is enabled by default as a simple and layered protection.
If you are also interested in Linux stateful firewalls using iptables, there are many guides on the Internet. See the Appendix for resources. For lots of iptables tutorials and examples, see http://www.linuxguruz.com/iptables/.
2.9 Security Features in the Kernel #
The following list shows tunable kernel parameters you can use to
secure your Linux server against attacks. Some are defaults
already within the SLE distributions. To check the current status of
any of these settings, you can query the kernel
(/proc/sys/...
contents). For example, the Source
Routing setting is located in the
/proc/sys/net/ipv4/conf/all/accept_source_route
file. Simply display the contents of a file to see how the current
running kernel is set up.
For each tunable kernel parameter shown, the change to the entry that
needs to be affected can be modified or added to the
/etc/sysctl.conf
configuration file to make the
change persistent after a reboots.
You can get a list of current kernel settings by using the command:
root #
sysctl -a
It is even a very good idea to store the output of the kernel settings
(for comparison or reference) by redirecting the output of the sysctl
command to a file, for example
root #
sysctl -A > /root/sysctl.settings.store
Because SUSE Linux Enterprise Server includes, by default, security-focused kernel
tuning parameters, you will find the existing
/etc/sysctl.conf
file to be sparsely populated.
You may choose to use the above mentioned “catalog” method
of storing the complete gamut of kernel settings and then choose those
parameters you want to be reset at reboot. You can place these in the
/etc/sysctl.conf
file where they will be picked up
upon a reboot or they can be inserted immediately (into the running
kernel) by running the command sysctl
-p
.
Many third-party applications like Oracle, SAP, DB2, Websphere, etc. recommend changing kernel parameters to ensure high performance for I/O or CPU processing. Having a full list of current settings can be helpful for reference.
2.9.1 Enable TCP SYN Cookie Protection (default in SUSE Linux Enterprise Server 12 SP5) #
A SYN attack is a denial-of-service attack that consumes all the
resources on a machine. Any server that is connected to a network is
potentially subject to this attack. To enable TCP SYN Cookie
Protection, edit the /etc/sysctl.conf
file and
ensure the following line and value exists:
net.ipv4.tcp_syncookies = 1
Despite the name of the configuration option, it applies to IP version 6 as well.
2.9.2 Disable IP Source Routing (default in SUSE Linux Enterprise Server 12 SP5) #
Source Routing is used to specify a path or route through the network from source to destination. This feature can be used by network people for diagnosing problems. However, if an intruder was able to send a source routed packet into the network, then they could intercept the replies and your server might not know that it is not communicating with a trusted server.
net.ipv4.conf.all.accept_source_route = 0
or
net.ipv6.conf.all.accept_source_route = 0
2.9.3 Disable ICMP Redirect Acceptance #
ICMP redirects are used by routers to tell the server that there is a
better path to other networks than the one chosen by the server.
However, an intruder could potentially use ICMP redirect packets to
alter the host's routing table by causing traffic to use a path you
did not intend. To disable ICMP Redirect Acceptance, edit the
/etc/sysctl.conf
file and add the following line:
net.ipv4.conf.all.accept_redirects = 0
or
net.ipv6.conf.all.accept_redirects = 0
2.9.4 Enable IP Spoofing Protection (default in SUSE Linux Enterprise Server 12 SP5) #
IP spoofing is a technique where an intruder sends out packets which claim to be from another host by manipulating the source address. IP spoofing is very often used for denial-of-service attacks. For more information on IP Spoofing, see http://en.wikipedia.org/wiki/IP_address_spoofing
net.ipv4.conf.all.rp_filter = 1
2.9.5 Enable Ignoring to ICMP Requests #
If you want or need Linux to ignore ping
requests,
edit the /etc/sysctl.conf
file and add the
following line:
net.ipv4.icmp_echo_ignore_all = 1
This cannot be done in many environments, as some monitoring systems use a rudimentary ICMP (ping) to determine the health of the device on the network (or at least its ability to respond).
2.9.6 Enable Ignoring Broadcasts Request (default in SUSE Linux Enterprise Server 12 SP5) #
If you want or need Linux to ignore broadcast requests.
net.ipv4.icmp_echo_ignore_broadcasts = 1
2.9.7 Enable Bad Error Message Protection (default in SUSE Linux Enterprise Server 12 SP5) #
To alert you about bad error messages in the network.
net.ipv4.icmp_ignore_bogus_error_responses = 1
2.9.8 Enable Logging of Spoofed Packets, Source Routed Packets, Redirect Packets #
To turn on logging for Spoofed Packets, Source Routed Packets, and
Redirect Packets, edit the /etc/sysctl.conf
file
and add the following line:
net.ipv4.conf.all.log_martians = 1
Because of the way SUSE Linux Enterprise Server is set up (with rsyslog) for network event tracking, keep in mind that this can cause a large amount of messages to be logged.
2.9.9 Buffer Overflow Attack Mitigation #
Starting with the 2.6.x kernel releases, Linux offers Address Space Layout Randomization (ASLR) and the No-eXecute (NX bit) for mitigation of buffer overflow attacks. For more information, see:
Since version 12, SUSE Linux Enterprise Server already comes with some buffer overflow attack mitigation techniques being enabled by default.
ASLR is enabled by default. This can be verified with the output of the following command. The expected result is 2:
tux >
cat /proc/sys/kernel/randomize_va_space
2
This randomizes the heap, stack, and load addresses of dynamically linked libraries. Programs that run privileged or process network data are already built using special compiler flags (PIE and _FORTIFY_SOURCE) to take even more advantage of randomizing the text and data segments as well.
Executable space protection prevents the execution of memory space that is not intended for execution. Linux makes use of the No eXecute bit. This is enabled by default on the SUSE Linux Enterprise Server kernel for the x86 and AMD64/Intel 64 architecture. Use of the NX bit has to be supported by each individual program. You can check if your system supports the NX bit:
tux >
dmesg | grep '[NX|DX]*protection'
[ 0.000000] NX (Execute Disable) protection: active
If NX is disabled, check your BIOS or UEFI for a setting that enables it and make sure that it is supported by your CPU.
Furthermore, since version 12, SUSE Linux Enterprise Server prevents leaking of
internal kernel addresses to make kernel exploits harder by setting
the kptr_restrict
:
tux >
cat /proc/sys/kernel/kptr_restrict
1
On CPU's that support it (newer AMD64/Intel 64 CPUs) the kernel also uses the SMEP protection by default that prevents direct execution of user space code from inside the kernel. This is often used by kernel exploits and therefore a good hardening measure.
2.9.10 File system hardening #
To mitigate vulnerabilities based on insecure file system access by
privileged programs (tmp-races, TOCTOU) the Linux kernel offers two
sysctl
variables which should already be enabled by
default on SUSE Linux Enterprise Server 12 SP5:
fs.protected_hardlinks
and
fs.protected_symlinks
or their corresponding
/proc
entries:
tux >
cat /proc/sys/fs/protected_hardlinks
1
By setting this to 1
, users can no longer create soft
or hard links to files which they do not own. This mitigates a commonly
used exploitation vector for programs which call open(2), creat(2) or
similar functions without care.
tux >
cat /proc/sys/fs/protected_symlinks
1
By setting this to 1
, symbolic links are permitted to
be followed only when outside a sticky world-writable directory, or when
the uid of the link and follower match, or when the directory owner
matches the symbolic link's owner.
2.9.11 Increased dmesg
Restrictions #
dmesg
provides all kinds of system internal
information, such as kernel addresses, crashes of services, and
similar things that could be used by local attackers. This is why the
access to dmesg
is restricted to root
by
default. The behavior is controlled by the
kernel.dmesg_restrict
option (defaults to 1). If
set to 0, any user can view the output of dmesg
.
2.9.12 Filter access to /dev/mem
(default in SUSE Linux Enterprise Server 12) #
/dev/mem
hosts an image of the system's main
memory, including kernel and user space memory. Allowing unfiltered
access to this information is dangerous and therefore the kernel on
SUSE Linux Enterprise Server has been compiled with CONFIG_STRICT_DEVMEM
enabled. This setting restricts user space access to
/dev/mem
to memory mapped peripherals.
