Linux SUSE Linux Enterprise User Manual

Download

SUSE Linux Enterprise

www.novell.com10 SP1

May08,2008 The Linux Audit Framework

The Linux Audit Framework

Legal Notice

This manual is protected under Novell intellectual property rights. By reproducing, duplicating or distributing this manual you explicitly agree to conform to the terms and conditions of this license agreement.

This manual may be freely reproduced, duplicated and distributed either as such or as part of a bundled package in electronic and/or printed format, provided however that the following conditions are fullled:

That this copyright notice and the names of authors and contributors appear clearly and distinctively on all reproduced, duplicated and distributed copies. That this manual, specically for the printed format, is reproduced and/or distributed for noncommercial use only. The express authorization of Novell, Inc must be obtained prior to any other use of any manual or part thereof.

For Novell trademarks, see the Novell Trademark and Service Mark list http://www.novell

.com/company/legal/trademarks/tmlist.html. * Linux is a registered trademark of

Linus Torvalds. All other third party trademarks are the property of their respective owners. A trademark symbol (®, ™ etc.) denotes a Novell trademark; an asterisk (*) denotes a third party trademark.

All information found in this book has been compiled with utmost attention to detail. However, this does not guarantee complete accuracy. Neither Novell, Inc., SUSE LINUX Products GmbH, the authors, nor the translators shall be held liable for possible errors or the consequences thereof.

Contents

About This Guide v

1 Understanding Linux Audit 1

1.1 Introducing the Components of Linux Audit . . . . . . . . . . . . . . 3

1.2 Conguring the Audit Daemon . . . . . . . . . . . . . . . . . . . . 5

1.3 Controlling the Audit System Using auditctl . . . . . . . . . . . . . . 10

1.4 Passing Parameters to the Audit System . . . . . . . . . . . . . . . 11

1.5 Understanding the Audit Logs and Generating Reports . . . . . . . . . 15

1.6 Querying the Audit Daemon Logs with ausearch . . . . . . . . . . . . 27

1.7 Analyzing Processes with autrace . . . . . . . . . . . . . . . . . . 31

1.8 Visualizing Audit Data . . . . . . . . . . . . . . . . . . . . . . . 32

2 Setting Up the Linux Audit Framework 35

2.1 Determining the Components to Audit . . . . . . . . . . . . . . . 36

2.2 Conguring the Audit Daemon . . . . . . . . . . . . . . . . . . . 37

2.3 Enabling Audit for System Calls . . . . . . . . . . . . . . . . . . . 38

2.4 Setting Up Audit Rules . . . . . . . . . . . . . . . . . . . . . . . 39

2.5 Adjusting the PAM Conguration . . . . . . . . . . . . . . . . . . 40

2.6 Conguring Audit Reports . . . . . . . . . . . . . . . . . . . . . 41

2.7 Conguring Log Visualization . . . . . . . . . . . . . . . . . . . . 44

3 Introducing an Audit Rule Set 47

3.1 Adding Basic Audit Conguration Parameters . . . . . . . . . . . . . 48

3.2 Adding Watches on Audit Log Files and Conguration Files . . . . . . . 49

3.3 Monitoring File System Objects . . . . . . . . . . . . . . . . . . . 50

3.4 Monitoring Security Conguration Files and Databases . . . . . . . . . 51

3.5 Monitoring Miscellaneous System Calls . . . . . . . . . . . . . . . . 54

3.6 Filtering System Call Arguments . . . . . . . . . . . . . . . . . . . 54

3.7 Managing Audit Event Records Using Keys . . . . . . . . . . . . . . 57

4 Useful Resources 59

A Creating Flow Graphs from the Audit Statistics 61

B Creating Bar Charts from the Audit Statistics 65

About This Guide

The Linux audit framework as shipped with this version of SUSE Linux Enterprise provides a CAPP-compliant auditing system that reliably collects information about any security-relevant events. The audit records can be examined to determine whether any violation of the security policies has been committed and by whom.

Providing an audit framework is an important requirement for a CC-CAPP/EAL certication. Common Criteria (CC) for Information Technology Security Information is an international standard for independent security evaluations. Common Criteria helps customers judge the security level of any IT product they intend to deploy in missioncritical setups.

Common Criteria security evaluations have two sets of evaluation requirements, functional and assurance requirements. Functional requirements describe the security attributes of the product under evaluation and are summarized under the Controlled Access Protection Proles (CAPP). Assurance requirements are summarized under the Evaluation Assurance Level (EAL). EAL describes any activities that must take place for the evaluators to be condent that security attributes are present, effective, and implemented. Examples for activities of this kind include documenting the developers' search for security vulnerabilities, the patch process, and testing.

This guide provides a basic understanding of how audit works and how it can be set up. For more information about Common Criteria itself, refer to the Common Criteria Web site [http://www.commoncriteria-portal.org].

This guide contains the following:

Understanding Linux Audit

Get to know the different components of the Linux audit framework and how they interact with each other. Refer to this chapter for detailed background information.

Setting Up the Linux Audit Framework

Follow the instructions to set up an example audit conguration from start to nish. If you need a quick start document to get you started with audit, this chapter is it. If you need background information about audit, refer to Chapter 1, Understanding

Linux Audit (page 1) and Chapter 3, Introducing an Audit Rule Set (page 47).

Introducing an Audit Rule Set

Learn how to create an audit rule set that matches your needs by analyzing an example rule set.

Useful Resources

Check additional online and system information resources for more details on audit.

1 Feedback

We want to hear your comments and suggestions about this manual and the other documentation included with this product. Please use the User Comments feature at the bottom of each page of the online documentation and enter your comments there.

2 Documentation Updates

For the latest version of this documentation, see the SLES 10 SP1 doc Web site [http://www.novell.com/documentation/sles10].

3 Documentation Conventions

The following typographical conventions are used in this manual:

• /etc/passwd: lenames and directory names

• placeholder: replace placeholder with the actual value

• PATH: the environment variable PATH

• ls, --help: commands, options, and parameters

• user: users or groups

•

Alt, Alt + F1: a key to press or a key combination; keys are shown in uppercase as

on a keyboard

•

File, File > Save As: menu items, buttons

vi The Linux Audit Framework

• ►amd64 ipf: This paragraph is only relevant for the specied architectures. The arrows mark the beginning and the end of the text block.◄

►ipseries s390 zseries: This paragraph is only relevant for the specied architectures. The arrows mark the beginning and the end of the text block.◄

•

Dancing Penguins (Chapter Penguins, ↑Another Manual): This is a reference to a chapter in another manual.

About This Guide vii

Understanding Linux Audit

Linux audit helps make your system more secure by providing you with a means to analyze what is going on on your system in great detail. It does not, however, provide additional security itself—it does not protect your system from code malfunctions or any kind of exploits. Instead, Audit is useful for tracking these issues and helps you take additional security measures, like Novell AppArmor, to prevent them.

Audit consists of several components, each contributing crucial functionality to the overall framework. The audit kernel module intercepts the system calls and records the relevant events. The auditd daemon writes the audit reports to disk. Various command line utilities take care of displaying, querying, and archiving the audit trail.

Audit enables you to do the following:

Associate Users with Processes

Audit maps processes to the user ID that started them. This makes it possible for the administrator or security ofcer to exactly trace which user owns which process and is potentially doing malicious operations on the system.

IMPORTANT: Renaming User IDs

Audit does not handle the renaming of UIDs. Therefore avoid renaming UIDs (for example, changing tux from uid=1001 to uid=2000) and obsolete UIDs rather than renaming them. Otherwise you would need to change auditctl data (audit rules) and would have problems retrieving old data correctly.

Understanding Linux Audit 1

Review the Audit Trail

Linux audit provides tools that write the audit reports to disk and translate them into human readable format.

Review Particular Audit Events

Audit provides a utility that allows you to lter the audit reports for certain events of interest. You can lter for:

• User

• Group

• Audit ID

• Remote Hostname

• Remote Host Address

• System Call

• System Call Arguments

• File

• File Operations

• Success or Failure

Apply a Selective Audit

Audit provides the means to lter the audit reports for events of interest and also to tune audit to record only selected events. You can create your own set of rules and have the audit daemon record only those of interest to you.

