![[*]](/usr/share/latex2html/icons/crossref.png){kind=icon}
for a detailed configuration example.
atsec is a trademark of atsec GmbH
ProLiant and Integrity are trademarks of Hewlett-Packard Company.
IBM, IBM logo, BladeCenter, eServer, iSeries, OS/400, PowerPC, POWER3, POWER4, POWER4+, pSeries, S390, xSeries, zSeries, zArchitecture, and z/VM are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both.
Intel, Itanium, Xeon, and Pentium are trademarks of Intel Corporation in the United States, other countries, or both.
Java and all Java-based products are trademarks of Sun Microsystems, Inc., in the United States, other countries, or both.
Linux is a registered trademark of Linus Torvalds.
UNIX is a registered trademark of The Open Group in the United States and other countries.
This document is provided AS IS with no express or implied warranties. Use the information in this document at your own risk.
This document may be reproduced or distributed in any form without prior permission provided the copyright notice is retained on all copies. Modified versions of this document may be freely distributed provided that they are clearly identified as such, and this copyright is included intact.
This document is derived from the "EAL3 Evaluated Configuration Guide for Red Hat Enterprise Linux" Copyright (c) 2003, 2004 by atsec GmbH, and IBM Corporation or its wholly owned subsidiaries.
Changes Copyright (c) 2004, 2005, 2006, 2007 by atsec inc., Hewlett-Packard Company or its wholly owned subsidiaries.
The Red Hat Enterprise Linux (RHEL) distribution is designed to provide a secure and reliable operating system for a variety of purposes. Because security requirements obviously depend on the applications and environment, it is not possible to simply certify that the system is "secure", a more precise definition is needed.
The Common Criteria (CC) provides a widely recognized methodology for security certifications. A CC evaluation is fundamentally a two-step process, consisting of defining the "security target" which describes the features that are to be evaluated, and then testing and verifying that the system actually implements these features with a sufficient level of assurance.
This document is a security guide that explains how to set up the evaluated configuration, and provides information to administrators and ordinary users to ensure secure operation of the system. It is intended to be self-contained in addressing the most important issues at a high level, and refers to other existing documentation where more details are needed.
The document primarily addresses administrators, but the §6 section in this guide is intended for ordinary users of the system as well as administrators.
Knowledge of the Common Criteria is not required for readers of this document.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 (http://www.ietf.org/rfc/rfc2119.txt).
Note that the terms "SHOULD" and "SHOULD NOT" are avoided in this document. Requirements are either absolute (and marked with MUST and equivalent terms), or entirely optional (in the sense of not affecting required security functions) and marked with RECOMMENDED, MAY or OPTIONAL.
If you follow the requirements in this document when setting up and using the system, your configuration will match the evaluated configuration. Certain configuration options are marked as OPTIONAL and you MAY modify them as needed, but you MUST NOT make other changes, because they will make the system fail to match the evaluated configuration.
Of course, you MUST always use common sense. This document is not a formal specification, and legitimate reasons may exist to modify the system setup in ways not described here if that is necessary for the system to fulfill its intended purpose. Specifically, applying security patches released by the vendor is strongly RECOMMENDED even though that will cause a deviation from the evaluated configuration.
In cases where the requirements and recommendations in this document conflict with those in other sources (such as the online documentation), the information in this Configuration Guide has higher precedence. You MUST follow the steps described here to reach the evaluated configuration, even if other documentation describes different methods.
The usual convention is used in this guide when referring to manual pages that are included in the software distribution. For example, the notation ls(1) means that running the man -S 1 ls command will display the manual page for the ls command from section one of the installed documentation. In most cases, the -S flag and the section number may be omitted from the command, they are only needed if pages with the same name exist in different sections.
A system can be considered to be "CC compliant" if it matches an evaluated and certified configuration. This implies various requirements concerning hardware and software, as well as requirements concerning the operating environment, users, and the ongoing operating procedures.
Strictly speaking, an evaluation according to the CC represents the results of investigation of the security properties of the target system according to defined guidelines. It should not be considered as a guarantee for fitness for any specific purpose, but should provide help in deciding the suitability of the system considering how well the intended use fits the described capabilities. It is intended to provide a level of assurance about the security functions that have been examined by a neutral third party.
The Security Target defines the security features of the product, the feature lists are based on Protection Profiles that are independent of the specific product. For this evaluation, the system offers a choice of two different variants of the evaluated configuration:
The CAPP mode, matching the Controlled Access Protection Profile. In this mode, the only supported access control mechanism is discretionary access control, based on user and group IDs, permission bits, and access control lists (ACLs).
The LSPP/RBAC mode, matching both the Labeled Security Protection Profile (which is based on and includes all features of CAPP), and the Role-Based Access Control Protection Profile. In addition to discretionary access control, this mode supports two forms of mandatory access control; multilevel security (MLS) and role-based access control.
This guide will specify which mode the information applies to by referring to either CAPP mode or LSPP/RBAC mode as appropriate. If not specified, the instructions apply to both configurations.
The hardware MUST be the one of the HP systems specified in the Security Target document:
HP Intel Itanium2 (single and multi-core) processor based servers:
HP Integrity Superdome product line
HP Integrity rx product line
HP Integrity cx product line
HP Integrity BL product line
Intel Xeon based servers with EM64T 64bit extensions (single and multi-core), and HP AMD Opteron processor (single and multi-core):
HP Proliant ML product line (EM64T capable models)
HP Proliant DL product line (EM64T capable or Opteron models)
HP ProLiant BL product line (EM64T capable or Opteron models)
HP Intel Pentium and Xeon processor based servers without EM64T extensions:
HP Proliant ML product line (except EM64T capable models)
HP Proliant DL product line (except EM64T capable or Opteron models)
HP Proliant BL product line (except EM64T capable or Opteron models)
HP Intel Xeon processor based systems:
HP xw product line
HP Intel Pentium 4 processor based systems:
HP xw product line
HP Compaq dc series product line
It is NOT permitted to install the operating system within a nPar hardware partition.
The following types of printers in the CAPP mode are supported:
Printers supporting PCL version 4 (parallel, USB, and Ethernet)
Printers supporting PostScript level 1 (parallel, USB, and Ethernet)
The following types of printers in the LSPP/RBAC mode are supported:
Printers supporting PCL version 4 (parallel, USB)
Printers supporting PostScript level 1 (parallel, USB)
Only printers supporting the above mentioned languages and are connected to the specified interfaces are allowed.
All Ethernet and Token Ring network adapters supported by the operating system are permitted. Modems, ISDN and other WAN adapters are not supported in the evaluated configuration.
You MAY attach the following peripherals without invalidating the evaluation results. Other hardware MUST NOT be installed in or attached to the system.
All storage devices and backup devices supported by the operating system software (hard disks, CDROM and DVDROM drives, streamer drives, floppy disk drives) (except hot pluggable devices connected via USB or IEEE 1394 (Firewire) interfaces).
Operator console consisting of a keyboard, video monitor, and optionally mouse. Additionally, you MAY directly attach supported serial terminals (see section §4.9 of this guide), but not modems, ISDN cards, or other remote access terminals.
USB keyboards and mice MAY be attached, as some of the supported hardware platforms would otherwise not have supported console input devices. If a USB keyboard or mouse is used, it MUST be connected before booting the operating system, and NOT added later to a running system. Other hot-pluggable hardware that depends on the dynamic loading of kernel modules MUST NOT be attached. Examples of such unsupported hardware are USB and IEEE1394/FireWire peripherals other than mice and keyboards.
Running the certified software on other similar hardware may result in an equivalent security level, but the certification does not apply if the hardware is different from that used for the testing processes during the evaluation.
The software MUST match the evaluated configuration. In the case of an operating system, this also requires that the installed kernel, system, and application software are the same. The documentation (including this guide) will specify permitted variations, such as modifying certain configuration files and settings, and installing software that does not have the capability to affect the security of the system (typically those that do not require 'root' privileges).
Stated requirements concerning the operating environment MUST be met. Typical requirements include a secure location for the hardware (protected from physical access by unauthorized persons), as well as restrictions concerning permitted network connections.
For more information about these requirements, please refer to §1.4.
The operation of the system MUST be in agreement with defined organizational security policies, to ensure that actions by administrators and users do not undermine the system's security.
The security target covers one or more systems running RHEL, networked in a non-hostile network, with a well-managed and non-hostile user community. It is not intended to address the needs of a directly Internet-connected server, or the case where services are to be provided to potentially hostile users.
You MUST set up the server (or servers) in a physically secure environment, where they are protected from theft and manipulation by unauthorized persons.
You MUST ensure that all connections to peripheral devices and all network connections are protected against tampering, tapping and other modifications. Using the secured protocols SSHv2 or SSLv3 is considered sufficient protection for network connections. All other connections must remain completely within the physically secure server environment.
All components in the network such as routers, switches, and hubs that are used for communication are assumed to pass the user data reliably and without modification. Translations on protocols elements (such as NAT) are allowed as long as those modifications do not lead to a situation where information is routed to somebody other than the intended recipient system.
If other systems are connected to the network they MUST be configured and managed by the same authority using an appropriate security policy not conflicting with the security policy of the target of evaluation. All links from this network to untrusted networks (such as the Internet) need to be protected by appropriate measures like carefully configured firewall systems that prevent attacks from the untrusted networks.
Be aware that information passed to another system leaves the control of the sending system, and the protection of this information against unauthorized access needs to be enforced by the receiving system. If an organization wants to implement a consistent security policy covering multiple systems on a network, organizational procedures MUST ensure that all those systems can be trusted and are configured with compatible security configurations enforcing an organization wide security policy. How to do this is beyond the scope of this Configuration Guide. If you set up a communication link to a system outside your control, please keep in mind that you will not be able to enforce any security policy for any information you pass to such a system over the communication link or in other ways (for example, by using removable storage media).
Every person that has the ability to perform administrative actions by switching to root has full control over the system and could, either by accident or deliberately, undermine the security of the system and bring it into an insecure state. This Configuration Guide provides the basic guidance how to set up and operate the system securely, but is not intended to be the sole information required for a system administrator to learn how to operate Linux securely.
It is assumed, within this Configuration Guide, that administrators who use this guide have a good knowledge and understanding of operating security principles in general and of Linux administrative commands and configuration options in particular. It is strongly advised that an organization that wants to operate the system in the evaluated configuration nevertheless have their administrators trained in operating system security principles and RHEL security functions, properties, and configuration.
Every organization needs to trust their system administrators not to deliberately undermine the security of the system. Although the evaluated configuration includes audit functions that can be used to make users accountable for their actions, an administrator is able to stop the audit subsystem and reconfigure it such that his actions no longer get audited. Well trained and trustworthy administrators are a key element for the secure operation of the system. This Configuration Guide provides the additional information a system administrator should obey when installing, configuring and operating the system in compliance with the requirements defined in the Security Target for the Common Criteria evaluation.
The security target addresses the security needs of cooperating users in a benign environment, who will use the system responsibly to fulfill their tasks.
Note that system availability is not addressed in this evaluation, and a malicious user could disable a server through resource exhaustion or similar methods.
The requirements for users specifically include:
User accounts MUST be assigned only to those users with a need to access the data protected by the system, and who MUST be sufficiently trustworthy not to abuse those privileges. For example, the system cannot prevent data from being intentionally redistributed to unauthorized third parties by an authorized user.
All users of the system MUST be sufficiently skilled to understand the security implications of their actions, and MUST understand and follow the requirements listed in section §6 of this guide. Appropriate training MUST be available to ensure this.
In the LSPP/RBAC mode, procedures MUST exist for granting users authorization for access to specific security levels.
In the LSPP/RBAC mode, procedures MUST exist for establishing the security level of all information imported into the system, for establishing the security level for all peripheral devices (such as printers, tape drives, and disk drives) attached to the system, and marking a sensitivity label on all output generated.
In the RBAC/LSPP mode, administrators MUST assign roles to users in a way that ensures that the roles accurately refelect the user's job function, responsibilities, qualifications, and/or competencies within the organization.
In the RBAC/LSPP mode, users who create new data objects are considered owners for those data objects. The organization is considered to be the owner for the rest of the information under the system's control.
It is the responsibility of the system administrators to verify that these requirements are met, and to be available to users if they need your help in maintaining the security of their data.
The evaluation covers a fresh installation of RHEL Version 5 Server or Client, on one of the supported hardware platforms as defined in section §2.2.1 of this guide.
The evaluated configuration MUST be the only operating system installed on the server.
Please refer to section §2.2.1 of this guide for more information about the hardware platforms and supported peripherals.
This section describes the detailed steps to be performed when installing the RHEL operating system on the target server.
All settings listed here are REQUIRED unless specifically declared otherwise.
You will need the following components to install a system in the evaluated configuration as explained in the following sections. You will need:
The target system that will be installed, refer to section §2.2.1 of this guide for the list of supported hardware. The target system REQUIRES at least one local hard drive that will be erased and repartitioned for use by the evaluated configuration.
A static IP address if you are intending to attach the target system to a network; the evaluated configuration does not support DHCP. In addition, you will need to configure the netmask, gateway, and DNS server list manually.
An Internet-connected system equipped with the rpm and rpm2cpio package management tools. This system does not need to be in the evaluated evaluated configuration, and no packages will be installed on it. It is used to download and verify the installation packages.
A method to transfer the kickstart installation configuration and RPM packages to the target system. You can use any one of the following choices:
A CD-R containing the installation files.
A USB memory stick or USB external hard drive with a capacity of at least 32 MB, and formatted using either the vfat or ext3 file system.
A network server configured to provide the installation files via the HTTP or NFS protocol.
Note that a floppy disk drive is not suitable due to insufficient capacity.
You MUST download the distribution ISO images from the Red Hat Network on a separate Internet-connected computer, and either burn CD-Rs from them, or make the contents available on a file server via NFS or HTTP. The download location https://rhn.redhat.com/network/software/download_isos_full.pxt contains links to the platform-specific images.
You MUST use Red Hat Enterprise Linux 5 Server or Red Hat Enterprise Linux 5 Client. Make sure that you are using the appropriate version for your platform, as listed below.
You MUST verify that the image files are authentic. The RHN download site lists MD5 checksums for the image files. The web site uses SSL, you MUST check that the web server's certificate is valid to ensure that you are using authentic information. Run md5sum *.iso to view the checksums for the downloaded images, and compare them with those shown on the authenticated web page.
You MUST download several additional packages not included on the ISO images to set up the evaluated configuration. The packages are available at the following location:
ftp://ftp.redhat.com/pub/redhat/linux/eal/EAL4_RHEL5/HP/
Download the RPMs using a separate Internet-connected computer. Do NOT install the downloaded packages yet.
You MUST select the appropriate RPM packages for your architecture. The 64bit architectures support execution of both 64bit and 32bit binaries, but this functionality is disabled for ia64/Itanium in the evaluated configuration.
This is a 32bit-only platform. Use *.i686.rpm variants of packages if available, *.i386.rpm or *.noarch.rpm otherwise.
This system uses a 64bit kernel and 64bit userspace programs, and also supports running 32bit programs. Use the *.x86_64.rpm or *.noarch.rpm variants of packages. You may OPTIONALLY install the *.i386.rpm or *.i686.rpm variants of libraries (package names containing -libs or -devel) in addition to the 64bit versions.
This is a 64-bit-only platform. Use the *.ia64.rpm or *.noarch.rpm versions of packages. It can support 32bit i386 applications, but that functionality is disabled for the evaluated configuration.
The files needed are the capp-lspp-eal4-config-hp RPM and the unpacked kickstart file (contained within the capp-lspp-eal4-config-hp RPM).
The installation will use an SMP capable kernel on all systems by default.
You MUST have the Red Hat package signing key available to verify the integrity of the capp-lspp-eal4-config-hp RPM package. It is available at https://www.redhat.com/security/db42a60e.txt
On the download system, run the following commands to verify the package integrity:
rpm --import db42a60e.txt
rpm --checksig capp-lspp-eal4-config-hp-*.rpm
This MUST display the status "gpg OK". If it does not, you MUST NOT proceed with the installation using that file.
The web page https://www.redhat.com/security/team/key/ provides additional information about the usage of package signing keys.