2.10 AppArmor #
Included with SUSE Linux Enterprise Server, AppArmor is an application security tool designed to provide an easy-to-use security framework for your applications. AppArmor proactively protects the operating system and applications from external or internal threats, even zero-day attacks, by enforcing a specified behavior and preventing some unknown application flaws from being exploited. AppArmor security policies, called “profiles”, completely define which system resources and files can be accessed by each application. The profiles also define the access mode, for example read or write. Several default profiles are included with AppArmor, and using a combination of advanced static analysis and learning-based tools, AppArmor profiles for even very complex applications can be deployed successfully in a matter of hours.
AppArmor consists of:
A kernel extension which enforces the security profiles.
A collection of RPMs, also shipped with SUSE Linux Enterprise Server that provides:
A set of AppArmor profiles for numerous programs that ship with SUSE Linux Enterprise Server.
Tools to create and manage new and existing AppArmor profiles.
A YaST user interface to manage reports and notification of security events.
Documentation about the AppArmor tools.
It is best to reboot a system after completing installation, so that AppArmor can confine all system daemons.
For additional details and step-by-step instructions on the usage and configuration of AppArmor you can also refer to Book “Security and Hardening Guide”.
2.11 SELinux #
SELinux is an advanced technology for securing Linux systems. It is included with “basic enablement” in SUSE Linux Enterprise Server 12 SP5, and is included with some other distributions by default. Hardening Linux using SELinux technology, on its own, warrants its own security HOWTO and is out of scope for this guide. The book SELinux: NSA's Open Source Security Enhanced Linux contains a very good description of its setup and usage. As part of the “basic enablement”, SELinux will not be officially supported, but packages have now been added to SUSE Linux Enterprise Server 12 SP5 to enable its usage with minimal effort. While AppArmor is much easier to use and has a similar feature set, knowing both will most certainly be beneficial.
2.12 FTP, telnet
, and rlogin
(rsh) #
The programs/protocols of FTP, telnet
, and
rlogin
(rsh
) are vulnerable to
eavesdropping, which is one of the main reasons secure alternatives
such as ssh
, scp
or
sftp
should be used instead. It is highly recommended
not to run these insecure services. Because of the high risk,
this guide does not cover these services (other than vsftp). It would
also be a good idea (and part of our guidance, see next section) not to
have FTP and Telnet server RPMs installed on the system. Note that the
EAL 4+ evaluation had vsftp installed. The “vs” stands for
“very secure”, which is a differentiator here when
compared to normal FTP.
2.13 Removing Unnecessary Software Packages (RPMs) #
A very important step in securing a Linux system is to determine the primary function(s) or role(s) of the Linux server. Otherwise, it can be difficult to understand what needs to be secured and securing these Linux systems can prove ineffective. Therefore, it is critical to look at the default list of software packages and remove any unnecessary packages or packages that do not comply with your defined security policies.
Generally, an RPM software package consists of the following:
The package's meta data that is written to the RPM database upon installation.
The package's files and directories.
Scripts that are being executed before and after installation and removal.
Packages generally do not impose any security risk to the system unless they contain:
setuid or setgid bits on any of the installed files
group- or world-writable files or directories
a service that is activated upon installation/activated by default.
Assuming that neither of the three conditions above apply, a package is merely a collection of files. Neither installation nor deinstallation of such packages has influence on the security value of the system.
Nevertheless, it is useful to restrict the installed packages in your system to a minimum. Doing this will result in fewer packages that require updates and will simplify maintenance efforts when security alerts and patches are released. It is a best practice not to install, among others, development packages or desktop software packages (for example, an X Server) on production servers. If you do not need them, you should also not install, for example, the Apache Web server or Samba file sharing server.
Many third-party vendors like Oracle and IBM require a desktop environment and development libraries to run installers. To avoid this from having an impact on the security of their production servers, many organizations work around this by creating a silent installation (response file) in a development lab.
Also, other packages like FTP and Telnet daemons should not be
installed as well unless there is a justified business reason for it
(again, ssh
, scp
or
sftp
should be used as replacements).
One of the first action items should be to create a Linux image that only contains RPMs needed by the system and applications, and those needed for maintenance and troubleshooting purposes. A good approach is to start with a minimum list of RPMs and then add packages as needed. This process is time-consuming but usually worth the effort.
The SUSE Appliance Program includes a component called JeOS (Just Enough Operating System). JeOS has a very small footprint and can be customized to fit the specific needs of a system developer. Main uses of JeOS are for hardware/software appliance or virtual machine development. Key benefits of JeOS are efficiency, higher performance, increased security and simplified management.
If JeOS is not an option for you, a good choice is the minimal installation pattern.
To generate a list of all installed RPMs, use the following command:
root #
rpm -qa
To retrieve details about a particular RPM (from the RPM itself), run:
root #
rpm -qi PACKAGE_NAME
To check for and report potential conflicts and dependencies when deleting an RPM, run:
root #
rpm -e --test PACKAGE_NAME
This can be very useful, as running the removal command without a test can often yield a mass of complaints and require manual recursive dependency hunting.
2.14 Patching Linux Systems #
Building an infrastructure for patch management is another very important part of a proactive and secure production Linux environment.
It is recommended to have a written security policy and procedure to handle Linux security updates and issues. For example, a security policy should detail the time frame for assessment, testing, and roll out of patches. Network related security vulnerabilities should get the highest priority and should be addressed immediately within a short time frame. The assessment phase should occur within a testing lab, and initial roll out should occur on development systems first
A separate security log file should contain details on which Linux security announcements have been received, which patches have been researched and assessed, when patches have been applied, etc.
SUSE releases their patches in three categories, security, recommended and optional. There are a few options that can be used to keep systems patched, up to date and secure. Each system can register and then retrieve updates via the SUSE Update Web site using the included YaST tool—YaST Online Update. SUSE has also created the Subscription Management Tool (SMT), an efficient way to maintain a local repository of available/released patches/updates/fixes that systems can then pull from (reducing Internet traffic). SUSE also offers SUSE Manager for the maintenance, patching, reporting and centralized management of Linux systems, not only SUSE, but other distributions as well.
2.14.1 YaST Online Update #
On a per-server basis, installation of important updates and improvements is possible using the YaST Online Update tool. Current updates for the SUSE Linux Enterprise family are available from the product specific update catalogs containing patches. Installation of updates and improvements is accomplished using YaST and selecting
in the Group. All new patches (except the optional ones) that are currently available for your system will already be marked for installation. Clicking will then automatically install these patches.2.14.2 Automatic Online Update #
YaST also offers the possibility to set up an automatic update. Select
› . Configure a Daily or a Weekly update. Some patches, such as kernel updates, require user interaction, which would cause the automatic update procedure to stop. Check for the update procedure to proceed automatically.In this case, run a manual Online Update from time to install patches that require interaction.
When rpm
or
zypper
.
2.14.3 Subscription Management Tool—SMT #
The Subscription Management Tool for SUSE Linux Enterprise goes one step further than the Online Update process by establishing a proxy system with repository and registration targets. This helps customers centrally manage software updates within the firewall on a per-system basis, while maintaining their corporate security policies and regulatory compliance.
The downloadable SMT (http://download.suse.com/) is integrated with SUSE Customer Center (https://scc.suse.com/) and provides a repository and registration target that is synchronized with it. This can be very helpful in tracking entitlements in large deployments. The SMT maintains all the capabilities of SUSE Customer Center, while allowing a more secure centralized deployment. It is included with every SUSE Linux Enterprise subscription and is therefore fully supported.
The SMT provides an alternative to the default configuration, which requires opening the firewall to outbound connections for each device to receive updates. That requirement often violates corporate security policies and can be seen as a threat to regulatory compliance by some organizations. Through its integration with SUSE Customer Center, the SMT ensures that each device can receive its appropriate updates without the need to open the firewall, and without any redundant bandwidth requirements.