Guarantee the Availability of the Report Data

Audit reports are owned by root and therefore only removable by root. Unauthorized users cannot remove the audit logs.

Prevent Audit Data Loss

If the kernel runs out of memory, the audit daemon's backlog is exceeded, or its rate limit is exceeded, audit can trigger a shutdown of the system to keep events from escaping audit's control. This shutdown would be an immediate halt of the system triggered by the audit kernel component without any syncing of the latest

2 The Linux Audit Framework

logs to disk. The default conguration is to log a warning to syslog rather than to halt the system.

If the system runs out of disk space when logging, the audit system can be congured to perform clean shutdown (init 0). The default conguration tells the audit daemon to stop logging when it runs out of disk space.

1.1 Introducing the Components of Linux Audit

The following gure illustrates how the various components of audit interact with each other:

Figure 1.1

Introducing the Components of Linux Audit

kernel

audit

auditd

auditd.conf

auditctl

audit.rules

audit.log

audispd

autrace

aureport

ausearch

application

Straight arrows represent the data ow between components while dashed arrows represent lines of control between components.

auditd

The audit daemon is responsible for writing the audit messages to disk that were generated through the audit kernel interface and triggered by application and system activity. How the audit daemon is started is controlled by its conguration le,

Understanding Linux Audit 3

/etc/sysconfig/auditd. How the audit system functions once it is started is controlled by /etc/auditd.conf. For more information about auditd and its conguration, refer to Section 1.2, “Conguring the Audit Daemon” (page 5).

auditctl

The auditctl utility controls the audit system. It controls the log generation parameters and kernel settings of the audit interface as well as the rule sets that determine which events are tracked. For more information about auditctl, refer to Section 1.3,

“Controlling the Audit System Using auditctl” (page 10).

audit rules

The le /etc/audit.rules contains a sequence of auditctl commands that are loaded at system boot time immediately after the audit daemon is started. For more information about audit rules, refer to Section 1.4, “Passing Parameters to

the Audit System” (page 11).

aureport

The aureport utility allows you to create custom reports from the audit event log. This report generation can easily be scripted and the output used by various other applications, for example, to plot these results. For more information about aureport, refer to Section 1.5, “Understanding the Audit Logs and Generating Reports” (page 15).

ausearch

The ausearch utility can search the audit log le for certain events using various keys or other characteristics of the logged format. For more information about ausearch, refer to Section 1.6, “Querying the Audit Daemon Logs with ausearch” (page 27).

audispd

The audit dispatcher daemon (audispd) can be used to relay event notications to other applications instead of or in addition to writing them to disk in the audit log.

autrace

The autrace utility traces individual processes in a fashion similar to strace. The output of autrace is logged to the audit log. For more information about autrace, refer to Section 1.7, “Analyzing Processes with autrace” (page 31).

4 The Linux Audit Framework

1.2 Conguring the Audit Daemon

Before you can actually start generating audit logs and process them, congure the audit daemon itself. Congure how it is started in the /etc/sysconfig/auditd conguration le and congure how the audit system functions once the daemon has been started in /etc/auditd.conf.

The most important conguration parameters in /etc/sysconfig/auditd are:

AUDITD_LANG="en_US" AUDITD_DISABLE_CONTEXTS="no"

AUDITD_LANG

The locale information used by audit. The default setting is en_US. Setting it to none would remove all locale information from audit's environment.

AUDITD_DISABLE_CONTEXTS

Disable system call auditing by default. Set to no for full audit functionality including le and directory watches and system call auditing.

The /etc/auditd.conf conguration le determines how the audit system functions once the daemon has been started. For most use cases, the default settings shipped with SUSE Linux Enterprise should sufce. For CAPP environments, most of these parameters need tweaking. The following list briey introduces the parameters available:

log_file = /var/log/audit/audit.log log_format = RAW priority_boost = 3 flush = INCREMENTAL freq = 20 num_logs = 4 dispatcher = /usr/sbin/audispd disp_qos = lossy max_log_file = 5 max_log_file_action = ROTATE space_left = 75 space_left_action = SYSLOG action_mail_acct = root admin_space_left = 50 admin_space_left_action = SUSPEND disk_full_action = SUSPEND disk_error_action = SUSPEND

Understanding Linux Audit 5

Depending on whether you want your environment to satisfy the requirements of CAPP, you need to be extra restrictive when conguring the audit daemon. Where you need to use particular settings to meet the CAPP requirements, a “CAPP Environment” note tells you how to adjust the conguration.

log_file and log_format

log_file species the location where the audit logs should be stored. log_format determines how the audit information is written to disk. Possible

values for log_format are raw (messages are stored just as the kernel sends them) or nolog (messages are discarded and not written to disk). The data sent to the audit dispatcher is not affected if you use the nolog mode. The default setting is raw and you should keep it if you want to be able to create reports and queries against the audit logs using the aureport and ausearch tools.

NOTE: CAPP Environment

In a CAPP environment, have the audit log reside on its own partition. By doing so, you can be sure that the space detection of the audit daemon is accurate and that you do not have other processes consuming this space.

priority_boost

Determine how much of a priority boost the audit daemon should get. Possible values are 0 to 3, with 3 assigning the highest priority. The values given here translate to negative nice values, as in 3 to -3 to increase the priority.

flush and freq

Species whether, how, and how often the audit logs should be written to disk. Valid values for flush are none, incremental, data, and sync. none tells the audit daemon not to make any special effort to write the audit data to disk. incremental tells the audit daemon to explicitly ush the data to disk. A frequency must be specied if incremental is used. A freq value of 20 tells the audit daemon to request the kernel to ush the data to disk after every 20 records. The data option keeps the data portion of the disk le in sync at all times while the sync option takes care of both metadata and data.

6 The Linux Audit Framework

NOTE: CAPP Environment

In a CAPP environment, make sure that the audit trail is always fully up to date and complete. Therefore, use sync or data with the flush parameter.

num_logs

Specify the number of log les to keep if you have given rotate as the max_log_file_action. Possible values range from 0 to 99. A value less than 2 means that the log les are not rotated at all. As you increase the number of les

to rotate, you increase the amount of work required of the audit daemon. While doing this rotation, auditd cannot always service new data that is arriving from the kernel as quickly, which can result in a backlog condition (triggering auditd to react according to the failure ag, described in Section 1.3, “Controlling the Audit System

Using auditctl” (page 10)). In this situation, increasing the backlog limit is recom-

mended. Do so by changing the value of the -b parameter in the /etc/audit

.rules le.

dispatcher and disp_qos

The dispatcher is started by the audit daemon during its start. The audit daemon relays the audit messages to the application specied in dispatcher. This application must be a highly trusted one, because it needs to run as root. disp_qos determines whether you allow for lossy or lossless communication between the audit daemon and the dispatcher. If you choose lossy, the audit daemon might discard some audit messages when the message queue is full. These events still get written to disk if log_format is set to raw, but they might not get through to the dispatcher. If you choose lossless the audit logging to disk is blocked until there is an empty spot in the message queue. The default value is lossy.

max_log_file and max_log_file_action

max_log_file takes a numerical value that species the maximum le size in

megabytes the log le can reach before a congurable action is triggered. The action to be taken is specied in max_log_file_action. Possible values for

max_log_file_action are ignore, syslog, suspend, rotate, and keep_logs. ignore tells the audit daemon to do nothing once the size limit is

reached, syslog tells it to issue a warning and send it to syslog, and suspend causes the audit daemon to stop writing logs to disk leaving the daemon itself still alive. rotate triggers log rotation using the num_logs setting. keep_logs

Understanding Linux Audit 7

also triggers log rotation, but does not use the num_log setting, so always keeps all logs.