Next, on the download system, unpack the contents of the capp-lspp-eal4-config-hp RPM into a temporary directory:
mkdir inst
cd inst
rpm2cpio ../capp-lspp-eal4-config-hp-*.rpm | cpio -id
This will create the following directory structure in the current working directory:
# this guide, and supporting documentation
./usr/share/doc/capp-lspp-eal4-config-hp-*/
GPL.txt
README*.txt
RHEL5-CC-EAL4-HP-Configuration-Guide.*
# the kickstart configuration used to automate the installation
./usr/share/capp-lspp/kickstart/
ks-i386.cfg
ks-ia64.cfg
ks-x86_64.cfg
# the evaluated configuration reconfiguration script
./usr/sbin/
capp-lspp-config
# configuration files used for the evaluated configurartion
./usr/share/capp-lspp/
auditd.conf
[...]
xinetd.conf
Depending on the installation method you choose, do one of the following steps:
Burn a CD-R containing the kickstart files from ./usr/share/capp-lspp/kickstart/ and the downloaded RPM packages, with all files at the top directory level (no subdirectories).
Copy the kickstart files from ./usr/share/capp-lspp/kickstart/ and the downloaded RPM packages onto a USB memory stick or USB external hard drive (with a capacity of at least 32 MB, and formatted using either the vfat or ext3 file system). Put all files at the top directory level (no subdirectories).
Configure a network server to provide the installation files via the HTTP or NFS protocol. Put all the downloaded RPM packages and the kickstart files from ./usr/share/capp-lspp/kickstart/ into a single directory with no subdirectories.
You MAY make changes to specific sections of the kickstart configuration. You MUST NOT change any settings not explicitly listed in this section.
Default: us
You MAY select a different keyboard mapping.
Default: -default=en_US.UTF-8 en_US.UTF-8
You MAY add additional language support, but MUST NOT change the default language or remove the en_US language support. (Users MAY configure individual language preferences to override the default.)
Default: America/New_York
You MAY select a different time zone.
Default: -disabled
You MAY enable the firewall and modify the firewall settings, but this is beyond the scope of this guide.
Default: -enforcing
For the LSPP/RBAC mode, you MUST NOT disable SELinux. If SELinux is disabled, the system will not be compliant with LSPP/RBAC any more. Nevertheless, it will still fulfill CAPP.
You MAY delete packages from the optional packages list
You MAY modify the default partitioning scheme in this function in the kickstart file, search for the following comment text:
## Required partitions, resize as appropriate
## Optional partitions, (de)activate and resize as appropriate
Note that you will have an opportunity to modify the partition settings during the install, please refer to section §2.2.5 of this guide for more information. Alternatively, you MAY use the Logical Volume Manager (LVM) to resize and add partitions after the installation is complete as documented in the lvm(8) manual page.
It is RECOMMENDED that you disconnect all network connections until the post-install system configuration is finished. You MAY use a network if required for the installation (for example when using a NFS or HTTP network server instead of CD-ROMs). If you do use a network, you MUST ensure that this network is secure.
Launch the installation boot program contained on the CD-ROM. The details of how to do this depend on the hardware platform, please refer to the hardware manuals and the Red Hat Enterprise Linux Installation Guide. Typically, insert the first CD and boot from CD-ROM.
At the boot loader prompt, you MUST initiate the preconfigured "kickstart" install using a configuration file specific for the evaluated configuration. The installer supports multiple methods to locate the kickstart information file.
You MAY use DHCP to temporarily configure the network during the installation process, but you MUST assign a static IP address for use in the evaluated configuration.
Please refer to the Red Hat Enterprise Linux Installation Guide for more information.
The first boot parameter is the name of the booted kernel image, this is always linux for installation.
You MUST use the ks= boot parameter that selects a kickstart based automated installation.
Choose the appropriate kickstart file for your architecture and distribution:
# ia64 (Itanium 2)
ks-ia64.cfg
# i386/i686 (Xeon without EM64T 64-bit support)
ks-i386.cfg
# x86_64 (Xeon with EM64T 64-bit support, or Opteron)
ks-x86_64.cfg
The installation process will prompt for all needed information, alternatively you MAY supply the following command line parameters to automate the installation:
Select one of the supported methods for accessing the distribution media:
method=cdrom:
method=nfs:server.example.com:/path/to/files/
method=http://server.example.com/path/to/files/
method=hd://sda1/path/to/files/
Use this network interface for the kickstart installation, default eth0.
Configure the network parameters for the installation. See also ksdevice.
Specify the fully qualified host name for the system, for example:
hostname=rhel5.example.com
(This parameter is specific to the CAPP/LSPP kickstart install and not generally available)
Comma separated lists of disk devices, without /dev/ prefix. Delete all data from the specified disk(s) and partition them for the evaluated configuration. This will DESTROY the data on these disks without prompting, use with care. Example:
instdisk=sda,sdb
(This parameter is specific to the CAPP/LSPP kickstart install and not generally available)
You MAY use a serial console to control the installation. Add the following parameter to activate a serial console attached to the first serial port (COM1):
console=ttyS0
You MAY use a computer using terminal emulation software and a null modem cable instead of a standalone serial terminal. You MUST ensure that the serial terminal is secure.
Examples:
# kickstart on USB storage device, install from CD
linux ks=hd:sda1:/ks-ia64.cfg method=cdrom:
# interactive network install, get IP address via DHCP
linux ks=http://example.com/rhel4/ks-ia64.cfg
# noninteractive network install (all on a single line)
linux ip=172.16.204.4 netmask=255.255.255.0 gateway=172.16.204.1
dns=172.16.204.1
ks=http://example.com/rhel4/ks-i386.cfg
method=cdrom:
hostname=rhel4.example.com
instdisk=sda
You MAY manually edit the partitioning instructions during the kickstart process. This section describes the partitioning requirements.
Set up the REQUIRED / (root) and /var/log partitions, and as many additional mounted partitions as appropriate. /var/log REQUIRES at least 100 MB of space in order to be able to install and launch the audit system, but this does not include the additional space needed for saved audit logs. You MAY use a /var/log/audit/ partition separate from /var/log/ to ensure that audit data is stored separately from other system logs. Please refer to section §5.3 of this guide for more information.
Some configurations (recognized automatically by the installation program) need a separate /boot partition formatted as an ext3 file system. If the installation program warns about the partitioning being invalid and that it may result in an unbootable system, add the /boot partition.
The ia64 (Itanium) systems require a separate partition /boot/efi/ to contain the boot loader data and program files, which MUST be formatted using the vfat file system. On ia64 systems, there is no need for a separate /boot partition in addition to the REQUIRED /boot/efi/ partition.
It is RECOMMENDED to also use separate partitions for /var, /var/log/audit/, /home and /tmp. The following table shows a RECOMMENDED partitioning scheme together with minimum sizes for the partitions. Using more space is RECOMMENDED:
/boot 75 MB # non-ia64, if needed by installer
/boot/efi 75 MB # REQUIRED on ia64 as vfat partition
/ 1200 MB
/tmp 200 MB
/home 100 MB
/var 384 MB
/var/log/audit 100 MB needed for install, >>1GB for use
All mounted partions MUST be of type ext3 or swap and formatted, except for /boot/efi/ which MUST be type vfat.
Configuring a swap partition at least as large as the installed RAM is RECOMMENDED.
The following transcript shows an example of the interactions during the pre-install phase of the configuration:
------------------------------------------------------------------------------- *** Common Criteria configuration kickstart ***
Using volume group 'VolGroup01'. (Answer '!' at any prompt to get an interactive shell)
Installation source [cdrom:] ?
Available destination disks: sda 70001
Install on which disk(s), comma separated [sda] ?
Hostname (fully qualified) [rhel5.example.com] ?
Network interface [eth0] ?
IP address [] ? 192.168.1.4
Netmask [255.255.255.0] ?
Gateway [] ? 192.168.1.254
Nameserver list (comma separated) [] ? 192.168.1.3,192.168.11.1
Manually edit partitioning instructions (y/n) [n] ?
--- WARNING ------------------------------------------------- This is your last chance to stop the installation. Continuing will erase the destination disk and install noninteractively. Answer 'n' if you need to edit your settings.
Okay to proceed with install on sda (y/n) [n] ? y -------------------------------------------------------------------------------
After the main part of the installation is complete, you will be interactively prompted for additional settings.
This is where you can choose between the two configuration variants of the evaluated configuration, select capp for the CAPP mode, and select lspp for the LSPP/RBAC mode.
The following transcript shows an example of the interactions during the post-install phase of the configuration:
------------------------------------------------------------------------------- *** Common Criteria configuration kickstart ***
Protection profile (capp or lspp) [capp] ? lspp
Please verify the system time and date:
Local time: Mon May 7 21:23:58 CEST 2007
Universal time (UTC): Mon May 7 19:23:58 UTC 2007
If the time or time zone is wrong, please correct it now using tools such as 'date', 'hwclock', or 'tzselect' as appropriate.
Is the time correct (y/n) [y] ? Bringing up loopback interface: [ OK ] Bringing up interface eth0: [ OK ]
Need to install the certification RPM and updated RPM packages:
capp-lspp-eal4-config-hp-[...].noarch.rpm [...] vixie-cron-4.1-68.el5.i386.rpm
Supply a web URL or a local (absolute) directory name.
If you need to mount a device containing the files, enter '!' and RETURN to get a shell prompt.
Location [ftp://ftp.redhat.com/pub/redhat/linux/eal/EAL4_RHEL5/HP/] ?
[...downloading...]
Preparing... ########################################### [100%]
1:audit-libs ########################################### [ 3%]
[...]
31:vixie-cron ########################################### [100%]
Switching SELinux to MLS mode... Fixing file labels... [...]
Please enter the password for the root account. Changing password for user root. New UNIX password: ********* Retype new UNIX password: ******** passwd: all authentication tokens updated successfully.
Create an administrative user account.
Real name (First Last) [] ? John Smith
Userid [jsmith] ? Changing password for user jsmith. New UNIX password: ******** Retype new UNIX password: ******** passwd: all authentication tokens updated successfully.
Add more administrative users (y/n) [n] ? --- Mon May 7 21:25:38 CEST 2007 script running: /usr/sbin/capp-lspp-config args -a --lspp [...] Reconfiguration successful. It is now necessary to reboot the system. After the reboot, your system configuration will match the evaluated configuration *** Reboot the system? (y/n) [y]: y ------------------------------------------------------------------------------- You MUST reboot the system when the installation is complete.
The automated kickstart post-install procedure reconfigures the system into the evaluated configuration noninteractively.
In RBAC/LSPP mode, you MUST follow the additional instructions in section §6.6 which are not automated.
This section describes the steps done during the automated process for informational purposes. It is NOT supported to attempt manually configuring the evaluated configuration.
You MAY use the capp-lspp-config after the initial installation script to re-run selected configuration steps automatically.
After software upgrades or installation of additional packages, these steps MUST be re-done or at least re-checked to ensure that the configuration remains secure.
The evaluated configuration disables direct "root" login over the network. All system administrators MUST log in using a non-root individual user ID, then use the su(8) command to gain superuser privileges for administrative tasks. This requires membership in the 'wheel' group of trusted users.
You MUST define at least one non-root user account with the useradd(8) command, and add this user account to the 'wheel' group. Note that the enhanced password quality checking mechanisms and the password expiry settings of the evaluated configuration are not active yet. You must manually set the password properties in accordance with the password policy.
The kickstart script creates one or more administrative user accounts in the postinstall section.
Here is an example of creating an additional user account:
useradd -m -c "John Doe" -G wheel jdoe
passwd jdoe
chage -m 1 -M 60 -W 7 jdoe
Please refer to sections §4.7 and §6.3 of this guide for more information on creating user accounts.
In LSPP/RBAC mode, the initial administrative users MUST be assigned to the sysadm_r role, and it is RECOMMENDED to permit the full range (SystemLow-SystemHigh) of MLS labels for them. For example:
semanage login -a -s staff_u -r SystemLow-SystemHigh jdoe
restorecon -r /home/jdoe
Several packages shipped on the installation media MUST be replaced with more recent versions to fix bugs or add additional features required for the evaluated configuration.
The kickstart script automatically installs the required updates in the postinstall section.
The kickstart script installs the capp-lspp-eal4-config-hp RPM package and runs the capp-lspp-config script contained within that RPM package noninteractively.
You MAY run the capp-lspp-config script interactively after installation is complete to verify and reset configuration settings to appropriate values for the evaluated configuration.
The capp-lspp-eal4-config-hp package contains EAL4 specific configuration files, and the script capp-lspp-config that sets up the evaluated configuration.
Run the following command to view a summary of the supported options:
capp-lspp-config -h
You MAY use the -a flag to automate the install and have it run without prompting. This is intended for people who are familiar with the process; if running it for the first time you SHOULD let it run interactively and verify the actions as described in this guide.
You MUST answer all questions asked by the script that are not marked as "optional" with y to achieve the evaluated configuration.
WARNING: Switching between CAPP and LSPP modes in an installed system is not supported. Please reinstall if you want to change the configuration type.
WARNING: The capp-lspp-config script will reboot the system as the final step in the process, as described in the manual instructions in section §3.16 of this guide. Remember to remove any CD-ROM from the drive and/or configure the system to boot from hard disk only.
The remaining steps in this chapter were done automatically if the kickstart install has completed successfully, or if you have run the script at a later time. You MAY skip ahead to section §4 of this guide.
The kickstart automated install uses a default package selection that contains all packages required for the evaluated configuration. It also installs several optional packages that you MAY remove once the installation is complete.
The following optional packages MAY be deleted from the system:
audit-libs-devel
autoconf
automake
bison
cvs
cyrus-sasl-devel.@@native@@
elinks
expect
expect-devel
flex
gcc
gcc-c++
keyutils-libs
keyutils-libs-devel
libattr-devel
libcap-devel
libselinux-devel.@@native@@
libsemanage-devel.@@native@@
libsepol-devel.@@native@@
libuser-devel.@@native@@
make
openssl-devel.@@native@@
pam-devel.@@native@@
pciutils-devel
perl-Digest-HMAC
perl-Digest-SHA1
python-devel
readline-devel
rpm-build
strace
swig
tcl
texinfo
tk
zlib-devel
In addition to the preselected packages, certain additional software from the RHEL CDs MAY be installed without invalidating the evaluated configuration. The rules described in section §4.4 of this guide MUST be followed to ensure that the security requirements are not violated.
Note: The system runlevel as specified in the 'initdefault' entry in /etc/inittab MUST remain at the default setting of '3' for these steps to be valid.
The following services are REQUIRED for runlevel 3:
auditd # the audit daemon
crond # vixie-cron
irqbalance # configures SMP IRQ balancing
kudzu # new device discovery
network # network interface configuration
syslog # system logging
The following services are OPTIONAL for runlevel 3:
cups # print subsystem
gpm # console mouse management
mdmonitor # software raid monitoring
postfix # SMTP MTA
rawdevices # Raw partition management (eg. for Oracle)
sshd # Secure Shell
vsftpd # FTP server
xinetd # Internet Services
You MUST ensure that all REQUIRED services are active. You MAY enable or disable services from the OPTIONAL list as suitable for your configuration. All other services MUST be deactivated.
Use chkconfig SERVICENAME off to disable a service, and chkconfig SERVICENAME on to enable it. The following command lists the active services:
chkconfig --list | grep "3:on" | sort
Make sure that the audit subsystem is activated. If auditd is not running, all logins are automatically disabled in the evaluated configuration as required by CAPP/LSPP.
Login from the network with user ID 0 ('root') MUST NOT be permitted over the network. Administrators MUST use an ordinary user ID to log in, and then use the /bin/su - command to switch identities. For more information, refer to section §4.3 of this guide.
It is RECOMMENDED that you remind administrators of this by adding the following alias to the bash configuration file /etc/profile that disables the pathless 'su' command:
alias su="echo \"Always use '/bin/su -' (see Configuration Guide)\""
This alias can be disabled for the root user in /root/.bashrc:
unalias su
The restriction for direct root logins is enforced through two separate mechanisms. For network logins using ssh, the PermitRootLogin no entry in /etc/ssh/sshd_config MUST be set (see next section). Console and serial terminal logins use the pam_securetty.so PAM module in the /etc/pam.d/login file that verifies that the terminal character device used is listed in the file /etc/securetty.
The file /etc/securetty MUST NOT be changed from the secure default settings.
Note that some systems configure the serial console, for example ttyS0, as a secure terminal. If that is the case (it is listed in the /etc/securetty file), you MUST NOT allow non-administrative users to use this serial terminal.
SSH protocol version 1 MUST be disabled. It has known security deficiencies.
The ssh client MUST NOT be set up SUID root (the SUID bit was removed in the post-install configuration). This prevents the use of some authentication methods normally supported by OpenSSH, but does not affect the evaluated configuration that uses password authentication exclusively.