The SMT also enables customers to locally track their SUSE Linux Enterprise devices (that is servers, desktops, or Point of Service terminals) throughout their enterprise. Now they can easily determine how many entitlements are in need of renewal at the end of a billing cycle without having to physically walk through the data center to manually update spreadsheets.
The SMT informs the SUSE Linux Enterprise devices of any available software updates. Each device then obtains the required software updates from the SMT. The introduction of the SMT improves the interaction among SUSE Linux Enterprise devices within the network and simplifies how they receive their system updates. The SMT enables an infrastructure for several hundred SUSE Linux Enterprise devices per instance of each installation (depending on the specific usage profile). This offers more accurate and efficient server tracking.
In a nutshell, the Subscription Management Tool for SUSE Linux Enterprise provides customers with:
Assurance of firewall and regulatory compliance
Reduced bandwidth usage during software updates
Full support under active subscription from SUSE
Maintenance of existing customer interface with SUSE Customer Center
Accurate server entitlement tracking and effective measurement of subscription usage
Automated process to easily tally entitlement totals (no more spreadsheets!)
Simple installation process that automatically synchronizes server entitlement with SUSE Customer Center
2.14.4 SUSE Manager #
SUSE Manager automates Linux server management, allowing you to provision and maintain your servers faster and more accurately. It monitors the health of each Linux server from a single console so you can identify server performance issues before they impact your business. And it lets you comprehensively manage your Linux servers across physical, virtual and cloud environments while improving data center efficiency. SUSE Manager delivers complete lifecycle cycle management for Linux:
Asset management
Provisioning
Package management
Patch management
Configuration management
Redeployment
For more information on SUSE Manager refer to https://www.suse.com/products/suse-manager/.
2.15 Securing the Network—Open Network Ports Detection #
Securing a server requires that you know what it is serving; what
services are running. Default server installations may have services
running that aren't self apparent and open network ports that they are
using. So, one of the most important tasks is to detect and close
network ports that are not needed. To get a list of listening network
ports (TCP and UDP sockets), you can use the netstat
service run the following command:
root #
netstat -tulp
Be aware that netstat
output can be wider than a
default terminal screen. If the screen is too narrow, the options
described above will likely cause the output to wrap and be less
legible.
Below is an example of output from the above command:
Proto Recv-Q Send-Q Local Foreign State PID/Program Address Address name tcp 0 0 *:auth *.* LISTEN 2328/xinetd tcp 0 0 local[...].:smtp *.* LISTEN 2360/sendmail:acce tcp 0 0 *:ssh *.* LISTEN 2317/sshd
From the output above you can see that three TCP-based services are running and listening: xinetd, sendmail, and sshd. Sendmail should not be configured to listen for incoming network connections unless the server running it is a designated as a mail or relay server. Running a port scan from another server can confirm that, but make sure to obtain proper permissions to scan/probe a machine on a production network.
Some organizations consider port scans without permission a security offense.
Using the Nmap tool, such a probe/scan can be run:
root #
nmap -sTU REMOTE_HOST
Starting nmap 3.70 ( http://www.insecure.org/nmap/ ) at 2004-12-10 22:51 CST
Interesting ports on venus (192.168.2.101):
(The 3131 ports scanned but not shown below are in state: closed)
PORT STATE SERVICE
22/tcp open ssh
113/tcp open auth
Nmap run completed -- 1 IP address (1 host up) scanned in 221.669 seconds
Note that running the nmap
command can take quite a
while (in this example almost 4 minutes) depending on the options used.
If you remove the UDP port scan (leave out the -U
option), then nmap
will finish the port scan nearly
immediately. The results of nmap
can vary widely and
might not show all listening network sockets depending on the status of
the SuSEfirewall2 (or other) and if it has been set up to block any
ports.
From the sample run above, you see that the
xinetd
daemon is listening on
port auth (port 113) for IDENT
(for more information
on this service, see
Section 2.16, “xinetd
Services - Disabling”). You
can also see that sendmail is not listening for remote incoming network
connections.
Another method to list all of the TCP and UDP sockets to which programs
are listening (on a host) is to use the lsof
command –
which lists open files:
root #
lsof -i -n | egrep 'COMMAND|LISTEN|UDP'
COMMAND PID USER FD TYPE DEVICE SIZE/OFF NODE NAME
sshd 2317 root 3u IPv6 6579 0t0 TCP *:ssh (LISTEN)
xinetd 2328 root 5u IPv4 6698 0t0 TCP *:auth (LISTEN)
sendmail 2360 root 3u IPv4 6729 0t0 TCP 127.0.0.1:smtp (LISTEN)
2.16 xinetd
Services - Disabling #
The xinetd
daemon is a
replacement for inetd, the Internet services daemon. It monitors the
ports for all network services configured in
/etc/xinetd.d
, and starts the services in response
to incoming connections. To check if
xinetd
is enabled and running,
execute:
root #
systemctl status xinetd
To check the current status of the
xinetd
service, execute:
root #
systemctl status xinetd
If xinetd
is active, it is very
important to see which services are active and being controlled by
xinetd
. The following command
will list all services configured in /etc/xinetd.d
and whether xinetd
monitors the
ports for these services:
root #
chkconfig -list | awk '/xinetd based services/,/""/'
xinetd based services:
chargen: off
chargen-udp: off
cups-lpd: off
cvs: off
daytime: off
daytime-udp: off
discard: off
discard-udp: off
echo: off
echo-udp: off
netstat: off
rsync: off
sane-port: off
servers: off
services: off
svnserve: off
swat: off
systat: off
tftp: on
time: off
time-udp: off
vnc: off
To get a list of only active services for which xinetd monitors the
ports, you could run (where the -v
option of grep does
an inverse-match) :
root #
chkconfig --list | awk '/xinetd based services/,/""/' | grep -v off
xinetd based services:
tftp: on
In the above example you can see that the
telnet-server
package is not
installed on the system. If the Telnet Server package
telnet-server
would be
installed, it would show up on the list whether it is active. Here is
an example how to disable a service. Assuming the
tftp
service is active, run
the following commands to disable it and to see how the telnet service
entries are being updated:
tux >
sudo systemctl status tftp
tftp.service - Tftp Server
Loaded: loaded (/usr/lib/systemd/system/tftp.service; static)
Active: active (running) since Fri 2014-09-05 07:56:23 CEST; 2h 1min ago
# sudo cat /etc/xinetd.d/tftp | grep disable
disable = no
# sudo systemctl disable tftp
# systemctl status tftp
tftp.service - Tftp Server
Loaded: loaded (/usr/lib/systemd/system/tftp.service; static)
Active: inactive (dead)
# sudo cat /etc/xinetd.d/tftp | grep disable
disable = yes
For the TFTP service it would be better to remove the package from the system since removal is always safer than disabling (when possible):
root #
rpm -e tftp
2.16.1 Inventory xinetd
services #
It is important to investigate all active
xinetd
services and to disable
them (or remove their packages) if they are not needed. To find out
what a service does, here is a viable approach. Using the
tftp
service as an example
and assuming its function is unknown and it is listed as an active
service. Execute the following commands:
root #
grep " server" /etc/xinetd.d/tftp
server = /usr/sbin/in.tftpd
server_args = -s /tftpboot
To read the manual:
root #
man in.tftpd
TFTPD(8) System Manager's Manual TFTPD(8)
NAME
tftpd - IPv4 Trivial File Transfer Protocol server
SYNOPSIS
in.tftpd [options...] directory...
DESCRIPTION
tftpd is a server for the Trivial File Transfer Protocol.
The TFTP protocol is extensively used to support remote boot-
ing of diskless devices. The server is normally started by
inetd, but can also run stand-alone.
[...]
To determine what package supplies the in.tftpd binary:
root #
rpm -qf /usr/sbin/in.tftpd
tftp-0.48-101.16
To get a description of TFTP and its usage, etc:
root #
rpm -qi tftp-0.48-101.16| awk '/Description/,/""/'
Description :
The Trivial File Transfer Protocol (TFTP) is normally used only for
booting diskless workstations and for getting or saving network
component configuration files.