NOTE: CAPP Environment

To keep a complete audit trail in CAPP environments, the keep_logs option should be used. If using a separate partition to hold your audit logs, adjust max_log_file and max_log_file_action to use the entire space available on that partition.

action_mail_acct

Specify an e-mail address or alias to which any alert messages should be sent. The default setting is root, but you can enter any local or remote account as long as e-mail and the network are properly congured on your system and /usr/lib/

sendmail exists.

space_left and space_left_action

space_left takes a numerical value in megabytes of remaining disk space that

triggers a congurable action by the audit daemon. The action is specied in

space_left_action. Possible values for this parameter are ignore, syslog, email, suspend, single, and halt. ignore tells the audit daemon to ignore

the warning and do nothing, syslog has it issue a warning to syslog, and email sends an e-mail to the account specied under action_mail_acct. suspend tells the audit daemon to stop writing to disk but remain alive while single triggers the system to be brought down to single user mode. halt triggers a full shutdown of the system.

NOTE: CAPP Environment

Make sure that space_left is set to a value that gives the administrator enough time to react to the alert and allows him to free enough disk space for the audit daemon to continue to work. Freeing disk space would involve calling aureport -t and archiving the oldest logs on a separate archiving partition or resource. The actual value for space_left depends on the size of your deployment. Set space_left_action to email.

admin_space_left and admin_space_left_action

admin_space_left takes a numerical value in megabytes of remaining disk

space. The system is already running low on disk space when this limit is reached

8 The Linux Audit Framework

and the administrator has one last chance to react to this alert and free disk space for the audit logs. The value of admin_space_left should be lower than the value for space_left. The values for admin_space_left_action are the same as for space_left_action.

NOTE: CAPP Environment

Set admin_space_left to a value that would just allow the administrator's actions to be recorded. The action should be set to single or halt.

disk_full_action

Specify which action to take when the system runs out of disk space for the audit logs. The possible values are the same as for space_left_action.

NOTE: CAPP Environment

As the disk_full_action is triggered when there is absolutely no more room for any audit logs, you should bring the system down to single-user mode (single) or shut it down completely (halt).

disk_error_action

Specify which action to take when the audit daemon encounters any kind of disk error while writing the logs to disk or rotating the logs. The possible value are the same as for space_left_action.

NOTE: CAPP Environment

Use syslog, single, or halt depending on your site's policies regarding the handling of any kind of hardware failure.

Once the daemon conguration in /etc/sysconfig/auditd and /etc/auditd .conf is complete, the next step is to focus on controlling the amount of auditing the

daemon does and to assign sufcient resources and limits to the daemon so it can operate smoothly.

Understanding Linux Audit 9

1.3 Controlling the Audit System Using auditctl

auditctl is responsible for controlling the status and some basic system parameters of the audit daemon. It controls the amount of auditing performed on the system. Using audit rules, auditctl controls which components of your system are subjected to the audit and to what extent they are audited. Audit rules can be passed to the audit daemon on the auditctl command line as well as by composing a rule set and instructing the audit daemon to process this le. By default, the rcaudit script is congured to check for audit rules under /etc/audit.rules. For more details on audit rules, refer to Section 1.4, “Passing Parameters to the Audit System” (page 11).

The main auditctl commands to control basic audit system parameters are:

• auditctl -e to enable or disable audit

• auditctl -f to control the failure ag

• auditctl -r to control the rate limit for audit messages

• auditctl -b to control the backlog limit

• auditctl -s to query the current status of the audit daemon

The -e, -f, -r, and -b options can also be specied in the audit.rules le to avoid having to enter them each time the audit daemon is started.

Audit status messages include information on each of the above-mentioned parameters. The following example highlights the typical audit status message. This message is output to the terminal any time you query the status of the audit daemon with auditctl

-s or change the status ag with auditctl -e flag.

Example 1.1

Querying the audit Status

AUDIT_STATUS: enabled=1 flag=2 pid=3105 rate_limit=0 backlog_limit=8192 lost=0 backlog=0

10 The Linux Audit Framework

Table 1.1

Audit Status Flags

CommandMeaning [Possible Values]Flag

auditctl -e [0|1]

Set the enable ag. [0|1]enabled

auditctl -f [0|1|2]

Set the failure ag. [0..2] 0=silent, 1=printk, 2=panic (immediate halt without syncing pending data to disk)

flag

—Process ID under which auditd is running.pid

auditctl -r

rate

Set a limit in messages per second. If the rate is not zero and it is exceeded, the action specied in the failure ag is triggered.

rate_limit

auditctl -b

backlog

Specify the maximum number of outstanding audit buffers allowed. If all buffers are full, the action specied in the failure ag is triggered.

backlog_limit

—Count the current number of lost audit

messages.

lost

—Count the current number of outstanding

audit buffers.

backlog

1.4 Passing Parameters to the Audit System

Commands to control the audit system can be invoked individually from the shell using auditctl or batch read from a le using auditctl -R. This second method is used by the init scripts to load rules from the le /etc/audit.rules after the audit daemon has been started. The rules are executed in order from top to bottom. Each of

Understanding Linux Audit 11

these rules would expand to a separate auditctl command. The syntax used in the rules le is the same as that used for the auditctl command.

Changes made to the running audit system by executing auditctl on the command line are not persistent across system restarts. For changes to persist, add them to the /etc/ audit.rules le and, if they are not currently loaded into audit, restart the audit system to load the modied rule set by using the rcauditd restart command.

Example 1.2

Example Audit Rules—Audit System Parameters

-b 1000❶

-f 1❷

-r 10❸

-e 1❹

❶

Specify the maximum number of outstanding audit buffers. Depending on the level of logging activity, you might need to adjust the number of buffers to avoid causing too heavy an audit load on your system.

❷

Specify the failure ag to use. See Table 1.1, “Audit Status Flags” (page 11) for possible values.

❸

Specify the maximum number of messages per second that may be issued by the kernel. See Table 1.1, “Audit Status Flags” (page 11) for details.

❹

Enable or disable the audit subsystem.

Using audit, you can track any kind of le system access to important les, congurations or resources. You can add watches on these and assign keys to each kind of watch for better identication in the logs.

12 The Linux Audit Framework

Example 1.3

Example Audit Rules—File System Auditing

-w /etc/shadow❶

-w /etc -p rx❷

-w /etc/passwd -k fk_passwd -p rwxa❸

❶

The -w option tells audit to add a watch to the le specied, in this case /etc/ shadow. All system calls requesting access permissions to this le are analyzed.

❷

This rule adds a watch to the /etc directory and applies permission ltering for read and execute access to this directory (-p wx). Any system call requesting any of these two permissions is analyzed. Only the creation of new les and the deletion of existing ones are logged as directory-related events. To get more specic events for les located under this particular directory, you should add a separate rule for each le. A le must exist before you add a rule containing a watch on it. Auditing les as they are created is not supported.

❸

This rule adds a le watch to /etc/passwd. Permission ltering is applied for read, write, execute, and attribute change permissions. The -k option allows you to specify a key to use to lter the audit logs for this particular event later.

System call auditing lets you track your system's behavior on a level even below the application level. When designing these rules, consider that auditing a great many system calls may increase your system load and cause you to run out of disk space due. Consider carefully which events need tracking and how they can be ltered to be even more specic.

Example 1.4

Example Audit Rules—System Call Auditing

-a entry,always -S mkdir❶

-a entry,always -S access -F a1=4❷

-a exit,always -S ipc -F a0=2❸

-a exit,always -S open -F success!=0❹

-a task,always -F auid=0❺

-a task,always -F uid=0 -F auid=501 -F gid=wheel❻

❶

This rule activates auditing for the mkdir system call. The -a option adds system call rules. This rule triggers an event whenever the mkdir system call is entered (entry, always). The -S option adds the system call to which this rule should be applied.

❷

This rule adds auditing to the access system call, but only but only if the second argument of the system call (mode) is 4 (R_OK). entry,always tells audit to

Understanding Linux Audit 13

add an audit context to this system call when entering it and write out a report as soon as the call exits.

❸

This rule adds an audit context to the IPC multiplexed system call. The specic

ipc system call is passed as the rst syscall argument and can be selected using

-F a0=ipc_call_number.

❹

This rule audits failed attempts to call open.

❺

This rule is an example of a task rule (keyword: task). It is different from the other rules above in that it applies to processes that are forked or cloned. To lter these kind of events, you can only use elds that are known at fork time, such as UID, GID, and AUID. This example rule lters for all tasks carrying an audit ID of 0.