The SSH Server MUST be configured to reject attempts to log in as root.
The permitted authentication mechanisms are per-user (nonempty) passwords and per-user DSS public key authentication. All other authentication methods MUST be disabled.
The setting PAMAuthenticationViaKbdInt MUST be disabled, since this would otherwise circumvent the disabled root logins over the network.
This results in the following option set for the SSH daemon that MUST be set in /etc/ssh/sshd_config:
# /etc/ssh/sshd_config # # CAPP/LSPP configuration. Please read the Evaluated Configuration Guide # before making changes. # # Cryptographic settings. Disallow the obsolete (and # insecure) protocol version 1, and hardcode a strong # cipher. Protocol 2 Ciphers 3des-cbc
# Configure password-based login. This MUST use the PAM # library exclusively, and turn off the builtin password # authentication code. UsePAM yes ChallengeResponseAuthentication yes PasswordAuthentication no PermitRootLogin no PermitEmptyPasswords no
# No other authentication methods allowed IgnoreRhosts yes RhostsRSAAuthentication no HostbasedAuthentication no PubkeyAuthentication no RSAAuthentication no KerberosAuthentication no GSSAPIAuthentication no
# Other settings, MAY change "X11Forwarding" to "yes" X11Forwarding no Subsystem sftp /usr/libexec/openssh/sftp-server
All other options MUST NOT be changed from the defaults or from those settings specified here. Specifically, you MUST NOT add other authentication methods (AFS, Kerberos, host-based) to those permitted here.
For the LSPP/RBAC mode, please refer to section §6.6 for more information.
In CAPP mode, the xinetd super server is not used in the evaluated configuration, but MAY be used to start non-root network processes. The file /etc/xinetd.conf contains default settings, these can be overridden by service-specific entry files stored in the directory /etc/xinetd.d/.
The xinetd.conf(5) man page contains more information on xinetd and configuration examples.
In the LSPP/RBAC mode, xinetd is used to launch sshd at the appropriate MLS level based on the MLS level of the incoming network connection. Refer to section §6.6 for more information.
The evaluated configuration OPTIONALLY includes FTP services. Note that FTP does not provide support for encryption, so this is only RECOMMENDED for anonymous access to non-confidential files. If you do not specifically need FTP, it is RECOMMENDED that you disable the vsftpd(8) service.
Use the chkconfig(8) command to control the FTP service:
# Activate FTP service
chkconfig vsftpd on
# Disable FTP service
chkconfig vsftpd off
The vsftpd service uses several additional configuration files. In /etc/vsftpd/vsftpd.conf the configuration of the ftp daemon is specified. In addition, the file /etc/vsftpd.ftpusers is used for access control. Users listed in that file can NOT log in via FTP. This file initially contains all system IDs and the root user. It can be augmented with other IDs according to the local needs, but the root entry MUST NOT be removed. The ftpusers file is not checked by the ftp daemon itself but by a PAM module. Please see section §4.16.2 of this guide for details.
The setup of /etc/vsftpd/vsftpd.conf depends on the local needs. Please refer to vsftpd.conf(5) for details.
The default configuration permits anonymous FTP. This setting is only suitable for distribution of public files for which no read access control is needed. The default configuration uses the following /etc/vsftpd/vsftpd.conf settings:
anonymous_enable=YES
local_enable=NO
Yoy MAY disable anonymous FTP with the following /etc/vsftpd/vsftpd.conf setting:
anonymous_enable=NO
You MAY enable FTP authentication for local user accounts with the following /etc/vsftpd/vsftpd.conf setting:
local_enable=YES
You MUST use the following option in /etc/vsftpd/vsftpd.conf to activate PAM session handling:
session_support=YES
It is RECOMMENDED to use the more secure alternatives sftp(1) or scp(1) to copy files among users, and to use FTP only for legacy applications that do not support this alternative.
This section applies in CAPP mode only. In the LSPP/RBAC mode, the mail system will not work correctly as it is not MLS aware. It does not need to be specifically disabled as it does not have any MLS override privileges.
You MUST disable the execution of user-specified programs when receiving mail via entries in the $HOME/.forward files of individual users. Add the following line to the /etc/postfix/main.cf file:
allow_mail_to_commands = alias
It is RECOMMENDED that you set up an alias for root in the /etc/aliases file. Specify one or more user names of administrators to whom mail addressed to root will be forwarded, for example with the following entry in the /etc/aliases file:
root: jdoe, jsmith
You must then rebuild the aliases database and reload the database using the following commands:
newaliases
postfix reload
Please see postfix(1), master(8), aliases(5), newaliases(1), and the documentation in /usr/share/doc/postfix*/ for details.
Use of the Cups printing system is OPTIONAL, if the service is active you MUST configure the settings described in this section.
In the evaluated configuration, the cupsd process runs as user lp and group lp.
Verify that the printer daemon is able to access your printer devices with these permissions. You MAY need to reconfigure the printer device access rights to match, for example by setting the device owner for the /dev/lp* devices to the lp user in the /etc/udev/permissions.d/50-udev.permissions file. Please refer to the cupsd.conf(5) and cupsd(8) man pages for more information.
This section applies only to the LSPP/RBAC mode.
You MUST configure the printer device file with the correct SELinux type (printer_device_t and the MLS range corresponding to the range of levels permitted to print on that device, i.e.:
cupsdisable
chcon -t printer_device_t /dev/lp0
chcon -l s0-s3:c0,c1,c7 /dev/lp0
cupsenable
The print queue MUST be disabled while changing the device settings.
Please refer to the cupsd(8) and cupsd.conf(5) manual pages for additional information.
The PAM subsystem is responsible for maintaining passwords and other authentication data. Because this is a security-critical system, understanding how it works is very important. In addition to the pam(8) manual page, full documentation is available in /usr/share/doc/pam-*/txts/ and includes "The Linux-PAM System Administrator's Guide" (pam.txt) as well as information for writing PAM applications and modules. Detailed information about modules is available in /usr/share/doc/pam-*/txts/README.pam_* as well as manual pages for individual modules, such as pam_stack(8).
The PAM configuration is stored in the /etc/pam.d/ directory. Note that the documentation refers to a file /etc/pam.conf that is not used by RHEL (PAM was compiled to ignore this file if the /etc/pam.d/ directory exists).
Each service (application) that uses PAM for authentication uses a service-name to determine its configuration, stored in the /etc/pam.d/SERVICE_NAME file. The special service-name OTHER (case insensitive) is used for default settings if there are no specific settings.
The configuration file for the service contains one entry for each module, in the format:
module-type control-flag module-path args
Comments MAY be used extending from '#' to the end of the line, and entries MAY be split over multiple lines using a backslash at the end of a line as a continuation character.
The module-type defines the type of action being done. This can be one of four types:
Authenticates users (determines that they are who they claim to be). It can also assign credentials, for example additional group memberships beyond those specified through /etc/passwd and /etc/groups. This additional functionality MUST NOT be used.
Account management not related to authentication, it can also restrict access based on time of day, available system resources or the location of the user (network address or system console).
Manages resources associated with a service by running specified code at the start and end of the session. Typical usage includes logging and accounting, and initialization such as auto mounting a home directory.
Used for updating the password (or other authentication token), for example when using the passwd(1) utility to change it.
The control-flag specifies the action that will be taken based on the success or failure of an individual module. The modules are stacked (executed in sequence), and the control-flags determine which final result (success or failure) will be returned, thereby specifying the relative importance of the modules.
Stacked modules are executed in the order specified in the configuration file.
The control-flag can be specified as either a single keyword, or alternatively with a more elaborate syntax that allows greater control. RHEL uses only the single keyword syntax by default.
The following keywords control how a module affects the result of the authentication attempt:
If this module returns a failure code, the entire stack will return failure. The failure will be reported to the application or user only after all other modules in the stack have been run, to prevent leakage of information (for example, ask for a password even if the entered username is not valid).
Same as required, but return failure immediately not executing the other modules in the stack. Can be used to prevent a user from entering a password over an insecure connection.
Return success immediately if no previous required modules in the stack have returned failure. Do not execute succeeding modules.
The return code of this module is ignored, except if all other modules in the stack return an indeterminate result (PAM_IGNORE).
The module-path specifies the filename of the module to be run (relative to the directory /lib/security/, and the optional args are passed to the module - refer to the module's documentation for supported options.
You MUST restrict authentication to services that are explicitly specified. The 'other' fallback MUST be disabled by specifying the pam_deny.so module for each module-type in the 'other' configuration. This ensures that access decisions within the PAM system are handled only by the service specific PAM configuration.
Note that RHEL uses the pam_stack(8) module to unify commonly used configuration options within single files, rather than having redundant information in multiple files. You MUST verify that the shared settings are applicable to services that use pam_stack, and keep in mind that a change to the shared file will affect several services.
You MUST add the pam_wheel.so module to the 'auth' module_type configuration for the 'su' service to restrict use of su(1) to members of the 'wheel' group.
You MUST add the pam_tally.so module to the auth and account module_type configurations of login, sshd and vsftpd. This ensures that accounts are disabled after several failed login attempts. The pam_tally.so module is used in the auth stack to increment a counter in the file /var/log/faillog, and in the account stack to either deny login after too many failed attempts, or to reset the counter to zero after successful authentication. The evaluated configuration uses a lockout after five failed attempts, corresponding to the setting deny=5, you MAY decrease the number for stricter enforcement. Be aware that this can be used in denial-of-service attacks to lock out legitimate users. Please refer to section §4.7 of this guide for more information.
You MUST use the pam_passwdqc.so password quality checking module to ensure that users will not use easily-guessable passwords.
You MUST use the pam_loginuid.so module for all authentication paths where human users are identified and authenticated, and add the require_auditd option for all cases where the authentication method is accessible to non- administrative users. This module sets the persistent login user ID and prevents login in case the audit system is inoperable for fail-secure operation.
When using the LSPP/RBAC mode, you MUST use the pam_selinux.so and pam_namespace.so modules as specified in the configuration files below. Please refer to section §4.16.2 for more information about the pam_namespace.so module's functionality.
The system supports many other PAM modules apart from the ones shown here. In general, you MAY add PAM modules that add additional restrictions. You MUST NOT weaken the restrictions through configuration changes of the modules shown here or via additional modules. Also, you MUST NOT add PAM modules that provide additional privileges to users (such as the pam_console.so module).
You MUST NOT run the authconfig(8) tool to modify the authentication configuration.
Following are the pam configuration files:
This file contains common settings that are shared by multiple services using authentication. The pam_passwdqc.so module is configured to enforce the minimum password length of 8 characters. Note that the pam_passwdqc.so module is not part of a default installation, it was added previously as described in section §4.13 of this guide.
The pam_tally module MUST be used to block the user after 5 failed login attempts.
The remember option to pam_unix.so prevents users from reusing old passwords. Hashes of old passwords are stored in the file /etc/security/opasswd with the exception of passwords changed by the "root" user. Note that this file MUST exist, otherwise users cannot change passwords. Use the following commands to create it:
touch /etc/security/opasswd
chmod 600 /etc/security/opasswd
The file /etc/pam.d/system-auth MUST be set up with the following content:
#%PAM-1.0 # # pam.d/system-auth - PAM master configuration for CAPP/LSPP compliance # see the Evaluated Configuration Guide for more info #
auth required pam_env.so auth required pam_unix.so nullok try_first_pass
account required pam_unix.so
password required pam_passwdqc.so min=disabled,disabled,16,12,8 \
random=42
password required pam_unix.so nullok use_authtok md5 \
shadow remember=7
session required pam_limits.so session required pam_unix.so
This file configures the behavior of the login program. It allows root login only for terminals configured in /etc/securetty. If the file /etc/nologin is present, then only root can log in.
The pam_loginuid.so module is by default configured without the require_auditd option, which assumes that all terminals available for login are in physically secure locations and accessible only for authorized administrators. This permits administrators to log in on the console even if the audit subsystem is not available. If any serial terminals are attached and available for arbitrary users, you MUST add the require_auditd option to ensure the CAPP/LSPP-compliant fail-secure operating mode that disables login if audit is not working. Please refer to section §4.9 of this guide for more information.
This is the configuration for the CAPP mode:
#%PAM-1.0 # # pam.d/login - PAM login configuration for CAPP compliance # see the Evaluated Configuration Guide for more info # # If serial terminals are in use, pam_loginuid.so MUST use the 'require_auditd' # option for LSPP-complaint fail-secure auditing. The default mode assumes that # all terminals are in physically secure locations. #
auth required pam_securetty.so auth include system-auth auth required pam_tally2.so deny=5 onerr=fail
account required pam_nologin.so account include system-auth account required pam_tally2.so
password include system-auth
# pam_selinux.so close should be the first session rule session required pam_selinux.so close session include system-auth session required pam_loginuid.so session optional pam_console.so # pam_selinux.so open should only be followed by sessions to be # executed in the user context session required pam_selinux.so open
This is the configuration for the RBAC/LSPP mode, adding the "select_context" option and the pam_namespace module:
#%PAM-1.0 # # pam.d/login - PAM login configuration for LSPP compliance # see the Evaluated Configuration Guide for more info # # If serial terminals are in use, pam_loginuid.so MUST use the 'require_auditd' # option for LSPP-complaint fail-secure auditing. The default mode assumes that # all terminals are in physically secure locations. #
auth required pam_securetty.so auth include system-auth auth required pam_tally2.so deny=5 onerr=fail
account required pam_nologin.so account include system-auth account required pam_tally2.so
password include system-auth
# pam_selinux.so close should be the first session rule session required pam_selinux.so close session include system-auth session required pam_loginuid.so session optional pam_console.so # pam_selinux.so open should only be followed by sessions to be # executed in the user context session required pam_selinux.so open select_context session required pam_namespace.so
This configuration applies for all PAM usage for which no explicit service is configured. It will log and block any attempts.
#%PAM-1.0 # # pam.d/other - PAM other configuration for CAPP/LSPP compliance # see the Evaluated Configuration Guide for more info #
auth required pam_warn.so auth required pam_deny.so
account required pam_warn.so account required pam_deny.so
password required pam_warn.so password required pam_deny.so
session required pam_warn.so session required pam_deny.so
This file configures the PAM usage for SSH. This is similar to the login configuration. The securetty entry is not applicable to network logins, and the pam_loginuid.so module MUST be configured to prevent network login if the audit system is not available. Note that pam_loginuid.so MUST run in the account stack, it does not work in the account or auth stacks due to the OpenSSH privilege separation mechanism.
This is the configuration for the CAPP mode:
#%PAM-1.0 # # pam.d/sshd - pam.d/sshd configuration for CAPP compliance # see the Evaluated Configuration Guide for more info #
auth include system-auth auth required pam_tally2.so deny=5 onerr=fail
account required pam_nologin.so account include system-auth account required pam_tally2.so
password include system-auth
session required pam_selinux.so close session include system-auth session required pam_loginuid.so require_auditd
This is the configuration for the RBAC/LSPP mode, adding the pam_namespace module:
#%PAM-1.0 # # pam.d/sshd - pam.d/sshd configuration for LSPP compliance # see the Evaluated Configuration Guide for more info #
auth include system-auth auth required pam_tally2.so deny=5 onerr=fail
account required pam_nologin.so account include system-auth account required pam_tally2.so
password include system-auth
session required pam_selinux.so close session include system-auth session required pam_loginuid.so require_auditd session required pam_namespace.so
This file configures the behavior of the 'su' command. Only users in the trusted 'wheel' group can use it to become 'root', as configured with the pam_wheel module.
#%PAM-1.0 # # pam.d/su - PAM su configuration for CAPP/LSPP compliance # see the Evaluated Configuration Guide for more info #
auth sufficient pam_rootok.so auth required pam_wheel.so use_uid auth include system-auth
account include system-auth
password required pam_deny.so
session include system-auth session optional pam_xauth.so
The password branch is disabled because forcing the root user to change the root password is not desired for this program,
This file configures the authentication for the FTP daemon. With the listfile module, users listed in /etc/vsftpd.ftpusers are denied FTP access to the system. Note that the setting is relevant only for authentication of incoming connections, and does not prevent local users from using the ftp(1) client to access other machines on the network.
#%PAM-1.0 # # pam.d/vsftpd - vsftpd configuration for CAPP/LSPP compliance # see the Evaluated Configuration Guide for more info
auth required pam_listfile.so item=user sense=deny \
file=/etc/vsftpd/ftpusers onerr=succeed
auth required pam_shells.so
auth include system-auth
auth required pam_tally2.so deny=5 onerr=fail
account include system-auth account required pam_tally2.so
password required pam_deny.so
session include system-auth session required pam_loginuid.so require_auditd
pam_deny.so is used in the password stack because the FTP protocol has no provisions for changing passwords.