To get a list of what files are installed via the TFTP package (RPM),
execute the rpm
command with the following options:
root #
rpm -ql tftp-0.48-101.16
/etc/xinetd.d/tftp
/usr/bin/tftp
/usr/sbin/in.tftpd
/usr/share/doc/packages/tftp
/usr/share/doc/packages/tftp/README
/usr/share/doc/packages/tftp/README.security
/usr/share/doc/packages/tftp/sample.rules
/usr/share/man/man1/tftp.1.gz
/usr/share/man/man8/in.tftpd.8.gz
/usr/share/man/man8/tftpd.8.gz
This example described what could be done to find out information
about services (specifically ones started by
xinetd
) even if an online
manual did not exist for the binary in.ftpd
. This
example yielded a man page – but you may not always find one. The
RPM commands in the example are very commonly used for a variety of
reasons. It is also possible to use the YaST software management
interface to retrieve all of the resultant information – however
having a knowledge of RPM command syntax can save quite a bit of time.
Again to disable the TFTP service, execute the following command:
root #
systemctl disable tftp
The xinetd
daemon is quite
flexible and has many features. Here are a few functionalities of
xinetd
:
Access control for TCP, UDP, and RPC services.
Access limitations based on time.
Provides mechanisms to prevent DoS attacks.
For more specific information on Xinetd, review the documentation and usage examples at the xinetd GitHub page https://github.com/xinetd-org/xinetd.
2.17 Securing Postfix #
Postfix is a replacement for Sendmail and has several security advantages over Sendmail. Postfix is the default mail system in SUSE Linux Enterprise Server and consists of several small programs that each perform their own small task—most of these programs run in a chroot jail. This is one of the reasons Postfix is recommended over Sendmail.
Linux servers that are not dedicated mail or relay servers should not
accept external e-mails. However, it is important for production
servers to send local e-mails to a relay server—some security
setups (for example the seccheck scripts) can be configured to send
e-mails to someone other than root
, even off the local system.
Verify the following parameters in
/etc/postfix/main.cf
are set to ensure that
Postfix accepts only local e-mails for delivery (look toward the
bottom of the file as the top portion is mostly commented-out example
entries and explanations):
mydestination = $myhostname, localhost.$mydomain, localhost inet_interfaces = localhost
The mydestination
parameter lists all domains to
receive e-mails for. The inet_interfaces
parameter
specifies the network to listen on. After reconfiguring Postfix, a
restart of the mail system is necessary:
root #
systemctl restart postfix
Verify that Postfix is not listening for network requests (incoming) by running one of these commands from another host:
nmap -sT -p 25 REMOTE_HOST telnet <remote_host> 25
Running these commands on the local host will provide inaccurate results because Postfix is supposed to accept connections from the local node. Use an external host for correct results.
2.18 File Systems: Securing NFS #
NFS (Network File System) allows servers to share files over a network. But like all network services using NFS involves risks.
Here are some basic rules:
NFS should not be enabled if not needed.
If NFS is truly needed, use a TCP wrapper to restrict remote access.
Ensure to export only to those hosts that really need access.
Use a fully qualified domain name to diminish any spoofing attempts.
Export only as read-only whenever possible.
Use NFS only over TCP.
If you do not have shared directories to export, then ensure that the NFS service is not enabled nor running on the system:
Check the NFS service status:
root #
systemctl status nfsserver
Check the current targets:
root #
ls /etc/systemd/system/*.wants/nfsserver.service
2.18.1 Enabling and Starting NFS Server #
If NFS must be used, it can be activated using the following commands
on SUSE Linux Enterprise Server or more simply and securely with the YaST
plug-in (ncurses). Access it directly from command line with
yast nfs_server
or yast
nfs_client
– or manually:
root #
systemctl enable nfs
systemctl start nfs
Portmapper information:
root #
rpcinfo -p SERVERNAME
program vers proto port
100000 2 tcp 111 portmapper
100000 2 udp 111 portmapper
100003 2 udp 2049 nfs
100003 3 udp 2049 nfs
100003 2 tcp 2049 nfs
100003 3 tcp 2049 nfs
100005 1 udp 623 mountd
100005 1 tcp 626 mountd
100005 2 udp 623 mountd
100005 2 tcp 626 mountd
100005 3 udp 623 mountd
100005 3 tcp 626 mountd
If you run it from an "untrusted" server or network, you should get the following output:
root #
rpcinfo -p SERVERNAME
No remote programs registered.
2.18.2 Exporting NFS #
To allow a client access to a file system or directory, the
/etc/exports
file serves as the access control
list. To give the network "network.example.com" read-only
access to /pub
, the entries in
/etc/exports
would look like as follows:
/pub *.network.example.com(ro,sync)
It is very important not to give write access to
NFS clients if not absolutely needed! Entries in
/etc/exports
are exported read-only
(ro
option) by default. To allow servers
sles-ha1
,
sles-ha2
and
sles-ha3
read-write access
to the /data/MYSQL
directory, the entries in
/etc/exports
would look like as follows:
/data/MYSQL sles-ha1.example.com(rw,sync) sles-ha2.example.com(rw,sync) sles-ha3.example.com(rw,sync)
Note that the options must not be separated from
the host names or networks with whitespace(s). Also, fully qualified
domain names should always be used to diminish
spoofing attempts. All entries in /etc/exports
are exported with the root_squash
option (“root
squashing”) by default. This means that a root user on a client
machine does not have root privileges (root access) to files on
exported NFS. It is not recommended to turn “root
squashing” off using the no_root_squash
option! After you have made all your entries in
/etc/exports
, you can export all file
systems/directories using the following command:
root #
exportfs -a
To unexport all shared file systems/directories, run:
root #
exportfs -ua
To see all shared file systems/directories, run:
root #
showmount -e localhost
Export list for localhost:
/pub *.network.example.com/data/MYSQL
sles-ha1.example.com,sles-ha2.example.com,sles-ha3.example.com
2.18.3 Using NFS over TCP #
If you need NFS, it is recommended to use NFS only over TCP since NFS
over UDP is not secure. All 2.4 and later kernels support NFS over TCP
on the client side. Server support for TCP appeared in later 2.4
kernels and beyond. To mount a shared directory using NFS over TCP, it
is necessary to use the proto=tcp
mount option:
root #
mount -o proto=tcp SERVERNAME:/pub /usr/local/pub
Verify that the target directory, in this case
/usr/local/pub
, exists on the client:
root #
mount [...] SERVERNAME:/pub on
/usr/local/pub type nfs (rw,proto=tcp,addr=192.168.1.110)
To have the shared directory mounted on the client at boot time, use
the /etc/fstab
file. For the above example, the
/etc/fstab
entry could look like this:
SERVERNAME:/pub /usr/local/pub nfs rsize=8192,wsize=8192,timeo=14,intr,tcp 0 0
2.19 Copying Files Using SSH Without Providing Login Prompts #
This example is needed in some cases to enable files to be copied over the network using SSH without providing an interactive login prompt. This allows trusted hosts to be set up—an example of federation.
SSH can allow a forced command using the “command” option.
Using this option it is possible to disable scp (secure copy) and
enforce every passed ssh
command to be ignored. On
the server side where you want to retrieve the file from, add the
following entry to the beginning of the SSH key in the
~/.ssh/authorized_keys
file (the
~
represents a particular users home directory –
root
's home directory is /root
– other
users typically reside in
/home/USERNAME
):
command="/bin/cat ~/<file_name>" ssh-rsa AAAAB3N...{etc}
To copy now the file from the remote server, run the following command:
ssh USER@SERVERNAME LOCAL_FILE
Since /bin/cat
is run on the server side, its output
needs to be redirected to the local file.