❻

This last rule makes heavy use of lters. All lter options are combined with a logical AND operator, meaning that this rule applies to all tasks that carry the audit ID of 501, have changed to run as root, and have wheel as the group. A process is given an audit ID on user login. This ID is then handed down to any child process started by the initial process of the user. Even if the user changes his identity, the audit ID stays the same and allows tracing actions to the original user.

TIP: Filtering System Call Arguments

For more details on ltering system call arguments, refer to Section 3.6, “Filter-

ing System Call Arguments” (page 54).

You can not only add rules to the audit system, but also remove them. Delete rules are used to purge the rule queue of rules that might potentially clash with those you want to add. There are different methods for deleting the entire rule set at once or for deleting system call rules or le and directory watches:

Example 1.5

Deleting Audit Rules and Events

-D❶

-d entry,always -S mkdir❷

-W /etc❸

❶

Clear the queue of audit rules and delete any preexisting rules. This rule is used as the rst rule in /etc/audit.rules les to make sure that the rules that are about to be added do not clash with any preexisting ones. The auditctl

14 The Linux Audit Framework

-D command is also used before doing an autrace to avoid having the trace rules clash with any rules present in the audit.rules le.

❷

This rule deletes a system call rule. The -d option must precede any system call rule that should be deleted from the rule queue and must match exactly.

❸

This rule tells audit to discard the rule with the directory watch on /etc from the rules queue. This rule deletes any rule containing a directory watch on /etc regardless of any permission ltering or key options.

To get an overview of which rules are currently in use in your audit setup, run auditctl -l. This command displays all rules with one rule per line.

Example 1.6

Listing Rules with auditctl -l

LIST_RULES: exit,always watch=/etc perm=rx LIST_RULES: exit,always watch=/etc/passwd perm=rwxa key=fk_passwd LIST_RULES: exit,always watch=/etc/shadow perm=rwxa LIST_RULES: entry,always syscall=mkdir LIST_RULES: entry,always a1=4 (0x4) syscall=access LIST_RULES: exit,always a0=2 (0x2) syscall=ipc LIST_RULES: exit,always success!=0 syscall=open

NOTE: Creating Filter Rules

You can build very sophisticated audit rules by using the various lter options. Refer to the auditctl(8) man page for more information about options available for building audit lter rules and audit rules in general.

1.5 Understanding the Audit Logs and Generating Reports

To understand what the aureport utility does, it is vital to know how the logs generated by the audit daemon are structured and what exactly is recorded for an event. Only then can you decide which report types are most appropriate for your needs.

Understanding Linux Audit 15

1.5.1 Understanding the Audit Logs

The following examples highlight two typical events that are logged by audit and how their trails in the audit log are read. The audit log or logs (if log rotation is enabled) are stored in the /var/log/audit directory. The rst example is a simple less command. The second example covers a great deal of PAM activity in the logs when a user tries to remotely log in to a machine running audit.

Example 1.7

A Simple Audit Event—Viewing the Audit Log

type=SYSCALL msg=audit(1175176190.105:157): arch=40000003 syscall=5 success=yes exit=4 a0=bfba161c a1=8000 a2=0 a3=8000 items=1 ppid=4457 pid=4462 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts0 comm="less" exe="/usr/bin/less" subj=unconstrained key="LOG_audit_log" type=CWD msg=audit(1175176190.105:157): cwd="/tmp" type=PATH msg=audit(1175176190.105:157): item=0 name="../var/log/audit/audit.log" inode=458325 dev=03:01 mode=0100640 ouid=0 ogid=0 rdev=00:00

The above event, a simple less /var/log/audit/audit.log, wrote three messages to the log. All of them are closely linked together and you would not be able to make sense of one of them without the others. The rst message reveals the following information:

type

The type of event recorded. In this case, it assigns the SYSCALL type to an event triggered by a system call (less or rather the underlying open). The CWD event was recorded to record the current working directory at the time of the syscall. A PATH event is generated for each path passed to the system call. The open system call takes only one path argument, so only generates one PATH event. It is important to understand that the PATH event reports the pathname string argument without any further interpretation, so a relative path requires manual combination with the path reported by the CWD event to determine the object accessed.

msg

A message ID enclosed in brackets. The ID splits into two parts. All characters before the : represent a UNIX epoch time stamp. The number after the colon represents the actual event ID. All events that are logged from one application's system call have the same event ID. If the application makes a second system call, it gets another event ID.

16 The Linux Audit Framework

arch

References the CPU architecture of the system call. Decode this information using the -i option on any of your ausearch commands when searching the logs.

syscall

The type of system call as it would have been printed by an strace on this particular system call. This data is taken from the list of system calls under /usr/include/

asm/unistd.h and may vary depending on the architecture. In this case, syscall=5 refers to the open system call (see man open(2)) invoked by the

less application.

success

Whether the system call succeeded or failed.

exit

The exit value returned by the system call. For the open system call used in this example, this is the le descriptor number. This varies by system call.

a0 to a3

The rst four arguments to the system call in numeric form. The values of these are totally system call dependent. In this example (an open system call), the following are used:

a0=bfba161c a1=8000 a2=0 a3=8000

a0 is the start address of the passed pathname. a1 is the ags. 8000 in hex notation translates to 100000 in octal notation, which in turn translates to O_LARGEFILE. a2 is the mode, which, because O_CREAT was not specied, is unused. a3 is not

passed by the open system call. Check the manual page of the respective system call to nd out which arguments are used with it.

items

The number of strings passed to the application.

ppid

The process ID of the parent of the process analyzed.

pid

The process ID of the process analyzed.

Understanding Linux Audit 17

auid

The audit ID. A process is given an audit ID on user login. This ID is then handed down to any child process started by the initial process of the user. Even if the user changes his identity (for example, becomes root), the audit ID stays the same. Thus you can always trace actions to the original user who logged in.

uid

The user ID of the user who started the process. In this case, 0 for root.

gid

The group ID of the user who started the process. In this case, 0 for root.

euid, suid, fsuid

Effective user ID, set user ID, and le system user ID of the user that started the process.

egid, sgid, fsgid

Effective group ID, set group ID, and le system group ID of the user that started the process.

tty

The terminal from which the application is started. In this case, a pseudoterminal used in an SSH session.

comm

The application name under which it appears in the task list.

exe

The resolved pathname to the binary program.

subj

auditd records whether the process is subject to any security context, such as AppArmor. unconstrained, as in this case, means that the process is not conned with AppArmor. If the process had been conned with audit, the binary pathname plus the AppArmor prole mode would have been logged.

key

If you are auditing a large number of directories or les, assign key strings each of these watches. You can use these keys with ausearch to search the logs for events of this type only.

18 The Linux Audit Framework

The second message triggered by the example less call does not reveal anything apart from just the current working directory when the less command was executed.

The third message reveals the following (the type and message ags have already been introduced):

item

In this example, item references the a0 argument—a path—that is associated with the original SYSCALL message. Had the original call had more than one path argument (such as a cp or mv command), an additional PATH event would have been logged for the second path argument.

name

Refers to the pathname passed as an argument to the less (or open) call.

inode

Refers to the inode number corresponding to name.

dev

Species the device on which the le is stored. In this case, 03:01, which stands for /dev/hda1 or “rst partition on the rst IDE device.”

mode

Numerical representation of the le's access permissions. In this case, root has read and write permissions and his group (root) has read access while the entire rest of the world cannot access the le at all.

ouid and ogid

Refer to the UID and GID of the inode itself.

rdev

Not applicable for this example. The rdev entry only applies to block or character devices, not to les.