The file /etc/login.defs defines settings that will be used by user management tools such as useradd(8). The file is not used during the authentication process itself.
The password aging settings defined in this file are used when creating users and when changing passwords, and stored in the user's /etc/shadow entry. Note that only the /etc/shadow entries are considered during authentication, so changes in /etc/login.defs will not retroactively change the settings for existing users.
The PASS_MIN_LEN setting has no effect in the evaluated configuration, the relevant settings are instead configured using the min= parameter to pam_passwdqc.so in the /etc/pam.d/system-auth file.
### /etc/login.defs # Global user account settings for the Common Criteria CAPP/LSPP configuration. # # *REQUIRED* # Directory where mailboxes reside, _or_ name of file, relative to the # home directory. If you _do_ define both, MAIL_DIR takes precedence. # QMAIL_DIR is for Qmail # # The setting is used only when creating or deleting users, and has # no effect on the mail delivery system. MAY be changed as required. # #QMAIL_DIR Maildir MAIL_DIR /var/spool/mail #MAIL_FILE .mail # # Password aging controls: # # PASS_MAX_DAYS Maximum number of days a password may be used. # PASS_MIN_DAYS Minimum number of days allowed between password changes. # PASS_MIN_LEN Minimum acceptable password length. # PASS_WARN_AGE Number of days warning given before a password expires. # PASS_MAX_DAYS 60 # MAY be changed, must be <= 60 PASS_MIN_DAYS 1 # MAY be changed, 0 < PASS_MIN_DAYS < PASS_MAX_DAYS PASS_MIN_LEN 5 # no effect in the evaluated configuration PASS_WARN_AGE 7 # MAY be changed # # Min/max values for automatic uid selection in useradd # # MAY be changed, 100 < UID_MIN < UID_MAX < 65535 # UID_MIN 500 UID_MAX 60000 # # Min/max values for automatic gid selection in groupadd # # MAY be changed, 100 < GID_MIN < GID_MAX < 65535 # GID_MIN 500 GID_MAX 60000 # # If defined, this command is run when removing a user. # It should remove any at/cron/print jobs etc. owned by # the user to be removed (passed as the first argument). # # MAY be activated as described in the "Managing user accounts" # section of the ECG. # #USERDEL_CMD /usr/sbin/userdel_local # # If useradd should create home directories for users by default # On RH systems, we do. This option is overridden with the -m flag on # useradd command line. # # MAY be changed. # CREATE_HOME yes # # The permission mask is initialized to this value. If not specified, # the permission mask will be initialized to 022. # # MAY be changed. # UMASK 077
You MUST set up the server in a secure location where it is protected from unauthorized access. Even though that is sufficient to protect the boot process, it is RECOMMENDED to configure the following additional protection mechanisms:
Ensure that the installed system boots exclusively from the disk partition containing RHEL, and not from floppy disks, USB drives, CD-ROMs, network adapters, or other devices.
Ensure that this setting cannot be modified, for example by using a BootProm/BIOS password to protect access to the configuration.
The GRUB boot loader is used on the x86 and Opteron platforms. It is highly configurable, and permits flexible modifications at boot time through a special-purpose command line interface. Please refer to the grub(8) man page or run info grub for more information.
Use the password command in /boot/grub/menu.lst to prevent unauthorized use of the boot loader interface. Using md5 encoded passwords is RECOMMENDED, run the command grub-md5-crypt to generate the encoded version of a password.
Protect all menu entries other than the default RHEL boot with the lock option, so that the boot loader will prompt for a password when the user attempts to boot from other media (such as a floppy) or sets other non-default options for the boot process. To implement this, add a line containing just the keyword lock after the title entry in the /boot/grub/menu.lst file.
Remove group and world read permissions from the grub configuration file if it contains a password by running the following command:
chmod 600 /boot/grub/menu.lst
All changes to the configuration take effect automatically on the next boot, there is no need to re-run an activation program.
The following example of the /boot/grub/menu.lst configuration file shows RECOMMENDED settings:
default=0
timeout=10
splashimage=(hd0,0)/boot/grub/splash.xpm.gz
password --md5 $1$O471l/$H/JW2MYeugX6Y1h3v.1Iz0
title Red Hat Enterprise Linux Server (2.6.18-8.1.3.lspp.81.el5)
root (hd0,0)
kernel /vmlinuz-2.6.18-8.1.3.lspp.81.el5 ro root=/dev/VolGroup00/LvRoot
initrd /initrd-2.6.18-8.1.3.lspp.81.el5.img
Note that the configuration shown here might not be exactly the configuration used on the installed system, depending on the kernel options needed for the hardware.
This section applies to the ia64 (Itanium) platform only.
On this platform, the filesystem /boot/efi/ is of type vfat with no built-in access control support. It MUST be mounted with the option umask=077 to ensure that only administrators are able to access the data.
Installing kernel RPM packages will automatically configure the necessary data in /boot/efi/ so that the builtin EFI boot loader will start the new kernel.
The file /boot/efi/efi/redhat/elilo.conf configures the boot options, including which kernel is booted and what kernel command line options are set. Please refer to the files /usr/share/doc/elilo-*/* for more information.
- This concludes the sections covered by the automated configuration script -
After all the changes described in this chapter have been done, you MUST reboot the system to ensure that all unwanted tasks are stopped, and that the running kernel, modules and applications all correspond to the evaluated configuration.
Please make sure that the boot loader is configured correctly for your platform.
Remember to remove any CD-ROM from the drive and/or configure the system to boot from hard disk only.
The system will then match the evaluated configuration. The server MAY then be connected to a secure network as described above.
To ensure that the systems remains in a secure state, special care MUST be taken during system operation.
Use the shutdown(8), halt(8) or reboot(8) programs as needed to shut down or reboot the system.
When powered on (or on initial program load of the logical partition on a host system), the system will boot into the RHEL operating system. If necessary (for example after a crash), a filesystem check will be performed automatically. In rare cases manual intervention is necessary, please refer to the e2fsck(8) and debugfs(8) documentation for details in this case.
In case a nonstandard boot process is needed (such as booting from floppy disk or CD-ROM to replace a defective hard drive), interaction with the boot loader and/or the host's management system can be used to modify the boot procedure for recovery.
For example, on systems using the grub boot loader you can use the following commands to launch a shell directly from the kernel, bypassing the normal init/login mechanism:
# view the current grub configuration
grub> cat (hd0,1)/boot/grub/menu.lst
# manually enter the modified settings
grub> kernel (hd0,1)/boot/vmlinuz root=/dev/sda1 init=/bin/sh
grub> initrd (hd0,1)/boot/initrd
grub> boot
Please refer to the relevant documentation of the boot loader, as well as the RHEL administrator guide, for more information.
Whenever you make changes to security-critical files, you MAY need to be able to track the changes made and revert to previous versions, but this is not required for compliance with the evaluated configuration.
The star(1) archiver is RECOMMENDED for backups of complete directory contents, please refer to section §6.6 of this guide. Regular backups of the following files and directories (on removable media such as tapes or CD-R, or on a separate host) are RECOMMENDED:
/etc/
/var/spool/cron/
Depending on your site's audit requirements, also include the contents of /var/log/ in the backup plan. In that case, the automatic daily log file rotation needs to be disabled or synchronized with the backup mechanism, refer to sections §5.2 and §5.3 of this guide for more information.
When SELinux is activated, the same actions as in the case of ACLs must be applied since the security related information is stored in extended attributes. Refer to section §6.6 for more information. This means that star MUST be used.
You MUST protect the backup media from unauthorized access, because the copied data does not have the access control mechanisms of the original file system. Among other critical data, it contains the secret keys used by the SSH and stunnel servers, as well as the /etc/shadow password database. Store the backup media at least as securely as the server itself.
A RECOMMENDED method to track changes is to use a version control system. RCS is easy to set up because it does not require setting up a central repository for the changes, and you can use shell scripting to automate the change tracking. RCS is not included in the evaluated configuration, see rcsintro(1) in the rcs RPM package for more information. Alternatively, you can create manually create backup copies of the files and/or copy them to other servers using scp(1).
System administration tasks require superuser privileges. Since directly logging on over the network as user 'root' is disabled, you MUST first authenticate using an unprivileged user ID, and then use the su command to switch identities. Note that you MUST NOT use the 'root' rights for anything other than those administrative tasks that require these privileges, all other tasks MUST be done using your normal (non-root) user ID.
You MUST use exactly the following su(1) command line to gain superuser access:
/bin/su -
This ensures that the correct binary is executed irrespective of PATH settings or shell aliases, and that the root shell starts with a clean environment not contaminated with the starting user's settings. This is necessary because the .profile shell configuration and other similar files are writable for the unprivileged ID, which would allow an attacker to easily elevate privileges to root if able to subvert these settings.
Administrators MUST NOT add any directory to the root user's PATH that are writable for anyone other than 'root', and similarly MUST NOT use or execute any scripts, binaries or configuration files that are writable for anyone other than 'root', or where any containing directory is writable for a user other than 'root'.
In the LSPP/RBAC mode, note that newrole(1) is available if the SELinux context needs to be changed to another one. As usual, this mechanism is completely decoupled from the DAC mechanisms. Please also refer to Section §4.16 for more information.
Additional software packages MAY be installed as needed, provided that they do not conflict with the security requirements.
Any additional software added is not intended to be used with superuser privileges. The administrator MUST use only those programs that are part of the original evaluated configuration for administration tasks, except if the administrator has independently ensured that use of the additional software is not a security risk.
Administrators MAY add scripts to automate tasks as long as those only depend on and run programs that are part of the evaluated configuration.
The security requirements for additional software are:
Kernel modules other than those provided as part of the evaluated configuration MUST NOT be installed or loaded. You MUST NOT load the tux kernel module (the in-kernel web server is not supported). You MUST NOT add support for non-ELF binary formats or foreign binary format emulation that circumvents system call auditing. You MUST NOT activate knfsd or export NFS file systems.
Device special nodes MUST NOT be added to the system.
SUID root or SGID root programs MUST NOT be added to the system. Programs which use the SUID or SGID bits to run with identities other than 'root' MAY be added if the numerical SUID and SGID values are not less than 500. This restriction is necessary to avoid conflict with the system user and group IDs such as the "bin" group.
The content, permissions, and ownership of all existing filesystem objects (including directories and device nodes) that are part of the evaluated configuration MUST NOT be modified. Files and directories MAY be added to existing directories provided that this does not violate any other requirement.
In the RBAC/LSPP mode, the administrator MUST NOT add new programs with automatic type transitioning capability, or add types providing such capability to existing programs. By convention this capability is indicated with an SELinux type name ending in _exec_t (shown using ls -lZ PROGRAMNAME , but the policy language would support type transitions for any type).
Programs automatically launched with 'root' privileges MUST NOT be added to the system. Exception: processes that immediately and permanently switch to a non privileged identity on launch are permitted, for example by using su USERID -c LAUNCH_COMMAND in the startup file, or alternatively by using the setgroups(2), setgid(2) and setuid(2) system calls in a binary. (seteuid(2) etc. are insufficient.)
Automatic launch mechanisms are:
Entries in /etc/inittab
Executable files or links in /etc/rc.d/init.d/ and its subdirectories
Entries in /etc/xinetd.conf
Scheduled jobs using cron (including entries in /etc/cron* files)
Examples of programs that usually do not conflict with these requirements and MAY be installed are compilers, interpreters, network services running with non-root rights, and similar programs. The requirements listed above MUST be verified in each specific case.
The cron(8) program schedules programs for execution at regular intervals. Entries can be modified using the crontab(1) program - the file format is documented in the crontab(5) manual page.
You MUST follow the rules specified for installation of additional programs for all entries that will be executed by the 'root' user. Use non-root crontab entries in all cases where 'root' privileges are not absolutely necessary.
Errors in the non interactive jobs executed by cron are reported in the system log files in /var/log/, and additionally via e-mail to the user who scheduled it.
Permission for users to schedule jobs with cron is controlled through the following allow and deny files:
/etc/cron.allow
/etc/cron.deny
The allow file has precedence if it exists, then only those users whose usernames are listed in it are permitted to use the service. If it does not exist, the deny file is used instead and all users who are not listed in that file can use the service. Note that the contents of these files are only relevant when the scheduling commands are executed, and changes have no effect on already scheduled commands.
The root user is always permitted to use the crontab(1) program on RHEL systems, even if listed in the /etc/cron.deny file.
In the RHEL distribution, the allow files do not exist, and deny files are used to prevent system-internal IDs and/or guest users from using these services. By default, the evaluated configuration permits everybody to use cron.
It is RECOMMENDED to restrict the use of cron to human users and disallow system accounts from using these mechanisms. For example, the following commands add all system accounts other than root to the deny files:
awk -F: '{if ($3>0 && $3<500) print $1}' /etc/passwd >/etc/cron.deny
chmod 600 /etc/cron.deny
Administrators MAY schedule jobs that will be run with the privileges of a specified user by editing the file /etc/crontab with an appropriate username in the sixth field. Entries in /etc/crontab are not restricted by the contents of the allow and deny files.
You MAY create a /etc/cron.allow file to explicitly list users who are permitted to use this service. If you do create the file, it MUST be owned by the user 'root' and have file permissions 0600 (no access for group or others).
If any filesystems need to be mounted in addition to those set up at installation time, appropriate mount options MUST be used to ensure that mounting the filesystem does not introduce capabilities that could violate the security policy.
The special-purpose proc, sysfs, devpts, selinuxfs, binfmt_misc, and tmpfs filesystems are part of the evaluated configuration. These are virtual filesystems with no underlying physical storage, and represent data structures in kernel memory. Access to contents in these special filesystems is protected by the normal discretionary access control policy and additional permission checks.
Note that changing ownership or permissions of virtual files and directories is generally NOT supported for the proc and sysfs filesystems (corresponding to directories /proc/ and /sys/), and attempts to do so will be ignored or result in error messages.
Note that use of the usbfs filesystem type is NOT permitted (and not needed) in the evaluated configuration.
A new file system can be integrated as part of the evaluated configuration, for example by installing an additional hard disk, under the following conditions:
The device is protected against theft or manipulation in the same way as the server itself, for example by being installed inside the server.
One or more new, empty, file systems in EXT3 format are created on it.
The file systems are mounted using the acl option, for example with the following setting in the /etc/fstab file:
/dev/sdc1 /home2 ext3 acl 1 2
Existing files and directories MAY then be moved onto the new file systems.
If a device containing a file system is ever removed from the system, the device MUST be stored within the secure server facility, or alternatively MUST be destroyed in a way that the data on it is reliably erased.
Alternatively, media MAY be accessed without integrating them into the evaluated configuration, for example DVDs or CD-ROMs.
CD/DVD devices MUST be accessed using the iso9660 filesystem type. Using the udf filesystem or using an automounter is NOT permitted in the evaluated configuration.
The following mount options MUST be used if the filesystems contain data that is not part of the evaluated configuration:
nodev,nosuid
Adding the noexec mount option to avoid accidental execution of files or scripts on additional mounted filesystems is RECOMMENDED.
You MAY use the context option to specify a SELinux context, including the MLS level and categories, for mounting filesystems that do not have associated label information. The specified context will then apply to all the data on that filesystem.
Be aware that data written to removable media is not reliably protected by the DAC permission mechanism, and should be considered accessible to anyone with physical access to the media. It is RECOMMENDED to add the ro option to mount the file system read-only.
Note that these settings do not completely protect against malicious code and data, you MUST also verify that the data originates from a trustworthy source and does not compromise the server's security. Specifically, be aware of the following issues:
Even unprivileged programs and scripts can contain malicious code that uses the calling user's rights in unintended ways, such as corrupting the user's data, introducing trojan horses in the system, attacking other machines on the network, revealing confidential documents, or sending unsolicited commercial e-mail ("spam").
Data on the additional filesystem MUST have appropriate access rights to prevent disclosure to or modification by unauthorized users. Be aware that imported data may have been created using user names and permissions that do not match your system's security policies.
You MUST NOT write data on removable file systems such as floppy disks, since it cannot be adequately protected by the system's access control mechanisms after being removed from the system. Please refer to section §4.6 of this guide for more information regarding non-filesystem-based backup.
Each new file system MUST be mounted on an empty directory that is not used for any other purpose. It is RECOMMENDED using subdirectories of /mnt for temporary disk and removeable storage media mounts.