Another approach is to replace the /bin/cat
(referenced above) with your own script that checks the passed SSH
commands by reading the environment variable
$SSH_ORIGINAL_COMMAND
. For example:
#!/bin/ksh if [[ $SSH_ORIGINAL_COMMAND = "File1" || $SSH_ORIGINAL_COMMAND = "File2" ]] then /bin/cat $SSH_ORIGINAL_COMMAND else echo "Invalid file name!" exit 1 fi
So you replace the /bin/cat
portion with the script
name in ~/.ssh/authorized_keys
, and run the
following command to copy Foo1
:
tux >
ssh USER@SERVERNAME Foo1 > LOCAL_FILE
To copy Foo 2
, run:
tux >
ssh USER@SERVERNAME Foo2 > LOCAL_FILE
With the modifications above, every other variety of passed parameters will return errors.
2.20 Checking File Permissions and Ownership #
The following sections deal with some ways the default permissions and
file settings can be modified to enhance the security of a host. It is
important to note that the use of the default SUSE Linux Enterprise Server
utilities like seccheck
- can be run to lock down
and improve the general file security and user environment. However, it
is beneficial to understand how to modify these things.
SUSE Linux Enterprise Server hosts include three default settings for file
permissions: permissions.easy
,
permissions.secure
, and
permissions.paranoid
, all located in the
/etc
directory. The purpose of these files is to
define special permissions, such as world-writable directories or, for
files, the setuser ID bit (programs with the setuser ID bit set do not
run with the permissions of the user that has launched it, but with the
permissions of the file owner, usually root
).
Administrators can use the file
/etc/permissions.local
to add their own settings.
The easiest way to implement one of the default permission rule-sets
above is to use the module in
YaST.
Each of the following topics will be modified by a selected rule-set, but the information is important to understand on its own.
2.21 Default umask #
The umask
(user file-creation mode mask) command is
a shell built-in command which determines the default file permissions
for newly created files. This can be overwritten by system calls but
many programs and utilities use umask
. By default,
SUSE sets umask
to 022
. You
can modify this by changing the value in
/etc/profile
.
The id
command will print out the current user
identity information. Example from a non-root prompt:
tux >
id
uid=1000(ne0) gid=100(users) groups=16(dialout),33(video),100(users)
And to determine the active umask, use the umask
command:
tux >
umask
0022
Now, for comparison's sake, the same commands from the root user:
root #
id uid=0(root) gid=0(root) groups=0(root)root #
umask 0022
2.22 SUID/SGID Files #
When the SUID (set user ID) or SGID (set group ID) bits are set on an executable, it executes with the UID or GID of the owner of the executable rather than that of the person executing it. This means that, for example, all executables that have the SUID bit set and are owned by root are executed with the UID of root. A good example is the passwd command that allows ordinary users to update the password field in the /etc/shadow file which is owned by root.
But SUID/SGID bits can be misused when the SUID/SGID executable has a security hole. Therefore, you should search the entire system for SUID/SGID executables and document it. For example, ensure that code developers do not set SUID/SGID bits on their programs if it is not an absolute requirement. Very often you can use workarounds like removing the executable bit for world/others. However, a better approach is to change the design of the software if possible.
To search the entire system for SUID or SGID files, you can run the following command:
root #
find / -path /proc -prune -o -type f -perm '/6000' -ls
The -prune
option in this example is used to skip the
/proc file system.
2.23 World-Writable Files #
World-writable files are a security risk since it allows anyone to modify them. Additionally, world-writable directories allow anyone to add or delete files. To locate world-writable files and directories, you can use the following command:
root #
find / -path /proc -prune -o -perm -2 ! -type l -ls
The ! -type l
parameter skips all symbolic links since
symbolic links are always world-writable. However, this is not a
problem as long as the target of the link is not world-writable, which
is checked by the above find command.
World-writable directories with sticky bit—directory or file
ownership where only the item's owner, the directory's owner, or
root
can rename or delete files—such as the
/tmp
directory do not allow anyone except the
owner of a file to delete or modify it in this directory. The sticky
bit makes files stick to the user who created it and it prevents other
users from deleting and renaming the files. Therefore, depending on the
purpose of the directory world-writable directories with sticky are
usually not an issue. An example is the /tmp
directory:
tux >
ls -ld /tmp
drwxrwxrwt 18 root root 16384 Dec 23 22:20 /tmp
The t
mode, which denotes the sticky bit, allows
files to be deleted and renamed only if the user is the owner of this
file or the owner of the directory.
SUSE Linux Enterprise Server supports file capabilities to allow more fine grained privileges to be given to programs rather than the full power of root:
root #
getcap -v /usr/bin/ping
/usr/bin/ping = cap_new_raw+eip
The previous command only grants the CAP_NET_RAW capability to whoever
executes ping. In case of vulnerabilities inside ping, an attacker can
gain at most this capability in contrast to full root when granting the
ping binary SUID root rights. Whenever possible, file capabilities
should be chosen in favor of the suid-to-root bit. But this only
applies when the binary is suid to root, not to other users such as
news
, lp
and alike.
2.24 Orphaned or Unowned Files #
Files not owned by any user or group might not necessarily be a security problem in itself. However, unowned files could pose a security problem in the future. For example, if a new user is created and the new users happens to get the same UID as the unowned files have, then this new user will automatically become the owner of these files.
To locate files not owned by any user or group, use the following command:
root #
find / -path /proc -prune -o -nouser -o -nogroup
2.25 Restricting Access to Removable Media #
In some environments it is required to restrict access to removable
media such as USB storage or optical devices. The tools coming with the
udisks2
package help with such a
configuration.
Create a user group whose users will be allowed to mount and eject removable devices, for example mmedia_all:
tux >
sudo
groupadd mmedia_allAdd a specific user
tux
to the new group:tux >
sudo
usermod -a -G mmedia_alltux
Create the
/etc/polkit-1/rules.d/10-mount.rules
file with the following content:tux >
cat /etc/polkit-1/rules.d/10-mount.rules polkit.addRule(function(action, subject) { if (action.id =="org.freedesktop.udisks2.eject-media" && subject.isInGroup("mmedia_all")) { return polkit.Result.YES; } }); polkit.addRule(function(action, subject) { if (action.id =="org.freedesktop.udisks2.filesystem-mount" && subject.isInGroup("mmedia_all")) { return polkit.Result.YES; } });Important: Naming of the Rules FileThe name of a rules file must start with a digit, otherwise it will be ignored.
Rules files are processed in alphabetical order. Functions are called in the order they were added until one of the functions returns a value. Therefore, to add an authorization rule that is processed before other rules, put it in a file in /etc/polkit-1/rules.d with a name that sorts before other rules files, for example
/etc/polkit-1/rules.d/10-mount.rules
. Each function should return a value frompolkit.Result
.Restart
udisks2
:root #
systemctl restart udisks2Restart
polkit
root #
systemctl restart polkit
2.26 Various Account Checks #
2.26.1 Unlocked Accounts #
It is important that all system and vendor accounts that are not used
for logins are locked. To get a list of unlocked accounts on your
system, you can check for accounts that do not
have an encrypted password string starting with !
or *
in the /etc/shadow
file.
If you lock an account using passwd
-l
, it will put a !!
in front of
the encrypted password, effectively disabling the password. If you
lock an account using usermod
-L
,
it will put a !
in front of the encrypted password.
Many system and shared accounts are usually locked by default by
having a *
or !!
in the password
field which renders the encrypted password into an invalid string.
Hence, to get a list of all unlocked (encryptable) accounts, run
(egrep
is used to allow use of
regular-expressions):
root #
egrep -v ':\*|:\!' /etc/shadow | awk -F: '{print $1}'
Also make sure all accounts have a x
in the
password field in /etc/passwd
. The following
command lists all accounts that do not have a x
in
the password field:
root #
grep -v ':x:' /etc/passwd
An x
in the password fields means that the password
has been shadowed, for example the encrypted password needs to be
looked up in the /etc/shadow
file. If the
password field in /etc/passwd
is empty, then the
system will not look up the shadow file and it will not prompt the user
for a password at the login prompt.