Example 1.8, “An Advanced Audit Event—Login via SSH” (page 20) highlights the

audit events triggered by an incoming SSH connection. Most of the messages are related to the PAM stack and reect the different stages of the SSH PAM process. Several of the audit messages carry nested PAM messages in them that signify that a particular stage of the PAM process has been reached. Although the PAM messages are logged by audit, audit assigns its own message type to each event:

Understanding Linux Audit 19

Example 1.8

An Advanced Audit Event—Login via SSH

type=USER_AUTH msg=audit(1175508928.540:4499): user pid=28731 uid=0 ❶ auid=0 msg='PAM: authentication acct=root : exe="/usr/sbin/sshd" (hostname=earth.example.com, addr=192.168.0.1, terminal=ssh res=success)' type=USER_ACCT msg=audit(1175508928.540:4500): user pid=28731 uid=0 ❷ auid=0 msg='PAM: accounting acct=root : exe="/usr/sbin/sshd" (hostname=earth.example.com, addr=192.168.0.1, terminal=ssh res=success)' type=CRED_ACQ msg=audit(1175508928.544:4501): user pid=28729 uid=0 ❸ auid=0 msg='PAM: setcred acct=root : exe="/usr/sbin/sshd" (hostname=earth.example.com, addr=192.168.0.1, terminal=/dev/pts/0 res=success)' type=USER_LOGIN msg=audit(1175508928.544:4502): user pid=28732 uid=0 ❹ auid=0 msg='uid=0: exe="/usr/sbin/sshd" (hostname=earth.example.com, addr=192.168.0.1, terminal=/dev/pts/0 res=success)' type=USER_START msg=audit(1175508928.548:4503): user pid=28732 uid=0 ❺ auid=0 msg='PAM: session open acct=root : exe="/usr/sbin/sshd" (hostname=earth.example.com, addr=192.168.0.1, terminal=/dev/pts/0 res=success)' type=CRED_REFR msg=audit(1175508928.548:4504): user pid=28732 uid=0 ❻ auid=0 msg='PAM: setcred acct=root : exe="/usr/sbin/sshd" (hostname=earth.example.com, addr=192.168.0.1, terminal=/dev/pts/0 res=success)'

❶

PAM reports that is has successfully requested user authentication for root from a remote host (earth.example.com, 192.168.0.1). The terminal where this is happening is ssh.

❷

PAM reports that it has successfully determined whether the user is authorized to log in at all.

❸

PAM reports that the appropriate credentials to log in have been acquired and that the terminal changed to a normal terminal (/dev/pts0).

❹

The user has successfully logged in. This event is the one used by aureport

-l to report about user logins.

❺

PAM reports that it has successfully opened a session for root.

❻

PAM reports that the credentials have been successfully reacquired.

1.5.2 Generating Custom Audit Reports

The raw audit reports stored in the /var/log/audit directory tend to become very bulky and hard to understand. To nd individual events of interest, you might have to

20 The Linux Audit Framework

read through thousands of other events before you spot the one that you want. To avoid this, use the aureport utility and create custom reports.

The following use cases highlight just a few of the possible report types that you can generate with aureport:

Read Audit Logs from Another File

When the audit logs have moved to another machine or when you want to analyze the logs of a number of machines on your local machine without wanting to connect to each of these individually, move the logs to a local le and have aureport analyze them locally:

aureport -if myfile

Summary Report ====================== Range of time: 04/19/2007 13:42:43.280 - 04/23/2007 21:11:21.533 Number of changes in configuration: 55 Number of changes to accounts, groups, or roles: 0 Number of logins: 20 Number of failed logins: 10 Number of users: 3 Number of terminals: 11 Number of host names: 5 Number of executables: 12 Number of files: 3 Number of AVC denials: 0 Number of MAC events: 0 Number of failed syscalls: 4 Number of anomaly events: 0 Number of responses to anomaly events: 0 Number of crypto events: 0 Number of process IDs: 544 Number of events: 2795

The above command, aureport without any arguments, provides just the standard general summary report generated from the logs contained in myfile. To create more detailed reports, combine the -if option with any of the options below. For example, generate a login report that is limited to a certain time frame:

aureport -l -ts 12:00 -te 13:00 -if myfile

Login Report ============================================ # date time auid host term exe success event ============================================

1. 04/23/2007 12:38:38 PM root earth /dev/pts/0 /usr/sbin/sshd yes 1624

Understanding Linux Audit 21

Convert Numeric Entities to Text

Some information, such as user IDs, are printed in numeric form. To convert these into a human-readable text format, add the -i option to your aureport command.

Create a Rough Summary Report

If you are just interested in the current audit statistics (events, logins, processes, etc.), run aureport without any other option:

aureport

Create a Summary Report of Failed Events

If you want to break down the overall statistics of plain aureport to the statistics of failed events, use aureport --failed:

aureport --failed

Failed Summary Report ====================== Range of time: 04/19/2007 13:42:43.280 - 04/23/2007 21:25:38.406 Number of changes in configuration: 0 Number of changes to accounts, groups, or roles: 0 Number of logins: 0 Number of failed logins: 10 Number of users: 1 Number of terminals: 6 Number of host names: 4 Number of executables: 4 Number of files: 1 Number of AVC denials: 0

22 The Linux Audit Framework

Number of MAC events: 0 Number of failed syscalls: 4 Number of anomaly events: 0 Number of responses to anomaly events: 0 Number of crypto events: 0 Number of process IDs: 21 Number of events: 32

Create a Summary Report of Successful Events

If you want to break down the overall statistics of a plain aureport to the statistics of successful events, use aureport --success:

aureport --success

Success Summary Report ====================== Range of time: 04/19/2007 13:42:43.280 - 04/23/2007 21:31:35.865 Number of changes in configuration: 55 Number of changes to accounts, groups, or roles: 0 Number of logins: 20 Number of failed logins: 0 Number of users: 3 Number of terminals: 10 Number of host names: 5 Number of executables: 12 Number of files: 4 Number of AVC denials: 0 Number of MAC events: 0 Number of failed syscalls: 0 Number of anomaly events: 0 Number of responses to anomaly events: 0 Number of crypto events: 0 Number of process IDs: 535 Number of events: 2787

Create Summary Reports

In addition to the dedicated summary reports (main summary and failed and success summary), use the --summary option with most of the other options to create summary reports for a particular area of interest only. Not all reports support this option, however. This example creates a summary report for user login events:

aureport -u --summary

User Summary Report =========================== total auid =========================== 5640 root 13 tux 3 geeko 3 news

Understanding Linux Audit 23

Create a Report of Events

To get an overview of the events logged by audit, use the aureport -e command. This command generates a numbered list of all events including date, time, event number, event type, and audit ID.

aureport -e

Event Report =========================== # date time event type auid ===========================

1. 04/23/2007 08:00:01 AM 1507 USER_ACCT unset

2. 04/23/2007 08:00:01 AM 1508 CRED_ACQ unset

3. 04/23/2007 08:00:01 AM 1509 LOGIN root

4. 04/23/2007 08:00:01 AM 1510 USER_START root

Create a Report from All Process Events

To analyze the log from a process's point of view, use the aureport -p command. This command generates a numbered list of all process events including date, time, process ID, name of the executable, system call, audit ID, and event number.

aureport -p

Process ID Report ====================================== # date time pid exe syscall auid event ======================================

1. 04/23/2007 08:00:01 PM 13097 /usr/sbin/cron 0 unset 1888

2. 04/23/2007 08:00:01 PM 13097 /usr/sbin/cron 0 unset 1889

3. 04/23/2007 08:00:01 PM 13097 ? 0 root 1890

Create a Report from All System Call Events

To analyze the audit log from a system call's point of view, use the aureport

-s command. This command generates a numbered list of all system call events including date, time, number of the system call, process ID, name of the command that used this call, audit ID, and event number.

aureport -s

Syscall Report ======================================= # date time syscall pid comm auid event =======================================

1. 04/23/2007 08:04:08 PM 5 13374 file root 1900

2. 04/23/2007 08:04:08 PM 5 13376 file root 1901

3. 04/23/2007 08:04:08 PM 5 13368 less root 1902

24 The Linux Audit Framework

Create a Report from All Executable Events

To analyze the audit log from an executable's point of view, use the aureport

-x command. This command generates a numbered list of all executable events including date, time, name of the executable, the terminal it is run in, the host executing it, the audit ID, and event number.

aureport -x

Executable Report ==================================== # date time exe term host auid event ====================================