For example:
# mount /dev/cdrom /mnt/cdrom -t iso9660 -o ro,nodev,nosuid,noexec
You MAY also add an equivalent configuration to /etc/fstab, for example:
/dev/cdrom /mnt/cdrom iso9660 ro,noauto,nodev,nosuid,noexec 0 0
You MUST NOT include the user flag, ordinary users are not permitted to mount filesystems. This is also enforced by the deletion of the SUID bit on the mount command.
Use the useradd(8) command to create new user accounts, then use the passwd(1) command to assign an initial password for the user. Alteratively, if the user is present when the account is created, permit them to choose their own password. Refer to the manual pages for useradd(8) and passwd(1) for more information.
If you assign an initial password for a new user, you MUST transfer this initial password in a secure way to the user, ensuring that no third party gets the information. For example, you can tell the password to a user personally known to you. If this is not possible, you MAY send the password in written form in a sealed letter. This applies also when you set a new password for a user in case the user has forgotten the password or it has expired. You MUST advise the user that he MUST change this initial password when he first logs into the system and select his own password in accordance with the rules defined in section §6.3 of this guide.
You MUST NOT use the -p option to useradd(8), specifying a password in that way would bypass the password quality checking mechanism.
The temporary password set by the administrator MUST be changed by the user as soon as possible. Use the chage(8) command with the -d option to set the last password change date to a value where the user will be reminded to change the password. The RECOMMENDED value is based on the settings in /etc/login.defs and is equivalent to today's date plus PASS_WARN_AGE minus PASS_MAX_DAYS.
Example:
useradd -m -c "John Doe" jdoe
passwd jdoe
chage -d $(date +%F -d "53 days ago") jdoe
The -m option to useradd(8) creates a home directory for the user based on a copy of the contents of the /etc/skel/ directory. Note that you MAY modify some default configuration settings for users, such as the default umask(2) setting or time zone, by editing the corresponding global configuration files:
/etc/profile
/etc/bashrc
/etc/csh.cshrc
If necessary, you MAY reset the user's password to a known value using passwd USER, and entering the new password. You cannot recover the previously used password, since the hash function used is not reversible.
You MAY use the usermod(8) command to change a user's properties. For example, if you want to add the user 'jdoe' to the wheel group, you could use the following:
# List the groups the user is currently a member of:
groups jdoe
# Add the additional group
usermod -G $(su jdoe -c groups | sed 's/ /,/g'),wheel jdoe
Users MAY be locked out (disabled) using passwd -l USER, and re-enabled using passwd -u USER.
The pam_tally.so PAM module enforces automatic lockout after excessive failed authentication attempts, as described in section §4.16.2 of this guide. Use the program pam_tally to view and reset the counter if necessary, as documented in the file /usr/share/doc/pam-*/txts/README.pam_tally. Note that the pam_tally mechanism does not prevent password guessing attacks, it only prevents use of the account after such an attack has been detected. Therefore, you MUST assign a new password for the user before reactivating an account. For example:
# view the current counter value
pam_tally --user jdoe
# set new password, and reset the counter
passwd jdoe
pam_tally --user jdoe --reset
The chage(1) utility MAY be used to view and modify the expiry settings for user accounts. Unprivileged users are able to view but not modify their own expiry settings.
The userdel(8) utility removes the user account from the system, but does not remove files outside the home directory (and the mail spool file), or kill processes belonging to this user. Use kill (or reboot the system) and find to do so manually if necessary, for example:
# Which user to delete?
U=jdoe
# Lock user account, but don't remove it yet
passwd -l $U
# Kill all user processes, repeat if needed (or reboot)
kill -9 `ps -la --User $U|awk '{print $4}'`
# Recursively remove all files and directories belonging to user
# (Careful - this may delete files belonging to others if they
# are stored in a directory owned by this user.)
find / -depth \( ! -fstype ext3 -prune -false \) \
-o -user $U -exec rm -rf {} \;
# Remove cron jobs
crontab -u $U -r
# Now delete the account
userdel $U
If you need to create additional groups or modify existing groups, use the groupadd(8), groupmod(8) and groupdel(8) commands.
Group passwords are NOT supported in the evaluated configuration, and have been disabled by removing the SUID bit from the newgrp(8) program. You MUST NOT re-enable this feature and MUST NOT use passwd(1) with the -g switch or the gpasswd(1) command to set group passwords.
In the LSPP/RBAC mode, by default, new users are automatically mapped to the standard SELinux user user_u which has the least powerful role user_r. Please refer to Section §4.15.4 on how to change this.
Please also refer to Section §4.16 for more information.
Associating network connections with labels is done with the aid of netlabelctl(8). By default, no label transfer is performed and the system is configured to accept any unlabeled communication coming from outside.
If labeling is employed in a network, a domain of interpretation (DOI) has to be agreed upon by all machines. It is specified by a numerical value. All machines within this DOI SHOULD interpret security levels in a similar manner. If one specific operation is forbidden on one machine for a certain securty level, it SHOULD also be forbidden for the corresponding security level on any other machine in the network. Note that the numberical values of the security levels need not be identical on all machines as the netlabel mechanism allows to install an appropriate mapping between them. For following command is used to set up a mapping which exchanges the meaning of local and remote categories 0 and 1 and provides a one-to-one mapping between the security levels 0 and 1 for the DOI 8
netlabelctl cipsov4 add std doi:8 tags:1 levels:0=0,1=1 categories:0=1,1=0
tags:1 specifies that CIPSO tag type number 1 is to be used. At present, this is the only choice supported by netlabel.
Use the following to make the DOI 8 be the default choice for CIPSPO on IPv4:
netlabelctl add map default protocol:cipsov4,8
If acceptance of unlabeled packets needs to be forbidden, the following command must be issued:
netlabelctl unlbl accept off
When no translation between security levels and categories is supposed to take place for a given DOI, a pass-through mapping can be established:
netlabelctl cipsov4 add pass doi:8 tags:1 pass
Refer to the netlabelctl(8) man page for a reference of further available options.
You MAY attach serial terminals to the system. They are activated by adding an entry in the file /etc/inittab for each serial terminal that causes init(8) to launch an agetty(8) process to monitor the serial line. agetty runs login(1) to handle user authentication and set up the user's session.
If you use serial terminals and require the CAPP/LSPP-compliant fail-safe audit mode, you MUST ensure that the file /etc/pam.d/login uses the option require_auditd for the pam_loginuid.so module in the session stack. Please refer to section §4.9 of this guide for more information about the needed PAM configuration.
For example, adding the following line to /etc/inittab activates a VT102-compatible serial terminal on serial port /dev/ttyS1, communicating at 19200 bits/s:
S1:3:respawn:/sbin/agetty 19200 ttyS1 vt102
The first field MUST be an unique identifier for the entry (typically the last characters of the device name). Please refer to the agetty(8) and inittab(5) man pages for further information about the format of entries.
You MUST reinitialize the init daemon after any changes to /etc/inittab by running the following command:
init q
The system supports SYSV-compatible shared memory, IPC objects, and message queues. If programs fail to release resources they have used (for example, due to a crash), the administrator MAY use the ipcs(8) utility to list information about them, and ipcrm(8) to force deletion of unneeded objects. Note that these resources are also released when the system is rebooted.
For additional information, please refer to the msgctl(2), msgget(2), msgrcv(2), msgsnd(2), semctl(2), semget(2), semop(2), shmat(2), shmctl(2), shmdt(2), shmget(2) and ftok(3) manual pages.
The system supports POSIX message queues for interprocess communication. Please refer to the mq_overview(7) man page for more information, including documentation of the supported system calls.
The optional mqueue filesystem is NOT supported in the evaluated configuration.
Message queue objects persist until deleted using the appropriate system calls such as mq_unlink(2) explicitly removes them, or until the system is rebooted.
The stunnel program is a flexible and secure solution for setting up encrypted network connections, enabling the use of strong encryption even for applications that are not able to use encryption natively. stunnel uses the OpenSSL library for its encryption functions, and the corresponding openssl(1) command line tool for key management.
Stunnel has three main operating modes:
Accept incoming SSL-encrypted TCP connections, and run a specific program to handle the request.
This is similar to how xinetd launches programs, and any program compatible with xinetd can also be used for this purpose. It must read and write the communication data on the stdin and stdout file descriptors and stay in the foreground. stunnel also supports switching user and group IDs before launching the program.
Open a SSL connection to a remote SSL-capable TCP server, and copy data to and from stdin and stdout.
Bind a TCP port to accept incoming unencrypted connections, and forward data using SSL to a prespecified remote server.
The following diagram shows a sample usage scenario:
+-----------+ +-----------+
| | | |
| | encrypted data stream | |
| stunnel=+==================================+=stunnel |
| |\ //|| |
|-----------| \ local / |-----------|
|1024 | \ local plain- / | 1024|
| | \ plain- text \ | |
| | / text data \ | |
| | / data stream \ | |
| ||/ stream \_=Client |
| Server=+*- | |
| | | |
| | | |
+-----------+ +-----------+
TCP ports TCP ports
In this scenario, neither the client nor the server have administrator privileges, they are running as normal user processes. Also, the client and server do not support encryption directly.
stunnel makes a secure communication channel available for the client and server. On the client, stunnel is accepting connections on TCP port 82. The client connects to this port on the local machine using normal unencrypted TCP, stunnel accepts the connection, and opens a new TCP connection to the stunnel server running on the remote machine. The stunnel instances use cryptographic certificates to ensure that the data stream has not been intercepted or tampered with, and then the remote stunnel opens a third TCP connection to the server, which is again a local unencrypted connection.
Any data sent by either the client or server is accepted by the corresponding stunnel instance, encrypted, sent to the other stunnel, decrypted and finally forwarded to the receiving program. This way, no modifications are required to the client and server.
To set up a secure connection compliant with the evaluated configuration, you MUST start the stunnel server(s) with administrator rights, and you MUST use a TCP port in the administrator-reserved range 1-1023 to accept incoming connections. A corresponding client which connects to the server MAY be started by any user, not just administrators.
stunnel MAY also be used by non-administratorive users to receive encrypted connections on ports in the range 1024-65536. This is permitted, but it is outside of the scope of the evaluated configuration and not considered to be a trusted connection.
Any network servers and clients other than the trusted programs described in this guide (stunnel, sshd, vsftpd, postfix and cupsd) MUST be run using non-administrator normal user identities. Programs run from stunnel MUST be switched to a non-root user ID by using the setuid and setgid parameters in the /etc/stunnel/*.conf configuration files.
It is RECOMMENDED configuring any such servers to accept connections only from machine-local clients, either by binding only the localhost IP address 127.0.0.1, or by software filtering inside the application. This ensures that the only encrypted connections are possible over the network. Details on how to do this depend on the software being used and are beyond the scope of this guide.
Please refer to the stunnel(8) and openssl(1) man pages for more information.
It is strongly RECOMMENDED that you have your server's certificate signed by an established Certificate Authority (CA), which acts as a trusted third party to vouch for the certificate's authenticity for clients. Please refer to the openssl(1) and req(1) man pages for instructions on how to generate and use a certificate signing request.
Create the server's private key and a certificate signing request (CSR) with the following commands:
touch /etc/stunnel/stunnel.pem
chmod 400 /etc/stunnel/stunnel.pem
openssl req -newkey rsa:1024 -nodes \
-keyout /etc/stunnel/stunnel.pem -out /etc/stunnel/stunnel.csr
You will be prompted for the information that will be contained in the certificate. Most important is the "Common Name", because the connecting clients will check if the hostname in the certificate matches the server they were trying to connect to. If they do not match, the connection will be refused, to prevent a 'man-in-the-middle' attack.
Here is a sample interaction:
Generating a 1024 bit RSA private key
..........++++++
.......++++++
writing new private key to '/etc/stunnel/stunnel.pem'
-----
You are about to be asked to enter information that will be incorporated
into your certificate request.
What you are about to enter is what is called a Distinguished Name or a DN.
There are quite a few fields but you can leave some blank
For some fields there will be a default value,
If you enter '.', the field will be left blank.
-----
Country Name (2 letter code) [PL]:US
State or Province Name (full name) [Some-State]:TX
Locality Name (eg, city) []:Austin
Organization Name (eg, company) [Stunnel Developers Ltd]:Example Inc.
Organizational Unit Name (eg, section) []:
Common Name (FQDN of your server) []:www.example.com
Common Name (default) []:localhost
The file /etc/stunnel/stunnel.pem will contain both the certificate (public key) and also the secret key needed by the server. The secret key will be used by non-interactive server processes, and cannot be protected with a passphrase. You MUST protect the secret key from being read by unauthorized users, to ensure that you are protected against someone impersonating your server.
Next, send the generated CSR file /etc/stunnel/stunnel.csr (not the private key) to the CA along with whatever authenticating information they require to verify your identity and your server's identity. The CA will then generate a signed certificate from the CSR, using a process analogous to openssl req -x509 -in stunnel.csr -key CA-key.pem -out stunnel.cert.
When you receive the signed certificate back from the CA, append it to the file /etc/stunnel/stunnel.pem containing the private key using the following command:
echo >> /etc/stunnel/stunnel.pem
cat stunnel.cert >> /etc/stunnel/stunnel.pem
Make sure that the resulting file contains no extra whitespace or other text in addition to the key and certificate, with one blank line separating the private key and certificate:
-----BEGIN RSA PRIVATE KEY-----
MIICXQIBAAKBgQCzF3ezbZFLjgv1YHNXnBnI8jmeQ5MmkvdNw9XkLnA2ONKQmvPQ
[...]
4tjzwTFxPKYvAW3DnXxRAkAvaf1mbc+GTMoAiepXPVfqSpW2Qy5r/wa04d9phD5T
oUNbDU+ezu0Pana7mmmvg3Mi+BuqwlQ/iU+G/qrG6VGj
-----END RSA PRIVATE KEY-----
-----BEGIN CERTIFICATE-----
MIIC1jCCAj+gAwIBAgIBADANBgkqhkiG9w0BAQQFADBXMQswCQYDVQQGEwJQTDET
[...]
bIbYKL6Q1kE/vhGmRXcXQrZzkfu8sgJv7JsDpoTpAdUnmvssUY0bchqFo4Hhzkvs
U/whL2/8RFv5jw==
-----END CERTIFICATE-----
You MAY distribute the original signed certificate (stunnel.cert in this example) to clients, it does not contain any confidential information. Never distribute the file containing the private key, that is for use by the stunnel server only.
When using externally signed certificates, you MUST use the option CApath in stunnel client configuration files along with the setting verify=2 or verify=3 to enable the clients to verify the certificate.
Alternatively, you MAY use a self-signed certificate instead of one signed by an external CA. This saves some time and effort when first setting up the server, but each connecting client MUST manually verify the certificate's validity. Experience shows that most users will not do the required checking and simply click "OK" for whatever warning dialogs that are shown, resulting in significantly reduced security. Self-signed certificates can be appropriate for controlled environments with a small number of users, but are not recommended for general production use.
Create a self-signed host certificate with the following commands:
# create secret key and self-signed certificate
openssl req -newkey rsa:1024 -nodes \
-keyout /etc/stunnel/stunnel.pem \
-new -x509 -sha1 -days 365 \
-out /etc/stunnel/stunnel.cert
# set appropriate file permissions
chmod 400 /etc/stunnel/*.pem
chmod 444 /etc/stunnel/*.cert
# append copy of certificate to private key
echo >> /etc/stunnel/stunnel.pem
cat /etc/stunnel/stunnel.cert >> /etc/stunnel/stunnel.pem
The secret key contained in the /etc/stunnel/stunnel.pem file MUST be kept secret.
You MAY distribute the public certificate contained in the /etc/stunnel/stunnel.cert file to clients. Make sure you do not accidentally distribute the secret key instead.
The client has no independent way to verify the validity of a self-signed certificate, each client MUST manually verify and confirm the validity of the certificate.
One method is to give a copy of the self-signed certificate to the client (using a secure transport mechanism, not e-mail), and import it into the client directly. The stunnel client uses the CAfile option for this purpose.
Alternatively, many client programs (not stunnel) can interactively import the certificate when connecting to the server. The client will display information about the server's certificate including an MD5 key fingerprint. You MUST compare this fingerprint with the original fingerprint of the server's certificate.
Run the following command on the server to display the original certificate's fingerprint:
openssl x509 -fingerprint -in /etc/stunnel/stunnel.cert
Most clients will store the certificate for future reference, and will not need to do this verification step on further invocations.