2.26.2 Unused Accounts #
All system or vendor accounts that are not being used by users, applications, by the system or by daemons should be removed from the system. You can use the following command to find out if there are any files owned by a specific account:
root #
find / -path /proc -prune -o -user ACCOUNT -ls
The -prune
option in this example is used to skip the
/proc file system. If you are sure that an account can be deleted, you
can remove the account using the following command:
root #
userdel -r ACCOUNT
Without the -r
option userdel
will not delete the
user's home directory and mail spool
(/var/spool/mail/USER
).
Note that many system accounts have no home directory.
2.27 Enabling Password Aging #
Password expiration is a general best practice—but might need to be excluded for some system and shared accounts (for example Oracle, etc.). Expiring password on those accounts could lead to system outages if the application account expires.
Typically a corporate policy should be developed that dictates rules/procedures regarding password changes for system and shared accounts. However, normal user account passwords should expire automatically. The following example shows how password expiration can be set up for individual user accounts.
The following files and parameters in the table can be used when a new
account is created with the useradd
command.
Settings such as these are stored for each user account in the
/etc/shadow
file. If using the YaST tool
( ) to add users, the
settings are available on a per-user basis. Here are the various
settings—some of which can also be system-wide (for example
modification of /etc/login.defs
and
/etc/default/useradd
):
|
|
Maximum number of days a password is valid. |
|
|
Minimum number of days before a user can change the password since the last change. |
|
|
Number of days when the password change reminder starts. |
|
|
Number of days after password expiration that account is disabled. |
|
|
Account expiration date in the format YYYY-MM-DD. |
Users created prior to these modifications will not be affected.
Ensure that the above parameters are changed in the
/etc/login.defs
and
/etc/default/useradd
files. Review of the
/etc/shadow
file will show how these settings get
stored after adding a user.
To create a new user account, execute the following command:
root #
useradd -c "TEST_USER" -g USERS TEST
The -g
option specifies the primary group for this
account:
root #
id TEST
uid=509(test) gid=100(users) groups=100(users)
The settings in /etc/login.defs
and
/etc/default/useradd
are recorded for the test
user in the /etc/shadow
file as follows:
root #
grep TEST /etc/shadow
test:!!:12742:7:60:7:14::
Password aging can be modified at any time by use of the
chage
command. To disable password aging for system
and shared accounts, you can run the following chage
command:
root #
chage -M -1 SYSTEM_ACCOUNT_NAME
To get password expiration information:
root #
chage -l SYSTEM_ACCOUNT_NAME
For example:
root #
chage -l TEST
Minimum: 7
Maximum: 60
Warning: 7
Inactive: 14
Last Change: Jan 11, 2015
Password Expires: Mar 12, 2015
Password Inactive: Mar 26, 2015
Account Expires: Never
2.28 Stronger Password Enforcement #
On an audited system it is important to restrict people from using simple passwords that can be cracked too easily. Writing down complex passwords is all right as long as they are stored securely. Some will argue that strong passwords protect you against dictionary attacks and those type of attacks can be defeated by locking accounts after a few failed attempts. However, this is not always an option. If set up like this, locking system accounts could bring down your applications and systems which would be nothing short of a denial-of-service attack - another issue.
At any rate, it is important to practice effective password management safety. Most companies require that passwords have at the very least a number, one lowercase letter, and one uppercase letter. Policies vary, but maintaining a balance between password strength/complexity and management is sometimes difficult.
2.29 Leveraging an Effective PAM stack #
Linux-PAM (Pluggable Authentication Modules for Linux) is a suite of shared libraries that enable the local system administrator to choose how applications authenticate users.
It is strongly recommended to familiarize oneself with the capabilities of PAM and how this architecture can be leveraged to provide the best authentication setup for an environment. This configuration can be done once – and implemented across all systems (a standard) or can be enhanced for individual hosts (enhanced security – by host / service / application). The key is to realize how flexible the architecture is.
To learn more about the PAM architecture, find PAM documentation in
the /usr/share/doc/packages/pam
directory (in
a variety of formats).
The following discussions are examples of how to modify the default PAM stacks—specifically around password policies—for example password strength, password re-use and account locking. While these are only a few of the possibilities, they serve as a good start and demonstrate PAM's flexibility.
2.29.1 Password Strength #
SUSE Linux Enterprise Server can leverage the
pam_cracklib
library to test
for weak passwords – and to suggest using a stronger one if it
determines obvious weakness. The following parameters represent an
example that could be part of a corporate password policy or something
required because of audit constraints.
The PAM libraries follow a defined flow. The best way to design the perfect stack usually is to consider all of the requirements and policies and draw out a flow chart.
|
|
Minimum length of password is 8 |
|
|
Minimum number of lowercase letters is 1 |
|
|
Minimum number of uppercase letters is 1 |
|
|
Minimum number of digits is 1 |
|
|
Minimum number of other characters is 1 |
To set up these password restrictions, use the
pam-config
tool and specify the parameters you want
to configure. For example, the minimum length parameter could be
modified like this:
sudo pam-config -a --cracklib-minlen=8 --cracklib-retry=3 \ --cracklib-lcredit=-1 --cracklib-ucredit=-1 --cracklib-dcredit=-1 \ --cracklib-ocredit=-1 --cracklib
Now verify that the new password restrictions work for new passwords.
Simply login to a non-root account and change the password using the
passwd
command. Note that the above requirements
are not enforced if you run the passwd
command
under root.
2.29.2 Restricting Use of Previous Passwords #
The pam_pwhistory module can be used to configure the number of previous passwords that cannot be reused. The following command implements password restrictions on a system so that a password cannot be reused for at least 6 months:
sudo pam-config -a --pwhistory --pwhistory-remember=26
Recall that in the section
Section 2.27, “Enabling Password Aging” we set
PASS_MIN_DAYS
to 7
, which
specifies the minimum number of days allowed between password changes.
Therefore, if pam_unix
is configured to
remember 26
passwords, then the previously used
passwords cannot be reused for at least 6 months (26*7 days).
The PAM configuration (/etc/pam.d/common-auth
)
resulting from the pam-config
command looks like the
following:
auth required pam_env.so auth required pam_unix.so try_first_pass account required pam_unix.so try_first_pass password requisit pam_cracklib.so password required pam_pwhistory.so remember=26 password optional pam_gnome_keyring.so use_authtok password required pam_unix.so use_authtok nullok shadow try_first_pass session required pam_limits.so session required pam_unix.so try_first_pass session optional pam_umask.so
Editing the PAM configuration files directly is not recommended. Use
the pam-config
command as shown above to fine-tune
PAM. For more information see man 8 pam-config
.
2.29.3 Locking User Accounts After Too Many Login Failures #
It is not generally recommend that a host automatically locks system and shared accounts after too many failed login or su attempts. This could lead to outages if the application's account gets locked because of too many login failures like in this example for an Oracle shared account:
root #
su oracle -c id
su: incorrect password
This could be an easy target for a denial-of-service attack. The
following example shows how to lock only individual user accounts
after too many failed su or login attempts. Add the following line
to the service files (such as /etc/pam.d/su
or /etc/pam.d/login
) that you want to
configure (but not to
/etc/pam.d/common-auth
!):
auth required pam_tally2.so deny=5 unlock_time=1200 [...]
The line counts failed login and failed su attempts for
all user accounts except the root account. The accounts will
be automatically unlocked after 1200 seconds.
The default location for attempted accesses is recorded in
/var/log/tallylog
.
If the user is authenticated and the login process continues on call to pam_setcred(3) it resets the attempts counter to 0.
SSH configuration requires an additional line in
/etc/pam.d/sshd
, so that the failed
login counter will reset after a successful login. The line is
account required pam_tally2.so
, as shown
in the following example:
#%PAM-1.0
auth requisite pam_nologin.so
auth include common-auth
auth required pam_tally2.so deny=5 unlock_time=1200
account requisite pam_nologin.so
account include common-account
account required pam_tally2.so
password include common-password
session required pam_loginuid.so
session include common-session
session optional pam_lastlog.so silent noupdate showfailed
If this line is omitted, pam_tally2
keeps the previous
failure count.