1. 04/23/2007 08:00:01 PM /usr/sbin/cron cron ? unset 1888

2. 04/23/2007 08:00:01 PM /usr/sbin/cron cron ? unset 1889

3. 04/23/2007 08:00:01 PM /usr/sbin/cron cron ? root 1891

Create a Report about Files

To generate a report from the audit log that focuses on le access, use the aureport -f command. This command generates a numbered list of all lerelated events including date, time, name of the accessed le, number of the system call accessing it, success or failure of the command, the executable accessing the le, audit ID, and event number.

aureport -f

File Report =============================================== # date time file syscall success exe auid event ===============================================

1. 04/23/2007 06:16:38 PM /var/log/audit/audit.log 5 yes /usr/bin/file root 1822

2. 04/23/2007 06:16:38 PM /var/log/audit/audit.log 5 yes /usr/bin/file root 1823

3. 04/23/2007 06:16:38 PM /var/log/audit/audit.log 5 yes /usr/bin/less root 1824

Create a Report about Users

To generate a report from the audit log that illustrates which users are running what executables on your system, use the aureport -u command. This command generates a numbered list of all user-related events including date, time, audit ID, terminal used, host, name of the executable, and an event ID.

aureport -u

User ID Report ==================================== # date time auid term host exe event ====================================

Understanding Linux Audit 25

1. 04/23/2007 08:00:01 PM unset cron ? /usr/sbin/cron 1888

2. 04/23/2007 08:00:01 PM unset cron ? /usr/sbin/cron 1889

3. 04/23/2007 08:00:01 PM root ? ? ? 1890

4. 04/23/2007 08:00:01 PM root cron ? /usr/sbin/cron 1891

5. 04/23/2007 08:00:01 PM root cron ? /usr/sbin/cron 1892

6. 04/23/2007 08:00:01 PM root cron ? /usr/sbin/cron 1893

7. 04/23/2007 08:04:01 PM unset ssh 192.168.0.20 /usr/sbin/sshd 1894

Create a Report about Logins

To create a report that focuses on the login attempts to your machine, run the aureport -l command. This command generates a numbered list of all loginrelated events including date, time, audit ID, host and terminal used, name of the executable, success or failure of the attempt, and an event ID.

aureport -l

Login Report ============================================ # date time auid host term exe success event ============================================

1. 04/23/2007 12:38:38 PM root earth.example.com /dev/pts/0 /usr/sbin/sshd yes 1624

2. 04/23/2007 01:38:12 PM root earth.example.com /dev/pts/1 /usr/sbin/sshd yes 1655

3. 04/23/2007 03:32:58 PM root 192.168.0.20 /dev/pts/0 /usr/sbin/sshd yes 1712

Limit a Report to a Certain Time Frame

To analyze the logs for a particular time frame, such as only the working hours of April 23, 2007, rst nd out whether this data is contained in the the current audit

.log or whether the logs have been rotated in by running aureport -t:

aureport -t

Log Time Range Report ===================== /var/log/audit/audit.log: 04/19/2007 13:42:43.280 - 04/23/2007 22:19:08.087

The current audit.log contains all the desired data. Otherwise, use the -if option to point the aureport commands to the log le that contains the needed data.

Then, specify the start date and time and the end date and time of the desired time frame and combine it with the report option needed. This example focuses on login attempts:

aureport -ts 04/23/2007 8:00 -te 04/23/2007 17:00 -l

26 The Linux Audit Framework

============================================ # date time auid host term exe success event ============================================

1. 04/23/2007 12:38:38 PM root earth.example.com /dev/pts/0 /usr/sbin/sshd yes 1624

2. 04/23/2007 01:38:12 PM root earth.example.com /dev/pts/1 /usr/sbin/sshd yes 1655

3. 04/23/2007 03:32:58 PM root sun.example.com /dev/pts/0 /usr/sbin/sshd yes 1712

The start date and time are specied with the -ts option. Any event that has a time stamp equal to or after your given start time appears in the report. If you omit the date, aureport assumes that you meant today. If you omit the time, it assumes that the start time should be midnight of the date specied. Use the 24 clock notation rather than the 12 hour one and adjust the date format to your locale (specied in /etc/sysconfig/audit under AUDITD_LANG, default is en_US).

Specify the end date and time with the -te option. Any event that has a time stamp equal to or before your given event time appears in the report. If you omit the date, aureport assumes that you meant today. If you omit the time, it assumes that the end time should be now. Use a similar format for the date and time as for -ts.

All reports except the summary ones are printed in column format and sent to stdout, which means that this data can be piped to other commands very easily. The visualization scripts introduced in Section 1.8, “Visualizing Audit Data” (page 32) are just one ex- ample of how to further process the data generated by audit.

1.6 Querying the Audit Daemon Logs with ausearch

The aureport tool helps you to create overall summaries of what is happening on the system, but if you are interested in the details of a particular event, ausearch is the tool to use. ausearch allows you to search the audit logs using special keys and search phrases that relate to most of the ags that appear in event messages in /var/log/ audit/audit.log. Not all record types contain the same search phrases. There are no hostname or uid entries in a PATH record, for example. When searching, make sure that you choose appropriate search criteria to catch all records you need. On the other hand, you could be searching for a specic type of record and still get various other related records along with it. This is caused by different parts of the kernel contributing additional records for events that are related to the one to nd. For example,

Understanding Linux Audit 27

you would always get a PATH record along with the SYSCALL record for an open system call.

TIP: Using Multiple Search Options

Any of the command line options can be combined with logical AND operators to narrow down your search.

Read Audit Logs from Another File

ausearch -option -if myfile

Convert Numeric Results into Text

Some information, such as user IDs are printed in numeric form. To convert these into human readable text format, add the -i option to your ausearch command.

Search by Audit Event ID

If you have previously run an audit report or done an autrace, you might want to analyze the trail of a particular event in the log. Most of the report types described in Section 1.5, “Understanding the Audit Logs and Generating Reports” (page 15) include audit event IDs in their output. An audit event ID is the second part of an audit message ID, which consists of a UNIX epoch time stamp and the audit event ID separated by a colon. All events that are logged from one application's system call have the same event ID. Use this event ID with ausearch to retrieve this event's trail from the log.

The autrace tool asks you to review the complete trail of the command traced in the logs using ausearch. autrace provides you with the complete ausearch command including the audit event ID.

In both cases, use a command similar to the following:

ausearch -a 5451

time->Wed Apr 25 21:59:44 2007 type=PATH msg=audit(1177531184.201:5451): item=0 name="/var/log/audit" inode=651613 dev=03:01 mode=040700 ouid=0 ogid=0 rdev=00:00 type=CWD msg=audit(1177531184.201:5451): cwd="/root" type=SYSCALL msg=audit(1177531184.201:5451): arch=40000003 syscall=5

28 The Linux Audit Framework

success=yes exit=4 a0=80624a0 a1=18800 a2=0 a3=80624a0 items=1 ppid=29163 pid=29433 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts2 comm="aureport" exe="/sbin/aureport" subj=unconstrained key="LOG_audit"

The ausearch -a command grabs all records in the logs that are related to the audit event ID provided and displays them. This option cannot be combined with any other option.

Search by Message Type

To search for audit records of a particular message type, use the ausearch -m

message_type command. Examples of valid message types include PATH, SYSCALL, and USER_LOGIN. Running ausearch -m without a message type

displays a list of all message types.

Search by Login ID

To view records associated with a particular login user ID, use the ausearch

-ul command. It displays any records related to the user login ID specied provided that user had been able to log in successfully.

Search by User ID

View records related to any of the user IDs (both user ID and effective user ID) with ausearch -ua. View reports related to a particular user ID with ausearch

-ui uid. Search for records related to a particular effective user ID, use the ausearch -ue euid. Searching for a user ID means the user ID of the user

creating a process. Searching for an effective user ID means the user ID and privileges that are required to run this process.

Search by Group ID

View records related to any of the group IDs (both group ID and effective group ID) with the ausearch -ga command. View reports related to a particular user ID with ausearch -gi gid. Search for records related to a particular effective group ID, use ausearch -ge egid.

Search by Command Line Name

View records related to a certain command, using the ausearch -c comm_name command, for example, ausearch -c less for all records related to the less command.