In the evaluated configuration, you MUST use one of the following cipher suite suites as defined in the SSL v3 protocol:
# Cipher Proto Key Authen- Encryption Message
# exchg tication auth code
#
RC4-SHA SSLv3 Kx=RSA Au=RSA Enc=RC4(128) Mac=SHA
DES-CBC3-SHA SSLv3 Kx=RSA Au=RSA Enc=3DES(168) Mac=SHA1
AES128-SHA SSLv3 Kx=RSA Au=RSA Enc=AES(128) Mac=SHA1
AES256-SHA SSLv3 Kx=RSA Au=RSA Enc=AES(256) Mac=SHA1
You MUST specify the cipher list in all stunnel client and server configuration files:
ciphers = RC4-SHA:DES-CBC3-SHA:AES128-SHA:AES256-SHA
options = NO_TLSv1
options = NO_SSLv2
For a service or tunnel that will only be used temporarily, simply launch the stunnel program from the command line and specify an appropriate configuration file. The tunnel will be available for multiple clients, but will not be started automatically after a reboot. To shut down the tunnel, search for the command line in the ps ax process listing, and use the kill(1) command with the PID shown for the stunnel process.
The RECOMMENDED method is to use two separate configuration files, one for server definitions (incoming connections use SSL), and one for client definitions (outgoing connections use SSL). More complex configurations will require additional configuration files containing individual service-specific settings. You MUST use the REQUIRED settings in all stunnel configuration files.
Use the following content for the file /etc/stunnel/stunnel-server.conf:
### /etc/stunnel/stunnel-server.conf
#
# The following settings are REQUIRED for CAPP compliance when used
# as a server, see ECG. File names MAY be changed as needed.
cert = /etc/stunnel/stunnel.pem
ciphers = RC4-SHA:DES-CBC3-SHA:AES128-SHA:AES256-SHA
options = NO_TLSv1
options = NO_SSLv2
#
# User and group ID MUST NOT be "root", but MAY be changed as needed.
setuid = nobody
setgid = nobody
#
# The following settings are RECOMMENDED
debug = 6
output = /var/log/stunnel-server.log
pid =
foreground = yes
#
# Individual service definitions follow
Use the following content for the file /etc/stunnel/stunnel-client.conf:
### /etc/stunnel/stunnel-client.conf
#
# The following settings are REQUIRED for CAPP compliance when used
# as a client, see ECG. File names MAY be changed as needed. You
# MAY use CApath instead of CAfile for externally signed certificates.
CAfile = /etc/stunnel/stunnel.cert
ciphers = RC4-SHA:DES-CBC3-SHA:AES128-SHA:AES256-SHA
options = NO_TLSv1
options = NO_SSLv2
client = yes
verify = 2
#
# User and group ID MUST NOT be "root", but MAY be changed as needed.
setuid = nobody
setgid = nobody
#
# The following settings are RECOMMENDED
debug = 6
output = /var/log/stunnel-client.log
pid =
foreground = yes
#
# Individual service definitions follow
The RECOMMENDED launch method for stunnel(8) is via the init(8) process. This requires adding new entries to /etc/inittab, the tunnels will be re-launched automatically whenever they are terminated, as well as after a reboot. The following are the RECOMMENDED /etc/inittab entries:
ts:3:respawn:/usr/sbin/stunnel /etc/stunnel/stunnel-server.conf tc:3:respawn:/usr/sbin/stunnel /etc/stunnel/stunnel-client.conf
Make sure you use the option foreground = yes in the configuration file when running from init (otherwise init will misinterpret the backgrounded server as having died and will try to restart it immediately, causing a loop), and use the output option to redirect the output to a log file.
If the client program supports SSL encryption, it will be able to communicate with the stunnel service directly. You MUST verify and accept the server's certificate if the client cannot recognize it as valid according to its known certification authorities.
If the client program does not support SSL directly, you can use stunnel as a client, or indirectly by setting up a proxy that allows the client to connect to an unencrypted local TCP port.
WARNING: The stunnel client does not verify the server's certificate by default. You MUST specify either verify = 2 or verify = 3 in the client configuration file to switch on certificate verification.
You MAY also activate client certificate verification in the server's configuration file, so that the server can verify the client's identity as well.
You MUST specify the ciphers supported in the evaluated configuration in the configuration file as described in the previous section.
Normal SMTP e-mail delivery is not encrypted, but most mail clients support the enhanced SMTPS protocol that uses SSL encryption. The protocol itself is unchanged other than being encrypted.
stunnel can easily be used as a proxy to receive SMTPS connections on the standard port expected by clients (465/tcp), and then forward the data to the mail server listening on the SMTP port (25/tcp). The mail server configuration does not need to be modified to support encryption of incoming mail.
To implement SSL support for incoming mail, add the following service definition to the /etc/stunnel/stunnel-server.conf configuration:
[inbound_mail]
accept = 465
connect = 127.0.0.1:25
The following shell script acts as a simple web server, reading requests from standard input and writing HTTP/HTML to standard output:
cat > /usr/local/sbin/webserver_test <<-__EOF__
#!/bin/sh
# Simple web server, can be run via stunnel or xinetd
#
# read and discard client data
dd bs=65536 count=1 >/dev/null 2>&1
#
# Send HTTP header
echo -e "HTTP/1.0 200\r"
echo -e "Content-type: text/html\r"
echo -e "\r"
#
# Send HTML output
echo "<html>"
echo "<h1>Test Page</h1>"
date
echo "<h2>Memory usage</h2>"
echo "<pre>"
free
echo "</pre>"
echo "</html>"
__EOF__
chmod +x /usr/local/sbin/webserver_test
Add the following entry to the /etc/stunnel/stunnel-server.conf configuration to make this service available using the encrypted HTTPS protocol:
[webserver_test]
accept = 443
exec = /usr/local/sbin/webserver_test
TIMEOUTclose = 0
Then, use a SSL-capable web browser to connect to port 443:
elinks https://localhost/
This example shows how to combine stunnel client and server definitions to implement an encrypted tunnel for applications that do not themselves support encryption.
First, on the server machine, set up a stunnel server definition that accepts SSL connections on TCP port 444, and reports memory usage statistics for the server to connecting clients. Add the following service definition to the /etc/stunnel/stunnel-server.conf configuration:
[free]
accept = 444
exec = /usr/bin/free
execargs = free
Then, on the client machine, add the following entry to the /etc/stunnel/stunnel-client.conf configuration, using the server's IP address instead of "127.0.0.1":
[free]
accept = 81
connect = 127.0.0.1:444
On the client machine, connect to the local stunnel proxy by running the following command as a normal user:
telnet localhost 81
This will open an unencrypted TCP connection to the client's local port 81, then stunnel builds an encrypted tunnel to the server's port 444 and transfers the decrypted data (in this case, the "free" output) back to the client. All unencrypted connections are machine local, and the data transferred over the network is encrypted.
The security of the operating system depends on correctly functioning hardware. For example, the memory subsystem uses hardware support to ensure that the memory spaces used by different processes are protected from each other.
The Abstract Machine Testing Utility (AMTU) is distributed as an RPM, and was installed previously as described in section §4.13 of this guide.
To run all supported tests, simply execute the amtu program:
amtu
In the LSPP/RBAC mode, running amtu is only supported from a ssh session, not on a local console.
A successful run is indicated by the following output:
Executing Memory Test...
Memory Test SUCCESS!
Executing Memory Separation Test...
Memory Separation Test SUCCESS!
Executing Network I/O Tests...
Network I/O Controller Test SUCCESS!
Executing I/O Controller - Disk Test...
I/O Controller - Disk Test SUCCESS!
Executing Supervisor Mode Instructions Test...
Privileged Instruction Test SUCCESS!
The program will return a nonzero exit code on failure, which MAY be used to automatically detect failures of the tested systems and take appropriate action.
Please refer to the amtu(8) man page for more details.
You MUST verify periodically that the system clock is sufficiently accurate, otherwise log and audit files will contain misleading information. When starting the system, the time and date are copied from the computer's hardware clock to the kernel's software clock, and written back to the hardware clock on system shutdown.
All internal dates and times used by the kernel, such as file modification stamps, use universal time (UTC), and do not depend on the current time zone settings. Userspace utilities usually adjust these values to the currently active time zone for display. Note that text log files will contain ASCII time and date representations in local time, often without explicitly specifying the time zone.
The date(1) command displays the current time and date, and can be used by administrators to set the software clock, using the argument mmddHHMMyyyy to specify the numeric month, day, hour, minute and year respectively. For example, the following command sets the clock to May 1st 2004, 1pm in the local time zone:
date 050113002004
The hwclock(8) can query and modify the hardware clock on supported platforms. The typical use is to copy the current value of the software clock to the hardware clock. Note that the hardware clock MAY be running in either local time or universal time, as indicated by the UTC setting in the /etc/sysconfig/clock file. The following command sets the hardware clock to the current time using UTC:
hwclock -u -w
Use the command tzselect(8) to change the default time zone for the entire system. Note that users MAY individually configure a different time zone by setting the TZ environment variable appropriately in their shell profile, such as the $HOME/.bashrc file.
SELinux is a set of services provided by the kernel which are used to implement labeleled access protection as well as mult-level and multi-category security. These services fulfill the requirements of the LSPP and RBAC profiles. While the mechanisms are enforced by the kernel itself, a comprehensive rulset (the policy) needs to be passed into the kernel from userland.
SELinux is optional in the CAPP mode, and you MAY choose any of the available policies (the default "targeted" policy is RECOMMENDED), or choose to use none of them, all of which is beyond the scope of this document. This section contains documentation and requirements for the LSPP/RBAC mode.
In the LSPP/RBAC mode, you MUST use the "mls" policy.
Many system-level applications have been extended to provide support for the new facilities offered by SELinux, see the HLD and LLD and the respective man pages for more details. Most utilities use the argument -Z to display an extended set of information.
Serveral tools are available for administrative purposes and SELinux usage. We will discuss the central ones in the following.
Note that there are several manual pages which contain information about the SELinux aspects of standard services like the Apache web server (httpd_selinux(8)), the Samba server (samba_selinux(8)) etc.
The evaluated configuration keeps the SELinux system enabled in a static configuration. This is essential for LSPP/RBAC compliance. You MAY modify the SELinux configuration, for example to add additional restrictions or define additional roles as described in Section §4.16.5.
The /etc/selinux/config file has the following content by default:
SELINUX=enforcing
SELINUXTYPE=targeted
You MUST NOT disable SELinux by using one of the settings SELINUX=disabled or SELINUX=permissive. You MUST NOT modify the SELinux policy in any other way than described below. You MAY switch from targeted to strict policy by setting SELINUXTYPE=strict.
The predefined roles available are defined in section §6.7.1.
Users with the same name, but a different postfix (_u instead of _r) are provided for the predefined roles. The newrole(8) command can be used to switch between identities, obviously assuming that the desired transition is permitted by the policy.
The SELinux root user is assigned with the sysadm_r role by the standard policy. This gives sufficient rights to change into different roles and perform all standard administrative tasks
Adding unprivileged users is easy because no special steps need to be taken to benefit from SELinux protection. Since DAC users automatically assume the user_r role, using useradd(8) is fully sufficient.
Associating DAC users with other roles than user_r MAY be accomplished with the help of the semanage(8) command. The following input is necessary to associate DAC user bob with the SELinux role sysadm_r:
semanage login -a -s sysadm_r bob
The administrator MAY also define new SELinux users which combine several roles. The following commands define, for instance, a staff_u user which is authorized for the roles sysadm_r and user_r:
semanage user -a -R "sysadm_r user_r" staff_u
Privileged users asssume the staff_r role instead of the regular user_r role. As we have mentioned before, the two roles are quite similar, but staff_r is allowed to change into higher privileged roles. The connection with a DAC user MAY be performed as follows:
useradd privuser
# Edit /etc/selinux/policy/users/local.users
load_policy /etc/selinux/policy/policy/policy.19
genhomedircon
restorecon -R /home/privuser
policy is a placeholder for the policy which is in use, i.e., targeted or strict. The following line needs to be added to /etc/selinux/policy/users/local.users:
user privuser roles { staff_r sysadm_r };
This change requires reloading the policy with load_policy(8). Calling genhomedircon(8) and restorecon(8) is required to fix the file labels in the user's home directory.
Users with different roles than user_r MUST NOT be used for anything else than administrative tasks. The provileged roles MUST be given up as soon as possible. Note that if DAC users other than root are able to use sysadm_r and system_r roles, then the same extra care as with root should be applied to them.
The SELinux policy is a key stone to the security of RHEL5, so it is not possible for the administrator ot modify the policy at will. Otherwise, this would allow to make the system inconsistent with the requirements imposed by the EAL4+ criteria. Nevertheless, it is possible to add new elements to the policy without compromising the system's security, and we will introduce in the following how to do this. Because policy writing is a complex task, the scope of this document can obviously not be a complete introduction to all possible features. For this, we need to point the reader to the literature.
To ensure integrity of the system, some restrictions apply. The flask definitions MUST NOT be changed. The policy definitions for the trusted applications MUST NOT be changed. This especially implies that the user_r role MUST NOT be allowed to switch to any other role in a modified policy, and all other predefined roles MUST NOT be allowed to switch to another role than permitted by the RHEL standard policy.
You MUST NOT modify the constraints contained in the policy/mls file in the policy source code. It is RECOMMENDED that you do not modify the base policy at all and only add additional policy modules using the semodule(8) facility.
To modify the standard policy, it is necessary to install the following packages:
rpm-build
libselinux-devel
selinux-policy (source RPM, *.src.rpm)
The following commands are required to extract the source of the standard policy from the SRPM:
# Install the reference policy and RHEL5 specific patches
rpm --install selinux-policy-*.el5.src.rpm
# Unpack the source and apply the patches
cd /usr/src/redhat/SPECS && rpmbuild -bp specfile
This places a copy of the (patched) reference policy in /usr/src/redhat/BUILD/serefpolicy-*. Note that the data MAY be copied to another directory.
Compiling and loading a modified policy is performed by the following steps:
make conf
make
make install
See the files INSTALL and README for a more detailled list of available Makefile targets. The above steps generate a binary version of the policy in /etc/selinux/NAME. Note that the policy has not yet been loaded into the kernel. This requires the following extra step:
make load
When adding new elements to a policy, it is RECOMMENDED to create three files:
The files are supposed to be in policy/modules/apps for applications, and policy/modules/services for (network) services, and policy/modules/admin for administrative applications. It is NOT allowed to add new definitions to the contents of modules/kernel. Note that a great many number of definitions are included in the policy distribution which can serve as working examples.
The policy defines access rules in terms of subject type, object type, and attempted operation. Section §6.4.3 describes this from the user's point of view. Roles provide constraints on permitted types. When defining a role, you need to define any new types needed for the policy (though it is preferable to re-use existing types as appropriate), define the subject role and the subject types that may be used while in that role, and then define the permitted access to object types using allow rules. Note that the MLS constraints will continue to limit the permitted access even when granting rights with allow rules.
You can use the dominance operator when defining roles to simplify the policy. This operator permits defining roles in terms of other roles, for example:
dominance {
role supervisor_r {
role manager_r;
role programmer_r;
role tester_r;
role technical_writer_r;
}
}
After such a definition, the supervisor_r MAY be listed as a permitted role for a user using semanage(8):
semanage user -a -R supervisor_r -P manager supervisor_u
semanage login -a -s supervisor_u phb
This will give the "supervisor" class users such as "phb" the right to use any of the roles listed in the role definition, using newrole -r to switch between them. When a new subsidiary role is added to supervisor_r, it becomes available to all users of the class, without needing to use the semanage command again to add it to each user.
After adding additional policy, you MUST ensure that new objects are correctly labeled to match the policy. Use the restorecon(8) tool to set individual files to the default context as defined in the "*.fc" labeling policy files, or run touch /.autorelabel and reboot.
This section describes functionality specific for the LSPP/RBAC configuration. In CAPP mode, SELinux is optional, and some of these mechanisms MAY be used optionally, but that is beyond the scope of this document for CAPP mode.
The administrator MUST be logged in at the SystemLow MLS level for general administrative tasks to ensure that system configuration files and binaries remain at the correct MLS level. Administrative actions at other MLS levels can lead to critical files becoming mislabeled and unreadable at lower levels.
By default, administrators MUST use the sysadm_r role for general system administration tasks, this role is largely equivalent to the traditional root role. Alternatively, you MAY define site specific administrative roles for more fine grained access control.