It is also possible to add the
lock_time=N
parameter,
and then optionally the
unlock_time=N
parameter.
For example, setting the lock_time=60
would
deny access for 60 seconds after a failed attempt. The
unlock_time=N
option
would then allow access after
N
seconds after an
account has been locked. If this option is used the user will be
locked out for the specified amount of time after he exceeded his
maximum allowed attempts. Otherwise the account is locked until
the lock is removed by a manual intervention of the system
administrator. See the pam_tally2
man page for
more information.
To exempt system and shared accounts from the
deny=N
parameter,
the per_user
parameter was added to the
module. The per_user
parameter instructs the module
not to use the
deny=N
limit for
accounts where the maximum number of login failures is set
explicitly. For example:
root #
pam_tally2 -u oracle
Login Failures Latest failure From
oracle 0 Fri Dec 10 23:57:55 -0600 2005 on unknown
To instruct the module to activate the
deny=N
limit for this
account again, run:
pam_tally2 -u oracle deny=n
By default, the maximum number of login failures for each account is
set to zero (0) which instructs pam_tally2 to leverage the
deny=N
parameter. To see
failed login attempts, run:
pam_tally2
Test these changes thoroughly on your system using
ssh
and su
, and make sure
the root
id does not get locked! To lock or unlock accounts
manually, run one of the following commands:
- Locking
passwd -l USER usermod -L USER
- Unlocking
passwd -u USER usermod -U USER
2.30 Restricting root
Logins #
By default, the root
user is assigned a password
and can log in using various methods—for example, on a local
terminal, in a graphical session, or remotely via SSH. These methods
should be restricted as far as possible. Shared usage of the root account
should be avoided. Instead, individual administrators should use tools such as
su
or sudo
(for more information,
type man 1 su
or man 8 sudo
) to obtain elevated
privileges. This allows associating root
logins with particular
users. This also adds another layer of security; not only the root
password, but both the root
and the password of
an an administrator's regular account would need to be compromised to gain full root
access. This section explains how to limit direct root logins on the different
levels of the system.
2.30.1 Restricting Local Text Console Logins #
TTY devices provide text-mode system access via the console. For desktop
systems these are accessed via the local keyboard or—in case of server
systems—via input devices connected to a KVM switch or a remote
management card (ILO, DRAC, etc).
By default, Linux offers 6 different consoles, that can be switched to
via the key combinations
Alt–F1 to
Alt–F6, when
running in text mode, or
Ctrl–Alt–F1 to
Ctrl–Alt–F6
when running in a graphical session. The
associated terminal devices are named tty1
.. tty6
accordingly.
The following steps restrict root access to the first TTY. Even this access method is only meant for emergency access to the system and should never be used for everyday system administration tasks.
The steps shown here are tailored towards PC architectures (x86 and
AMD64/Intel 64). On architectures such as POWER, different terminal
device names than tty1
may be used. Be careful not to
lock yourself out completely by specifying wrong terminal device names.
You can determine the device name of the terminal you are currently
logged into by running the tty
command. Be careful
not to do this in a virtual terminal, such as via SSH or in a graphical
session (device names /dev/pts/N
),
but only from an actual login terminal reachable via
Alt–FN.
Ensure that the PAM stack configuration file
/etc/pam.d/login
contains thepam_securetty
module in theauth
block:auth requisite pam_nologin.so auth [user_unknown=ignore success=ok ignore=ignore auth_err=die default=bad] pam_securetty.so noconsole auth include common-auth
This will include the
pam_securetty
module during the authentication process on local consoles, which restrictsroot
to logging-in only on TTY devices that are listed in the file/etc/securetty
.Remove all entries from
/etc/securetty
except one. This limits the access to TTY devices for root.# # This file contains the device names of tty lines (one per line, # without leading /dev/) on which root is allowed to login. # tty1
Check whether logins to other terminals will be rejected for
root
. A login ontty2
, for example, should be rejected immediately, without even querying the account password. Also make sure that you can still successfully login totty1
and thus thatroot
is not locked out of the system completely.
Do not add the pam_securetty
module to the
/etc/pam.d/common-auth
file. This would
break the su
and sudo
commands,
because these tools would then also reject root
authentications.
These configuration changes will also cause root logins on serial
consoles such as /dev/ttyS0
to be denied. In case you
require such use cases you need to list the respective tty devices
additionally in the /etc/securetty
file.
2.30.2 Restricting Graphical Session Logins #
To improve security on your server, avoid using graphical
environments at all. Graphical programs are often
not designed to be run as root
and are more likely to
contain security issues than console programs. If you require a graphical
login, use a non-root
login. Configure your system to disallow
root
from logging in to graphical sessions.
To prevent root
from logging in to graphical sessions, you
can apply the same basic steps as outlined in
Section 2.30.1, “Restricting Local Text Console Logins”.
Just add the pam_securetty
module to the PAM
stack file belonging to the display manager—for example,
/etc/pam.d/gdm
for GDM. The graphical
session also runs on a TTY device: by default, tty7
.
Therefore, if you restrict root
logins to tty1
,
then root
will be denied login in the graphical session.
2.30.3 Restricting SSH Logins #
By default, the root
user is also allowed to log
into a machine remotely via the SSH network protocol (if the SSH port is
not blocked by the firewall). To restrict this, make the following
change to the OpenSSH configuration:
Edit
/etc/ssh/sshd_config
and adjust the following parameter:PermitRootLogin no
Restart the
sshd
service to make the changes effective:systemctl restart sshd.service
Using the PAM pam_securetty
module is not suitable in
case of OpenSSH, because not all SSH logins go through the PAM
stack during authorization (for example, when using SSH public-key authentication).
In addition, an attacker could differentiate between a wrong password
and a successful login that was only rejected later on by policy.
2.31 Setting an Inactivity Timeout for Interactive Shell Sessions #
It can be a good idea to terminate an interactive shell session after a certain period of inactivity; for example, to:
prevent open, unguarded sessions.
avoid waste of system resources.
By default, there is no inactivity timeout for shells. Nothing will happen if a shell stays open and unused for days or even years. It is possible to configure most shells in a way that idle sessions terminate automatically after a certain amount of time. The following example shows how to set an inactivity timeout for a number of common types of shells.
The inactivity timeout can be configured for login shells only or for all interactive shells. In the latter case the inactivity timeout is running individually for each shell instance. This means that timeouts will accumulate. When a sub- or child-shell is started, then a new timeout begins for the sub- or child-shell, and only afterwards will the timeout of the parent continue running.
The following table contains configuration details for a selection of common shells shipped with SLE:
package | shell personalities | shell variable | time unit | readonly setting | config path (only login shell) | config path (all shells) |
---|---|---|---|---|---|---|
|
|
| seconds |
|
|
|
|
|
| seconds |
|
|
|
|
|
| minutes |
|
|
|
|
|
| seconds |
|
|
|
Every listed shell supports an internal timeout shell variable that can be set to a specific time value to cause the inactivity timeout. If you want to prevent users from overriding the timeout setting, you can mark the corresponding shell timeout variable as readonly. The corresponding variable declaration syntax is also found in the table above.
This feature is only helpful for avoiding risks if a user is forgetful or follows unsafe practices. It does not protect against hostile users. The timeout only applies to interactive wait states of a shell. A malicious user can always find ways to circumvent the timeout and keep their session open regardless.
To configure the inactivity timeout, you need to add the matching timeout
variable declaration to each shell's startup script. Use either the path for
login shells only, or the one for all shells, as listed in the table. The
following example uses paths and settings that are suitable for
bash
and ksh
to setup a read-only login shell timeout that cannot be overridden by users.
Create the file /etc/profile.d/timeout.sh
with the
following content:
# /etc/profile.d/timeout.sh for SUSE Linux # # Timeout in seconds until the bash/ksh session is terminated # in case of inactivity. # 24h = 86400 sec readonly TMOUT=86400
We recommend using the screen
tool in order to
detach sessions before logging out. screen
sessions
are not terminated and can be re-attached whenever required. An active
session can be locked without logging out (read about
Ctrl–a–x
/
lockscreen
in man screen
for details).