Understanding Linux Audit 29

Search by Executable Name

View records related to a certain executable with the ausearch -x exe command, for example ausearch -x /usr/bin/less for all records related to the /usr/bin/less executable.

Search by System Call Name

View records related to a certain system call with the ausearch -sc syscall command, for example, ausearch -sc open for all records related to the open system call.

Search by Process ID

View records related to a certain process ID with the ausearch -p pid command, for example ausearch -p 13368 for all records related to this process ID.

Search by Event or System Call Success Value

View records containing a certain system call success value with ausearch -sv success_value, for example, ausearch -sv yes for all successful system calls.

Search by Filename

View records containing a certain lename with ausearch -f filename, for example, ausearch -f /foo/bar for all records related to the /foo/ bar le. Using the lename alone would work as well, but using relative paths would not.

Search by Terminal

View records of events related to a certain terminal only with ausearch -tm term, for example, ausearch -tm ssh to view all records related to events on the SSH terminal and ausearch -tm tty to view all events

related to the console.

Search by Hostname

View records related to a certain remote hostname with ausearch -hn hostname, for example, ausearch -hn earth.example.com. You can use a hostname, fully qualied domain name, or numeric network address.

30 The Linux Audit Framework

Search by Key Field

View records that contain a certain key assigned in the audit rule set to identify events of a particular type. Use the ausearch -k key_field, for example, ausearch -k CFG_etc to display any records containing the CFG_etc key.

Limit a Search to a Certain Time Frame

Use -ts and -te to limit the scope of your searches to a certain time frame. The

-ts option is used to specify the start date and time and the -te option is used to specify the end date and time. These options can be combined with any of the above, except the -a option. The use of these options is similar to use with aureport.

1.7 Analyzing Processes with autrace

In addition to monitoring your system using the rules you set up, you can also perform dedicated audits of individual processes using the autrace command. autrace works similarly to the strace command, but gathers slightly different information. The output of autrace is written to /var/log/audit/audit.log and does not look any different from the standard audit log entries.

When performing an autrace on a process, make sure that any audit rules are purged from the queue to avoid these rules clashing with the ones autrace adds itself. Delete the audit rules with the auditctl -D command. This stops all normal auditing.

auditctl -D

No rules

autrace /usr/bin/less /etc/sysconfig/auditd

Waiting to execute: /usr/bin/less Cleaning up... No rules Trace complete. You can locate the records with 'ausearch -i -p 7642'

Always use the full path to the executable to track with autrace. After the trace is complete, autrace provides the event ID of the trace, so you can analyze the entire data trail with ausearch. To restore the audit system to use the audit rule set again, just restart the audit daemon with rcauditd restart.

Understanding Linux Audit 31

1.8 Visualizing Audit Data

Neither the data trail in /var/log/audit/audit.log nor the different report types generated by aureport, described in Section 1.5.2, “Generating Custom Audit

Reports” (page 20), provide an intuitive reading experience to the user. The aureport

output is formatted in columns and thus easily available to any sed, perl, or awk scripts that users might connect to the audit framework to visualize the audit data.

The visualization scripts (see Appendix A, Creating Flow Graphs from the Audit

Statistics (page 61) and Appendix B, Creating Bar Charts from the Audit Statistics

(page 65)) are one example of how to use standard Linux tools available with SUSE Linux Enterprise or any other Linux distribution to create easy-to-read audit output. The following examples help you understand how the plain audit reports can be transformed into human readable graphics.

The rst example illustrates the relationship of programs and system calls. To get to this kind of data, you need to determine the appropriate aureport command that delivers the source data from which to generate the nal graphic:

aureport -s

Syscall Report ======================================= # date time syscall pid comm auid event =======================================

1. 04/23/2007 08:04:08 PM 5 13374 file root 1900

2. 04/23/2007 08:04:08 PM 5 13376 file root 1901

3. 04/23/2007 08:04:08 PM 5 13368 less root 1902 ...

The rst thing that the visualization script needs to do on this report is to extract only those columns that are of interest, in this example, the syscall and the comm columns. The output is sorted and duplicates removed then the nal output is piped into the visualization program itself:

LC_ALL=C aureport -s -i | awk '/^[0-9]/ { printf "%s %s\n", $6, $4 }' | sort | uniq| ./mkgraph

NOTE: Adjusting the Locale

Depending on your choice of locale in /etc/sysconfig/auditd, your aureport output might contain an additional data column for AM/PM on time

32 The Linux Audit Framework

stamps. To avoid having this confuse your scripts, precede your script calls with LC_ALL=C to reset the locale and use the 24 hour time format.

Figure 1.2

Flow Graph—Program versus System Call Relationship

The second example illustrates the different types of events and how many of each type have been logged. The appropriate aureport command to extract this kind of information is aureport -e:

aureport -e -i --summary

Event Summary Report ====================== total type ====================== 2434 SYSCALL 816 USER_START 816 USER_ACCT 814 CRED_ACQ 810 LOGIN 806 CRED_DISP 779 USER_END 99 CONFIG_CHANGE 52 USER_LOGIN

Because this type of report already contains a two column output, it is just piped into the the visualization script and transformed into a bar chart.

aureport -e -i --summary | ./mkbar events

Understanding Linux Audit 33

Figure 1.3

Bar Chart—Common Event Types

For background information about the visualization of audit data, refer to the Web site of the audit project at http://people.redhat.com/sgrubb/audit/

visualize/index.html.

34 The Linux Audit Framework

Setting Up the Linux Audit Framework

This chapter shows how to set up a simple audit scenario. Every step involved in conguring and enabling audit is explained in detail. After you have learned to set up audit, consider a real-world example scenario in Chapter 3, Introducing an Audit Rule Set (page 47).

To set up audit on your SUSE Linux Enterprise, you need to complete the following steps:

Procedure 2.1

Setting Up the Linux Audit Framework

Make sure that all required packages are installed: audit, audit-libs, and optionally audit-libs-python. To use the log visualization as described in Section 2.7, “Conguring Log Visualization” (page 44), install gnuplot from the SUSE Linux Enterprise media and graphviz from the SUSE Linux Enterprise SDK.

Determine the components to audit. Refer to Section 2.1, “Determining the

Components to Audit ” (page 36) for details.

Check or modify the basic audit daemon conguration. Refer to Section 2.2,

“Conguring the Audit Daemon” (page 37) for details.

Enable auditing for system calls. Refer to Section 2.3, “Enabling Audit for System

Calls” (page 38) for details.

Compose audit rules to suit your scenario. Refer to Section 2.4, “Setting Up

Audit Rules” (page 39) for details.

Setting Up the Linux Audit Framework 35

Change your PAM conguration to enable audit IDs. Section 2.5, “Adjusting

the PAM Conguration” (page 40).

Generate logs and congure tailor-made reports. Refer to Section 2.6, “Cong-

uring Audit Reports” (page 41) for details.

Congure optional log visualization. Refer to Section 2.7, “Conguring Log

Visualization” (page 44) for details.

IMPORTANT: Controlling the Audit Daemon

Before conguring any of the components of the audit system, make sure that the audit daemon is not running by entering rcauditd status as root. On a default SUSE Linux Enterprise system, audit is started on boot, so you need to turn it off by entering rcauditd stop. Start the daemon after conguring it with rcauditd start.

2.1 Determining the Components to Audit

Before setting out to create your own audit conguration, determine to which degree you want to use it. Check the following rules of thumb to determine which use case best applies to you and your requirements:

• If you require a full security audit for CAPP/EAL certication, enable full audit for system calls and congure watches on various conguration les and directories, similar to the rule set featured in Chapter 3, Introducing an Audit Rule Set (page 47). Proceed to Section 2.3, “Enabling Audit for System Calls” (page 38).

• If you require an occasional audit of a system call instead of a permanent audit for system calls, use autrace. Proceed to Section 2.3, “Enabling Audit for System

Calls” (page 38).

• If you require le and directory watches to track access to important or securitysensitive data, create a rule set matching these requirements. Enable audit as described in Section 2.3, “Enabling Audit for System Calls” (page 38) and proceed to Section 2.4, “Setting Up Audit Rules” (page 39).