Polyinstantiation is a mechanism to create multiple versions of common directories for different MLS levels, which ensures that legacy programs that are not aware of MLS restrictions continue to work in this environment. For example, applications generally expect to be able to write to /tmp and $HOME.
The pam_namespace.so module is used within the PAM configuration files as described in section §4.16.2 to enable this functionality.
You MAY change the polyinstantiation configuration, but changes to settings other than the parent directory location could cause applications to fail.
Please refer to the namespace.conf(5) and pam_namespace(8) manual pages for more information.
The /usr/share/selinux/devel/lspp_policy.te file defines additional audit properties for changing security relevant attributes through the /proc/self/attr/ pseudofile interface. These are OPTIONAL and MAY be changed.
The following is the default file content:
## Customized SELinux policy for LSPP evaluated configuration
policy_module(lspp_policy,1.0)
############################################################################# ### Additional audit #############################################################################
gen_require(`
attribute domain;
')
# Audit setting of security relevant process attributes # These settings are OPTIONAL auditallow domain self:process setcurrent; auditallow domain self:process setexec; auditallow domain self:process setfscreate; auditallow domain self:process setsockcreate;
When the file is changed, you MUST rebuild and reload the policy module:
cd /usr/share/selinux/devel
make lspp_policy.pp
semodule -i lspp_policy.pp
Administrators MAY assign MLS levels and categories to data as appropriate. Symbolic names for levels and categories MAY be assigned in the /etc/selinux/mls/setrans.conf file as documented in the file comments and the mcs(8) man page.
Administrators MAY define new roles, including hierarchical roles defined in terms of other roles. This requires adding new SELinux policy modules for the new roles and the permitted access rules. Please refer to section §4.16.5 for more information.
This section applies to the LSPP/RBAC mode only.
You MUST use either CIPSO or Labeled IPsec to associate MLS labels with network data. This is NOT automatically configured as part of the installation because activating it is likely to lock the administrator out from remote administration. Use a local or serial console, not ssh, to configure labeled networking.
You MUST enable one of the labeled networking mechanisms even on a standalone system with no external network connections to ensure that system local (loopback) communication is covered by the MLS policy.
The following minimal example enables CIPSO data label packet tags for both local and remote network communication:
netlabelctl -p cipsov4 add pass doi:1 tags:1
netlabelctl -p map del default
netlabelctl -p map add default protocol:cipsov4,1
When in a network that also contains systems not using CIPSO, you MUST also disable reception of unlabeled packets:
netlabelctl -p unlbl accept off
Please refer to the netlabelctl(8) man page and the /usr/share/doc/netlabel_tools-*/netlabelctl.txt file for more information.
IPsec is usually used to ensure confidentiality and integrity of network communication, both of which are beyond the scope of this evaluation. Labeled IPsec uses IPsec mechanisms to associate network connections with specific SELinux contexts, including MLS information.
The ipsec-tools package contains documentation about setting up IPsec, including the setkey(8), racoon(8), and racoon.conf(5) manual pages.
You MUST use the -ctx option of the setkey tool when adding Security Policy Database (SPD) entries to specify SELinux label information for all defined security associations, this has the side effect of activating MLS checking for the corresponding connections.
Please refer to the mailing list posting
§
for a detailed configuration example.
The rbac-self-test(1) tool incorporates a file integrity check component that detects unexpected changes to important system programs and configuration files, and that checks that security critical system functions such as the core MLS permission checks operate correctly.
Administrators MAY use this tool to perform these checks, either by running it manually at the command line, or in an automated function. The tool MAY be configured to automatically enter a secure state (single user mode) if the test indicates a problem.
Please refer to the rbac-self-test(1) online documentation for more information about the tool.
It is RECOMMENDED that you review the system's configuration at regular intervals to verify if it still agrees with the evaluated configuration. This primarily concerns those processes that may run with 'root' privileges.
The permissions of the device files /dev/* MUST NOT be modified.
In particular, review settings in the following files and directories to ensure that the contents and permissions have not been modified:
/etc/audit/audit.rules
/etc/audit/auditd.conf
/etc/cron.allow
/etc/cron.d/*
/etc/cron.deny
/etc/cron.daily/*
/etc/cron.hourly/*
/etc/cron.monthly/*
/etc/cron.weekly/*
/etc/crontab
/etc/group
/etc/gshadow
/etc/hosts
/etc/inittab
/etc/ld.so.conf
/etc/localtime
/etc/login.defs
/etc/modprobe.conf
/etc/modprobe.d/*
/etc/pam.d/*
/etc/passwd
/etc/rc.d/init.d/*
/etc/securetty
/etc/security/opasswd
/etc/selinux/*
/etc/shadow
/etc/ssh/sshd_config
/etc/stunnel/*
/etc/sysconfig/*
/etc/sysctl.conf
/etc/vsftpd.ftpusers
/etc/vsftpd/vsftpd.conf
/etc/xinetd.conf
/etc/xinetd.d/*
/var/log/faillog
/var/log/lastlog
/var/spool/cron/tabs/*
Use the command lastlog to detect unusual patterns of logins.
Also verify the output of the following commands (run as 'root'):
crontab -l
find / \( -perm -4000 -o -perm -2000 \) -ls
find / \( -type f -o -type d -o -type b \) -perm -0002 -ls
find /bin /boot /etc /lib /sbin /usr \
! -type l \( ! -uid 0 -o -perm +022 \)
System log messages are stored in the /var/log/ directory tree in plain text format, most are logged through the syslogd(8) and klogd(8) programs, which MAY be configured via the /etc/syslog.conf file.
The logrotate(8) utility, launched from /etc/cron.daily/logrotate, starts a fresh log file every week or when they reach a maximum size and automatically removes or archives old log files. You MAY change the configuration files /etc/logrotate.conf and /etc/logrotate.d/* as required.
In addition to the syslog messages, various other log files and status files are generated in /var/log by other programs:
File Source -------------|--------------------------------------------------------------- audit Directory for audit logs boot.msg Messages from system startup lastlog Last successful log in (see lastlog(8)) vsftpd.log Transaction log of the VSFTP daemon localmessages Written by syslog mail Written by syslog, contains messages from the MTA (postfix) messages Written by syslog, contains messages from su and ssh news/ syslog news entries (not used in the evaluated configuration) warn Written by syslog wtmp Written by the PAM susbystem, see who(1)
Please see syslog(3), syslog.conf(5) and syslogd(8) man pages for details on syslog configuration.
The ps(1) command can be used to monitor the currently running processes. Using ps faux will show all currently running processes and threads.
The audit subsystem implements a central monitoring solution to keep track of security relevant events, such as changes and change attempts to security critical files.
This is accomplished through two separate mechanisms. All system calls are intercepted, and the kernel writes the parameters and return value to the audit log for those calls that are marked as security relevant in the filter configuration. In addition, some trusted programs contain audit-specific code to write audit trails of the actions they are requested to perform.
Please refer to "Setting up the audit configuration files" of this guide and the auditd(8), auditd.conf(8), and auditctl(8) man pages for more information.
The Controlled Access Protection Profile (CAPP) and the Labeled Security Protection Profile (LSPP) specify the auditing capabilities that a compliant system must support. The evaluated configuration described here is based on these requirements.
WARNING: Some of the CAPP/LSPP requirements can conflict with your specific requirements for the system. For example, a CAPP/LSPP-compliant system MUST disable logins if the audit subsystem is not working. Please ensure that you are aware of the consequences if you enable auditing.
CAPP/LSPP are designed for a multiuser system, with multiple unique users who maintain both shared and private resources. The auditing features are intended to support this mode of operation with a reliable trail of security-relevant operations. It is less useful for a pure application server with no interactive users.
Please be aware that the auditing subsystem will, when activated, cause some slowdown for applications on the server. The impact depends on what the application is doing and how the audit subsystem is configured. As a rule of thumb, applications that open a large number of separate files are most affected, and CPU-bound programs should not be measurably affected. You will need to balance the performance requirements against your security needs when deciding if and how you want to use auditing.
You MAY make changes to the set of system calls and events that are to be audited. CAPP/LSPP requires that the system has the capability to audit security relevant events, but it is up to you to choose how you want to use these capabilities. It is acceptable to turn off system call auditing completely even in an evaluated configuration, for example on a pure application server with no interactive users on the system.
The audit package provides a suggested audit configuration for CAPP systems in the /usr/share/doc/audit-*/capp.rules file, and for LSPP/RBAC systems in the /usr/share/doc/audit-*/lspp.rules file. It contains a suggested setup for a typical multiuser system, all access to security relevant files is audited, along with other security relevant events such as system reconfiguration. You MAY copy this file to /etc/audit/audit.rules and modify the configuration according to your local requirements, including the option of using an empty audit rules file to disable auditing if not required.
You MAY selectively disable and enable auditing for specific events or users as required by modifying the audit.rules file as documented in the auditctl(8) man page. It is RECOMMENDED that you always reconfigure the audit system by modifying the /etc/audit/audit.rules file and then running the following command to reload the audit rules:
auditctl -R /etc/audit/audit.rules
This procedure ensures that the state of the audit system always matches the content of the /etc/audit/audit.rules file. You SHOULD NOT manually add and remove audit rules and watches on the command line as those changes are not persistent.
Note that reloading audit rules involves initially deleting all audit rules, and for a short time the system will be operating with no or only a partial set of audit rules. It is RECOMMENDED to make changes to the audit rules when no users are logged in on the system, for example by using single user mode or a reboot to activate the changes.
Please refer to the auditctl(8) man page for more details.
Use the ausearch(8) tool to retrieve information from the audit logs. The information available for retrieval depends on the active filter configuration. If you modify the filter configuration, it is RECOMMENDED keeping a datestamped copy of the applicable configuration with the log files for future reference.
For example:
# search for events by process ID
ausearch -p 3000
# search for events with a specific login UID
ausearch -ul jdoe
Please refer to the ausearch(8) man page for more details.
Of course, you can use other tools such as plain grep(1) or scripting languages such as awk(1), python(1) or perl(1) to further analyze the text audit log file or output generated by the low-level ausearch tool.
If the audit daemon is terminated, no audit events are saved until it is restarted. To avoid lost audit records when you have modified the filter configuration, you MUST use the command service auditd reload to re-load the filters.
You MUST NOT use the KILL signal (-9) to stop the audit daemon, doing so would prevent it from cleanly shutting down.
The default audit configuration stores audit records in the /var/log/audit/audit.log file. This is configured in the /etc/audit/auditd.conf file. You MAY change the auditd.conf file to suit your local requirements.
The most important settings concern handling situations where the audit system is at risk of losing audit information, such as due to lack of disk space or other error conditions. You MAY choose actions appropriate for your environment, such as switching to single user mode (action single) or shutting down the system (action halt) to prevent auditable actions when the audit records cannot be stored. For example, the following settings are RECOMMENDED in the /etc/audit/auditd.conf file if a fail-secure audit system is required:
admin_space_left_action = single
disk_full_action = halt
disk_error_action = halt
Note however, that going to single user mode does not guarantee that no more audit events occur. This is because init will only terminate processes it knows about, i.e., processes it has started or that have an associated stop script. Even these processes will be terminated one after another, and init waits on each individual stop script to return, allowing other programs to generate audit events meanwhile. Hence going to single user mode comes with the risk of losing audit data.
One could write a script to terminate processes init does not know about. However, this comes with a race condition. If a program keeps checking for the number of instances of itself and keeps forking child processes to keep that number of instances, there is no guarantee that the kill script will ever finish.
Hence it is RECOMMENDED that you configure the audit log demon to halt the system if it is running out of audit space.
It is RECOMMENDED that you configure appropriate disk space thresholds and notification methods to receive an advance warning when the space for audit records is running low.
It is RECOMMENDED that you use a dedicated partition for the /var/log/audit/ directory to ensure that auditd has full control over the disk space usage with no other processes interfering.
In the LSPP/RBAC mode, the email notification mechanism is not supported and would most likely require a MLS-aware mail system to work as intended. In the CAPP mode, it may or may not work depending on your SELinux policy and settings.
Please refer to the auditd.conf(5) man page for more information about the storage and handling of audit records.
auditd writes audit records using the normal Linux filesystem buffering, which means that information can be lost in a crash because it has not been written to the physical disk yet. Configuration options control how auditd handles disk writes and allow the administrator to choose an appropriate balance between performance and reliability.
Any applications that read the records while the system is running will always get the most current data out of the buffer cache, even if it has not yet been committed to disk, so the buffering settings do not affect normal operation.
The default setting is flush = DATA, ensuring that record data is written to disk, but metadata such as the last file time might be inconsistent.
The highest performance mode is flush = none, but be aware that this can cause loss of audit records in the event of a system crash.
If you want to ensure that auditd always forces a disk write for each record, you MAY set the flush = SYNC option in /etc/audit/auditd.conf, but be aware that this will result in significantly reduced performance and high strain on the disk.
A compromise between crash reliability and performance is to ensure a disk sync after writing a specific number of records to provide an upper limit for the number of records lost in a crash. For this, use a combination of flush = INCREMENTAL and a numeric setting for the freq parameter, for example:
flush = INCREMENTAL
freq = 100
The audit record files are not protected against a malicious administrator, and are not intended for an environment where the administrators are not trustworthy.
The system uses various files in /etc/sysconfig to configure the system. Most files in this directory tree contain variable definitions in the form of shell variables that are either read by the rc scripts at system boot time or are evaluated by other commands at runtime. Note that changes will not take effect until the affected service is restarted or the system is rebooted.
The system provides a large amount of online documentation, usually in text format. Use the man program to read entries in the online manual, for example:
man ls
man man
to read information about the ls and man commands respectively. You can search for keywords in the online manual with the apropos(1) utility, for example:
apropos password
When this guide refers to manual pages, it uses the syntax ENTRY(SECTION), for example ls(1). Usually you do not need to provide the section number, but if there are several entries in different sections, you can use the optional -S switch and pick a specific one.
Some programs provide additional information GNU 'texinfo' format, use the info program to read it, for example:
info diff
Additional information, sorted by software package, can be found in the /usr/share/doc/*/ directories. Use the less(1) pager to read it, for example:
less /usr/share/doc/bash*/FAQ
Many programs also support a -help, -? or -h switch you can use to get a usage summary of supported command-line parameters.
A collection of How-To documents in HTML format can be found under /usr/share/doc/howto/en/html if the optional howtoenh package is installed.
Please see /usr/share/doc/howto/en/html/Security-HOWTO for security information. The HTML files can be read with the elinks browser.
The RHEL server documentation is also available in electronic form in the directories /usr/share/doc/rhel*.
Note that this Configuration Guide has precedence over other documents in case of conflicting recommendations.
You MUST authenticate (prove your identity) before being permitted to use the system. When the administrator created your user account, he or she will have assigned a user name and default password, and provided that information for you along with instructions how to access the system.
Logging in to the system will usually be done using the Secure Shell (SSH) protocol, alternatively a serial terminal may be available. Use the ssh command to connect to the system unless instructed otherwise by the administrator, for example:
ssh jdoe@172.16.0.1
The ssh(1) manual page provides more information on available options. If you need to transfer files between systems, use the scp(1) or sftp(1) tools.
If this is the first time you are connecting to the target system, you will be prompted if you want to accept the host key. If the administrator has provided a key fingerprint for comparison, verify that they match, otherwise type yes to continue. You MUST immediately change your initially assigned password with the passwd(1) utility.
You MUST NOT under any circumstances attempt to log in from an insecure device, such as a public terminal or a computer belonging to a friend. Even if the person owning the computer is trustworthy, the computer may not be due to having been infected with malicious code. Always remember that the device you are typing your password into has the ability to save and re-use your authentication information, so you are in effect giving the computer you are using the right to do any and all actions in your name. Insecure handling of authentication information is the leading cause for exploits of otherwise secure systems, and SSH can only protect the information during transit, and offers no protection at all against an insecure end point.
When you log out from the system and leave the device you have used for access (such as a terminal or a workstation with terminal emulation), you MUST ensure that you have not left information on the screen or within an internal buffer that should not be accessible to another user. You should be aware that some terminals also store information not displayed on the terminal (such as passwords, or the contents of a scrollback buffer). Nevertheless this information may be extractable by the next user unless the terminal buffer has been cleared. Safe options include completely shutting down the client software used for access, powering down a hardware terminal, or clearing the scrollback buffer by switching among virtual terminals in addition to clearing the visible screen area.
If you ever forget your password, contact your administrator who will be able to assign a new password.
You MAY use the chsh(1) and chfn(1) programs to update your login shell and personal information if necessary. Not all settings can be changed this way, contact your administrator if you need to change settings that require additional privileges.