2.32 Preventing Accidental Denial of Service #
Linux allows you to set limits on the amount of system resources that users and groups can consume. This is also very handy if bugs in programs cause them to use up too much resources (for example memory leaks), slow down the machine, or even render the system unusable. Incorrect settings can allow programs to use too many resources which may make the server unresponsive to new connections or even local logins (for example if a program uses up all available file handles on the host). This can also be a security concern if someone is allowed to consume all system resources and therefore cause a denial-of-service attack – either unplanned or worse, planned. Setting resource limits for users and groups may be an effective way to protect systems, depending on the environment.
2.32.1 Example for Restricting System Resources #
The following example demonstrates the practical usage of setting or
restricting system resource consumption for an Oracle user account.
For a list of system resource settings, see
/etc/security/limits.conf
or man
limits.conf
.
Most shells like Bash provide control over various resources (for example the maximum allowable number of open file descriptors or the maximum number of processes) that are available on a per/user basis. To examine all current limits in the shell execute:
root #
ulimit -a
For more information on ulimit
for the Bash shell,
examine the Bash man pages.
Setting "hard" and "soft" limits might not behave as expected when
using an SSH session. To see valid behavior it may be necessary to
login as root and then su to the id with limits (for example,
oracle
in these examples). Resource limits should
also work assuming the application has been started automatically
during the boot process. It may be necessary to set
UsePrivilegeSeparation
in
/etc/ssh/sshd_config
to "no" and restart the SSH
daemon (systemctl restart sshd
) if it
seems that the changes to resource limits are not working (via SSH).
However this is not generally recommended – as it weakens a systems
security.
ssh
You can add some extra security to your server by disabling password
authentication for SSH. Remember that you need to have SSH keys
configured, otherwise you cannot access the server. To disable
password login, add the following lines to
/etc/ssh/sshd_config
:
UseLogin no UsePAM no PasswordAuthentication no PubkeyAuthentication yes
In this example, a change to the number of file handles or open files
that the user oracle
can use is made by editing
/etc/security/limits.conf
as root
making
the following changes:
oracle soft nofile 4096 oracle hard nofile 63536
The soft limit in the first line defines the limit on the number of
file handles (open files) that the
oracle
user will have after
login. If the user sees error messages about running out of file
handles, then the user can increase the number of file handles like in
this example up to the hard limit (in this example 63536) by
executing:
root #
ulimit -n 63536
You can set the soft and hard limits higher if necessary.
It is important to be judicious with the usage of ulimits.
Allowing a "hard" limit for “nofile” for a user that
equals the kernel limit (/proc/sys/fs/file-max
)
is very bad! If the user consumes all the available file handles,
the system cannot initiate new logins as accessing (opening) PAM
modules which are required for performing a login will not be
possible.
You also need to ensure that pam_limits
is either
configured globally in /etc/pam.d/common-auth
, or
for individual services like SSH, su, login, and telnet in:
/etc/pam.d/sshd (for SSH) |
/etc/pam.d/su (for su) |
/etc/pam.d/login (local logins and telnet) |
If you do not want to enable it for all logins, there is a specific PAM
module that will read the
/etc/security/limits.conf
file. Entries in pam
configuration directives will have entries like:
session required /lib/security/pam_limits.so session required /lib/security/pam_unix.so
It is important to note that changes are not immediate and require a new login session:
root #
su – oracletux >
ulimit -n 4096
Note that these examples are specific to the Bash shell -
ulimit
options are different for other shells. The
default limit for the user oracle
is
4096
. To increase the number of file handles the
user oracle
can use to 63536
,
do:
root #
su – oracletux >
ulimit -n 4096tux >
ulimit -n 63536tux >
ulimit -n 63536
Making this permanent requires the addition of the setting,
ulimit -n 63536
, (again, for Bash) to the users
profile (~/.bashrc
or
~/.profile
file) which is the user start-up file
for the Bash shell on SUSE Linux Enterprise Server (to verify your shell run: echo
$SHELL
). To do this you could simply type (or
copy/paste – if you are reading this on the system) the following
commands for the Bash shell of the user oracle
:
root #
su - oracletux >
cat >> ~oracle/.bash_profile << EOF ulimit -n 63536 EOF
2.33 Displaying Login Banners #
It is often necessary to place a banner on login screens on all servers for legal/audit policy reasons and to deter intruders, etc.
If you want to print a legal banner after a user logs in using SSH,
local console, etc., you can leverage the /etc/motd
(motd = message of the day) file. The file exists on SUSE,
however it is empty. Simply add content to the file that is
applicable/required by the organization.
Try to keep the content to a single page (or less), as it will scroll the screen if it does not fit.
For SSH you can edit the “Banner” parameter in the
/etc/ssh/sshd_config
file which will then
appropriately display the banner text before the login prompt. For
local console logins you can edit the /etc/issue
file which will display the banner before the login prompt. For GDM,
you could make the following changes to require a user to acknowledge
the legal banner by selecting or
. Edit the
/etc/gdm/PreSession/Default
file and add the
following lines at the beginning of the script:
if ! gdialog --yesno '\nThis system is classified...\n' 10 10; then sleep 10 exit 1; fi
The This system is classified... test should be replaced with the valid text. It is important to note that this dialog will not prevent a login from progressing.
2.34 Miscellaneous #
2.34.1 Host-Based Linux Monitoring and Intrusion Detection #
Before you place a host into production or even on a network, consider
the use of a system integrity checker, like
seccheck
(already discussed in
Section 2.4, “Verifying Security Action with seccheck
”), so in case
of unauthorized changes, notifications will be issued. Also consider
the use of an intrusion detection environment, like AIDE, the
Advanced Intrusion Detection Environment.
AIDE is a GPL licensed and open source intrusion detection system. It could be considered a system fingerprinting mechanism. AIDE works by creating a database containing information about the files on your system. The database is created from rules laid out in the configuration file aide.conf. When AIDE is run, this database is referenced to check for changes (or created for the first time). Assuming a comparison check is being run, any changes not permitted by the configuration file are reported.
By leveraging AIDE—storing a copy of the host's database in a secure location—and comparing it (on a scheduled basis or as part of a forensic effort), system integrity/insurance can be a matter of heuristics and procedure. If an intruder compromises your system, the comparison effort will enable an administrator or security officer to know what has changed on the host. The initial database should be created as a final step—before a system gets deployed into production.
It is outside the scope of this article to cover Linux Monitoring and detailed Intrusion Detection systems (IDS) or solutions, however there is a plethora of information about configuring AIDE or other solutions and many informative articles on the Web.
2.34.2 Connect Accounting Utilities #
Here is a list of commands you can use to get data about user logins:
who
.
Lists currently logged in users.
w
.
Shows who is logged in and what they are doing.
last
.
Shows a list of last logged in users, including login time, logout
time, login IP address, etc.
lastb
.
Same as last
, except that by default it shows
/var/log/btmp
, which contains all the bad login
attempts.
lastlog
.
This command reports data maintained in
/var/log/lastlog
, which is a record of the last
time a user logged in.
ac
.
Available after installing the acct
package.
Prints the connect time in hours on a per-user basis or daily basis,
etc. This command reads /var/log/wtmp
.
dump-utmp
.
Converts raw data from /var/run/utmp or
/var/log/wtmp
into ASCII-parseable format.
Also check the /var/log/messages
file, or the
output of journalctl
if no logging facility is
running. See Book “Administration Guide”, Chapter 16 “journalctl
: Query the systemd
Journal” for more information on
the systemd
journal.
2.34.3 Other #
Finally, the following items are relevant to the system security as well, and misconfiguration can cause many problems – and should be reviewed:
Resolver (
/etc/hosts
,/etc/resolv.conf
,/etc/nsswitch.conf
).NTP configuration (
/etc/ntp.conf
).