In the LSPP/RBAC mode, additional information is associated with the user's login session.
When logging in at a local console, you will be prompted for the desired SELinux context, including role and MLS level. You MAY accept the default by pressing the Return key, or you MAY select a different role or MLS level from the set of roles and range of MLS levels assigned for your user identity by the system administrator.
You MAY use the newrole(1) command with the -r option to change to a different role from your set of permitted roles.
On a local console (or other secure device with a type listed in /etc/selinux/mls/contexts/securetty_types), you MAY use the newrole(1) command with the -l option to change your active MLS level within your permitted clearance range:
newrole -l Secret
When using ssh(1), you MAY specify a role for the session by appending a slash and the name of the role to the username, for example:
ssh jdoe/staff_r@192.168.1.1
ssh -l jdoe/staff_r 192.168.1.1
When labeled networking is active, the MLS level will be automatically set based on the level of the incoming network connection.
All users, including the administrators, MUST ensure that their authentication passwords are strong (hard to guess) and handled with appropriate security precautions. The password policy described here is designed to satisfy the requirements of the evaluated configuration. If your organization already has a password policy defined, your administrator MAY refer you to that policy if it is equivalently strong.
You MUST change the initial password set by the administrator when you first log into the system. You MUST select your own password in accordance with the rules defined here. You MUST also change the password if the administrator has set a new password, for example if you have forgotten your password and requested the administrator to reset the password.
Use the passwd(1) program to change passwords. It will first prompt you for your old password to confirm your identity, then for the new password. You will be prompted to enter the new password twice, to catch mistyped passwords.
The passwd(1) program will automatically perform some checks on your new password to help ensure that it is not easily guessable, but you MUST nevertheless follow the requirements in this chapter.
Note that the administrators MUST also ensure that their own passwords comply with this password policy, even in cases where the automatic checking is not being done, such as when first installing the system.
Your password MUST be a minimum of 8 characters in length. More than 8 characters MAY be used (it is RECOMMENDED to use more than 8, best is to use passphrases), and all characters are significant.
Use either a password containing at least one character each from all four character classes; or a passphrase containing 12 total characters from any three of the four character classes. The character classes are defined as follows:
Lowercase letters: abcdefghijklmnopqrstuvwxyz
Uppercase letters: ABCDEFGHIJKLMNOPQRSTUVWXYZ
Digits: 0123456789
Punctuation: !"#$%&'()*+,-./:;<=>?[\]^_`{|}~
You MUST NOT base the password on a dictionary word, your real name, login name, or other personal details (such as dates, names of relatives or pets), or names of real people or fictional characters.
Instead of a password, you MAY use a passphrase consisting of multiple unrelated words (at least three) joined with random punctuation characters. Such a passphrase MUST have a length of at least 16 characters. (This corresponds to automatically generated pass phrases constructed by choosing 3 words from a 4096 word dictionary and adding two punctuation characters from a set of 8, equivalent to 42 bits of entropy.)
You MUST NOT use a simple alphabetic string, palindrome or combinations of adjacent keyboard keys.
When you choose a new password, it MUST NOT be a simple variation or permutation of a previously used one.
You MUST NOT write the password on paper or store it on electronic devices in unprotected form. Storage in a secure location (such as an envelope in a safety deposit box, or encrypted storage on an electronic device) MAY be acceptable, contact your administrator first to ensure that the protection is strong enough to make password recovery infeasible for the types of attackers the system is intended to protect against.
The password is for you and you only. A password is like a toothbrush - you do not want to share it with anybody, even your best friend. You MUST NOT disclose your password to anybody else, or permit anybody else to use the system using your identity.
Note that administrators will never ask you for your password, since they do not need it even if they are required to modify settings affecting your user account.
You MUST NOT use the same password for access to any systems under external administration, including Internet sites. You MAY however use the same password for accounts on multiple machines within one administrative unit, as long as they are all of an equivalent security level and under the control of the same administrators.
You MUST inform the administrator and select a new password if you have reason to believe that your password was accidentally disclosed to a third party.
If the system notifies you that your password will expire soon or has expired, choose a new one as instructed. Contact your administrator in case of difficulty.
A RECOMMENDED method of generating passwords that fits these criteria while still being easy to memorize is to base it on letters of words in a sentence (NOT a famous quotation), including capitalization and punctuation and one or two variations. Example:
"Ask not for whom the bell tolls."
=> An4wtbt.
"Password 'P'9tw;ciSd' too weak; contained in RHEL documentation"
=> P'9tw;ciRd
Linux is a multiuser operating system. You can control which other users will be able to read or modify your files by setting the Unix permission bits and user/group IDs, or (if more precise control is needed) by using POSIX-style access control lists (ACLs).
Note that the administrators ('root') are able to override these permissions and access all files on the system. Use of encryption is RECOMMENDED for additional protection of sensitive data.
The 'umask' setting controls the permissions of newly created files and directories and specifies the access bits that will be removed from new objects. Ensure that the setting is appropriate, and never grant write access to others by default. The umask MUST include at least the 002 bit (no write access for others), and the RECOMMENDED setting is 027 (read-only and execute access for the group, no access at all for others).
Do not set up world-writable areas in the filesystem - if you want to share files in a controlled manner with a fixed group of other users (such as a project group), please contact your administrator and request the creation of a user group for that purpose.
Always remember that you are responsible for the security of the data you create and use. Choose permissions that match the protection goals appropriate for the content, and that correspond to your organization's security policy. Access to confidential data MUST be on a need-to-know basis, do not make data world-readable unless the information is intended to be public.
Whenever you start a program or script, it will execute with your access rights. This implies that a malicious program would be able to read and modify all files that you have access to. Never execute any code that you have received from untrustworthy sources, and do not run commands that you do not understand. Be aware that manipulations to the environment a program is run in can also cause security flaws, such as leaking sensitive information. Do not use the shell variables LD_LIBRARY_PATH or LD_PRELOAD that modify the shared library configuration used by dynamically linked programs unless the specific settings are approved by the administrator or your organizational policies.
Programs can be configured to run with the access rights of the program file's owner and/or group instead of the rights of the calling user. This is the SUID/SGID mechanism, which utilities such as passwd(1) use to be able to access security-critical files. You could also create your own SUID/SGID programs via chmod(1), but DO NOT do that unless you fully understand the security implications - you would be giving away your access privileges to whoever launches the SUID program. Please refer to the "Secure Programming HOWTO" in the unlikely case that you need to create such a program, there you will find explanations of the many aspects that must be considered, such as the risk of unintended shell escapes, buffer overflows, resource exhaustion attacks and many other factors. Note that SUID root programs MUST NOT be added to the evaluated configuration, the only permitted use of the SUID bit is for setting non-root user IDs.
Please refer to the chmod(1), umask(2), chown(1), chgrp(1), acl(5), getfacl(1), and setfacl(1) manual pages for information, or any of the many available books covering Linux security (cf. Appendix 'Literature'), or ask your system administrator for advice.
This section applies in the LSPP/RBAC mode only.
Mandatory access control is based on sensitivity labels, which are checked by the SELinux module. Although SELinux provides the ability to implement a wealth of other security policies, the system implements mandatory access controls based on the Bell/LaPadula model of labeled access controls.
A sensitivity label is a tag consisting of of a sensitivity (or a range of sensitivities) and a set of categories. It is part of the security context attached to every subject and object.
All subjects and objects are assigned a sensitivity label, consisting of a hierarchical security level (or a range of security levels) and a set of categories. Sensitivity labels are part of the security context and hence attached to all subjects and objects covered by the access control policies. The management of sensitivity labels can be performed for each subject and each object separately.
Note that in filesystems that do not support extended attributes, objects inherit the security context of their mount point. For example, the VFAT file system mounted under /boot/efi inherits the sensitivity label that was assigned to the mount point during the mount operation.
SELinux supports 16 hierarchical sensitivity levels (s0 to s15) and 1024 categories (c0 to c1023).
At login time, a user is assigned the default sensitivity label (the lower bound of the range of sensitivity labels) and the default set of categories associated with the account. Users cannot change their sensitivity label or categories to values outside the range assigned to the account. Administrators MAY change the mapping using the semanage(8) utility's login subfunction.
The access check algorithm implemented by the Mandatory Access Control policy is as follows:
When comparing two sensitivity labels A and B, sensitivity label A dominates sensitivity label B if label A is equal or greater than label B. For sensitivity labels of subjects with different lower and upper bounds (i.e. with a real "range"), the lower bound is used in the checks as the "effective" sensitivity level. If objects have a different lower and upper bound of a sensitivity label, the mechanism verifies whether the subjects "effective" sensitivity label is within the range of the object's label range.
A subject can perform a read (or equivalent) operation on an object only if its sensitivity label dominates the object's sensitivity label.
A subject can perform a write (or equivalent) operation on an object only if the object's sensitivity label is equal the subject's sensitivity label. Note that equality is still a domination; SELinux therefore further restricts the LSPP MAC policy. If users wish to write an object at a higher label, they can do so by transitioning to that label (using the newrole command), which effectively limits write-up to the upper bound of the user's label range.
A subject in the sysadm_r role allows the configuration of the SELinux mechanism. This effectively implements a MAC override privilege. In addition, administrators MAY change object labels using the chcon(1) utility's -l option.
This section applies in the LSPP/RBAC mode only.
The role-based access control policy (RBAC) provides further restrictions on access to objects in addition to the DAC and MLS policies. Each user always has one effective role (shown as the second colon-separated field in the output of id -Z), chosen from the set of rule which the system administrator has associated with your account.
You MAY choose a role when logging in to the system when prompted to choose a context, or you MAY use the newrole(1) with the -r option to choose a role when logged on already.
Your current role influences which access rights you have to objects. Each object has a type (shown as the third colon-separated field in the output of ls -lZ), and your role, the object's type, and the type of access attempted (such as file read or write) is used to decide if the access is permitted or not.
The system administrator MAY grant you the right to change the type of objects among a set of permitted types. Use the chcon(1) command with the -t option to do so.
Please refer to the newrole(1) and chcon(1) manual pages for more information.
This section applies in the LSPP/RBAC mode only.
Printing of files is performed by the lp(1) utility through the cupsd(8) service which enforce the file's sensitivity label to be printed together with the contents.
It is possible to define multiple print queues and assign a single security level to each. This allows unlabeled printing (i.e., for PostScript documents) in a secure manner.
For printing labeled data, the print spooler converts any input information into bitmaps and subsequently puts the applicable label information on a separate banner and trailer page and on the header and bottom of each page.
Please refer to section §6.5 for more information.
The system comes with various tools to archive data (tar, star, cpio). If ACLs are used, then only star MUST be used to handle the files and directories as the other commands do not support ACLs. The options -H=exustar -acl must be used with star. If the system is in the LSPP/RBAC mode (and thus SELinux is enabled), then star MUST be used to perform the backup.
Please see the star(1) man page for more information.
The remainder of this section applies in the LSPP/RBAC mode only.
As the security context of a persistent object (including the sensitivity label) is stored in the file's extended attributes, sensitivity labels can be exported from and imported into the system together with the file they are attached to. RHEL provides the star command for export and import of files with their extended attributes.
Users can export data without labels to single-level devices allocated to them, if:
the user's sensitivity label is in the set of the device's sensitivity label;
the device's sensitivity label equals the sensitivity label of the data written to it.
Revoked access rights are enforced upon the next access check.
When transporting data over the network, the IPSec or CIPSO protocols can be used to preserve the sensitivity labels of the data in transmission. Please refer to section §6.6 for more information.
On object creation, the object's sensitivity label is set according to the transition rules of the policy, which use the subjects context and other security attributes (like the containing object's security context), depending on the object class of the newly created object.
Data can be exported without a label to a single-level device only. The label of that device must equal the sensitivity label of the exported data. Upon import of unlabeled data, the label of the user causing the import is used for the imported data.
This section applies in the LSPP/RBAC mode only.
Each subject has a security context, which contains a role and an SELinux user identitiy. The TOE is already preconfigured with different roles:
The system administrator role sysadm_r allows the configuration of the SELinux mechanism, including modification of labels (MAC override), configuration and review of audit. The system administrator role sysadm_r usually runs at the lowest sensitivity level (SystemLow) and has MAC override privileges.
The audit administrator role auditadm_r allows the configuration of the audit subsystem as well as the review of the audit trails.
The staff_r role contains all users which have the right to change to the sysadm_r and auditadm_r roles. This allows the prevention of the login of user with immediate roles of auditadm_r and sysadm_r. users must have the respective auditadm_r or sysadm_r role in their set of allowed roles to be able to change to it.
Normal unprivileged users are assigned to the user_r role. This role does not allow the use of any security relevant mechanism. Users with this role cannot switch to other roles.
The administrator defines the default role for each user. Only administrators can modify subject / role associations. Users can change to another role if they have the role in their set of allowed roles. Note that for the TOE, this only applies to users in the staff_r role; these users are allowed to transition to the secadm_r or sysadm_r role, if they hold this role in their set of allowed roles.
Every subject and object has a role assigned in their security context.
Every subject can hold only one role at any time as their active role.
When checking an access request for RBAC, the SELinux Security Server uses:
the security context of the object,
the security context of the subject
the security class of the object
and looks up the security policy database to determine the set of allowed operations for the subject on the object. It returns three access vectors to the policy enforcement module:
all allowed operations.
the set of operations that will generate a log entry even if the operation is allowed
the set of operations that do not generate a log entry if the operation is denied.
The policy enforcement module allows an operation if the operation is allowed in the "allowed" access vector.
Revoked access rights are enforced upon the next access check.
The SELinux framework ensures that only valid SELinux labels are allowed. An SELinux label is only valid with regards to roles if a user-chosen role (e.g. newrole) is within the defined set of roles associated with the user.
For access control in RBAC, each user has one active role. A user may have more than one role he can use, but there is always only one active role. No user can have an empty set of roles. Associated with each role are certain domain types. The administrator configures which types belong to which roles and which roles a user can assume using the semanage(8) utility.
Access checks are done based on the current user's domain type, the accessed object's object type, and the access mode. Access rules are administrator-defined in the SELinux policy. Object types MAY be changed using the chcon(1) utility.
Changing to a new role is a privileged action which a user can do using the newrole(1) program. That program checks that the user may actually change to that new role (the role must be assigned to that user by the administrator).
If there are conflicting recommendations in this guide and in one of the sources listed here, the Configuration Guide has precedence concerning the evaluated configuration.
"RHEL5 Deployment Guide", http://www.redhat.com/docs/manuals/enterprise/RHEL-5-manual/Deployment_Guide-en-US/index.html
"RHEL5 Installation Guide", http://www.redhat.com/docs/manuals/enterprise/RHEL-5-manual/Installation_Guide-en-US/index.html
David A. Wheeler, "Secure Programming for Linux and Unix HOWTO", file:///usr/share/doc/howto/en/html_single/Secure-Programs-HOWTO.html, http://tldp.org/HOWTO/Secure-Programs-HOWTO/
Kevin Fenzi, Dave Wreski, "Linux Security HOWTO", file:///usr/share/doc/howto/en/html_single/Security-HOWTO.html, http://www.linuxsecurity.com/docs/LDP/Security-HOWTO/
Frank Mayer, Karl MacMillan, David Caplan, "SELinux by example", Prentice Hall, 2006, ISBM 0131963694
Ellen Siever, Stephen Spainhour, Stephen Figgins, & Jessica P. Hekman, "Linux in a Nutshell, 3rd Edition", O'Reilly 2000, ISBN 0596000251
Simson Garfinkel, Gene Spafford, Alan Schwartz, "Practical Unix & Internet Security, 3rd Edition", O'Reilly 2003, ISBN 0596003234
Aeleen Frisch, "Essential System Administration, 3rd Edition", O'Reilly 2002, ISBN 0596003439
Daniel J. Barrett, Richard Silverman, "SSH, The Secure Shell: The Definitive Guide", O'Reilly 2001, ISBN 0596000111
David N. Blank-Edelman, "Perl for System Administration", O'Reilly 2000, ISBN 1565926099
Shelley Powers, Jerry Peek, Tim O'Reilly, Mike Loukides, "Unix Power Tools, 3rd Edition", O'Reilly 2002, ISBN 0596003307
W. Richard Stevens, "Advanced Programming in the UNIX(R) Environment", Addison-Wesley 1992, ISBN 0201563177
Linda Mui, "When You Can't Find Your UNIX System Administrator", O'Reilly 1995, ISBN 1565921046