36 The Linux Audit Framework

2.2 Conguring the Audit Daemon

The basic setup of the audit daemon is done in /etc/auditd.conf:

The default settings work reasonably well for many setups. Some values, such as num_logs, max_log_file, space_left, and admin_space_left depend on the size of your deployment. If disk space is limited, you might want to reduce the number of log les to keep if they are rotated and you might want get an earlier warning if disk space is running out. For a CAPP-compliant setup, adjust the values for

log_file, flush, max_log_file, max_log_file_action, space_left, space_left_action, admin_space_left, admin_space_left_action, disk_full_action, and disk_error_action, as described in Section 1.2,

“Conguring the Audit Daemon” (page 5). An example CAPP-compliant conguration

looks like this:

log_file = path_to_separate_partition/audit.log log_format = RAW priority_boost = 3 flush = SYNC ### or DATA freq = 20 num_logs = 4 dispatcher = /usr/sbin/audispd disp_qos = lossy max_log_file = 5 max_log_file_action = KEEP_LOGS space_left = 75 space_left_action = EMAIL action_mail_acct = root admin_space_left = 50 admin_space_left_action = SINGLE ### or HALT

Setting Up the Linux Audit Framework 37

disk_full_action = SUSPEND ### or HALT disk_error_action = SUSPEND ### or HALT

The ### precedes comments where you can choose from several options. Do not add the comments to your actual conguration les.

TIP: For More Information

Refer to Section 1.2, “Conguring the Audit Daemon” (page 5) for detailed background information about the auditd.conf conguration parameters.

2.3 Enabling Audit for System Calls

A standard SUSE Linux Enterprise system has auditd running by default. There are different levels of auditing activity available:

Basic Logging

Out of the box without any further conguration, auditd logs only events concerning its own conguration changes to /var/log/audit/audit.log. No events (le access, system call, etc.) are generated by the kernel audit component until requested by auditctl. However, other kernel components and modules may log audit events outside of the control of auditctl and these appear in the audit log. By default, the only module that generates audit events is Novell AppArmor.

Advanced Logging with System Call Auditing

To audit system calls and get meaningful le watches, you need to enable audit contexts for system calls.

As you need system call auditing capabilities even when you are conguring plain le or directory watches, you need to enable audit contexts for system calls. To enable audit contexts for the duration of the current session only, execute auditctl -e 1 as root. To disable this feature, execute auditctl -e 0 as root.

To enable audit contexts for system calls permanently, open the /etc/sysconfig/ auditd conguration le as root and set AUDITD_DISABLE_CONTEXTS to no. Then restart the audit daemon with the rcauditd restart command. To turn this feature off temporarily, use auditctl -e 0. To turn it off permanently, set AUDITD_DISABLE_CONTEXTS to yes.

38 The Linux Audit Framework

2.4 Setting Up Audit Rules

Using audit rules, determine which aspects of the system should be analyzed by audit. Normally this includes important databases and security-relevant conguration les. You may also analyze various system calls in detail if a broad analysis of your system is required. A very detailed example conguration that includes most of the rules that are needed in a CAPP compliant environment is available in Chapter 3, Introducing an

Audit Rule Set (page 47).

A simple rule set for very basic auditing on a few important les and directories could look like this:

# basic audit system parameters

-D

-b 8192

-f 1

-e 1

# some file and directory watches with keys

-w /var/log/audit/ -k LOG_audit

-w /etc/auditd.conf -k CFG_audit_conf -p rxwa

-w /etc/audit.rules -k CFG_audit_rules -p rxwa

-w /etc/passwd -k CFG_passwd -p rwxa

-w /etc/sysconfig/ -k CFG_sysconfig

# an example system call rule

-a entry,always -S umask

### add your own rules

When conguring the basic audit system parameters, such as the backlog parameter

-b, test these settings with your intended audit rule set to determine whether the backlog size is appropriate for the level of logging activity caused by your audit rule set. If your chosen backlog size is too small, your system might not be able to handle the audit load and consult the failure ag (-f) when the backlog limit is exceeded.

IMPORTANT: Choosing the Failure Flag

When choosing the failure ag, note that -f 2 tells your system to perform an immediate shutdown without ushing any pending data to disk when the limits of your audit system are exceeded. Because this shutdown is not a clean shutdown, restrict the use of -f 2 to only the most security conscious environ-

Setting Up the Linux Audit Framework 39

ments and use -f 1 (system continues to run, issues a warning and audit stops) for any other setup to avoid loss of data or data corruption.

Directory watches produce less verbose output than separate le watches for the les under these directories. To get detailed logging for your system conguration in /etc/ sysconfig, for example, add watches for each individual le. Audit does not support globbing, which means you cannot just create a rule that says -w /etc/* and watches anything below /etc.

For better identication in the log le, a key has been added to each of the le and directory watches. Using the key, it is easier to comb the logs for events related to a certain rule. When creating keys, distinguish between mere log le watches and conguration le watches by using an appropriate prex with the key, in this case LOG for a log le watch and CFG for a conguration le watch. Using the lename as part of the key also makes it easier for you to identify events of this type in the log le.

Another thing to bear in mind when creating le and directory watches is that audit cannot deal with les that do not when the rules are created. Any le that is added to your system while audit is already running are not watched unless you extend the rule set to watch this new le.

For more information about creating custom rules, refer to Section 1.4, “Passing Param-

eters to the Audit System” (page 11).

IMPORTANT: Changing Audit Rules

Never change audit rules in a running audit system. Always stop the audit daemon with rcauditd stop before touching the audit conguration and reread the audit conguration by restarting the daemon with rcauditd start.

2.5 Adjusting the PAM Conguration

Audit allows you to consistently track a user's actions from login right through logout no matter which identities this user might adopt in the meantime. Audit does so by using audit IDs that are created upon login and handed down to any child process of the original login process. To enable this special feature of audit, you need to modify the PAM conguration of several components (login, sshd, gdm, crond, atd).

40 The Linux Audit Framework

To adjust the PAM conguration to enable audit IDs, proceed as follows:

Open the PAM conguration le for the application—/etc/pam .d/application— and add the following line before the common-session

line:

session required pam_loginuid.so session include common-session

Apply your changes and close the conguration le.

The changes in PAM conguration take effect as soon as the application is called again, for example, login, sshd, and the display managers log with an audit ID at the next login.

2.6 Conguring Audit Reports

To avoid having to dig through the raw audit logs to get an impression of what your system is currently doing, run custom audit reports at certain intervals. Custom audit reports enable you to focus on areas of interest and get meaningful statistics on the nature and frequency of the events you are monitoring. To analyze individual events in detail, use the ausearch tool.

Before setting up audit reporting, consider the following:

• What types of events do you want to monitor by generating regular reports? Select the appropriate aureport command lines as described in Section 1.5.2, “Generating

Custom Audit Reports” (page 20).

• What do you want to do with the audit reports? Decide whether to create graphical charts from the data accumulated or whether it should be transferred into any sort of spreadsheet or database. Set up the aureport command line and further processing similar to the examples shown in Section 2.7, “Conguring Log Visualization” (page 44) if you want to visualize your reports.

• When and at which intervals should the reports run? Set up appropriate automated reporting using cron.

Setting Up the Linux Audit Framework 41

For this example, assume that you are interested in nding out about any attempts to access your audit, PAM, and system conguration. Proceed as follows to nd out about le events on your system:

Generate a full summary report of all events and check for any anomalies in the summary report, for example, have a look at the “failed syscalls” record, because these might have failed due to insufcient permissions to access a le or a le not being there at all:

aureport

Summary Report ====================== Range of time: 04/19/2007 13:42:43.280 - 05/02/2007 14:05:49.302 Number of changes in configuration: 215 Number of changes to accounts, groups, or roles: 0 Number of logins: 47 Number of failed logins: 16 Number of users: 5 Number of terminals: 17 Number of host names: 7 Number of executables: 16 Number of files: 173 Number of AVC denials: 0 Number of MAC events: 0 Number of failed syscalls: 799 Number of anomaly events: 0 Number of responses to anomaly events: 0 Number of crypto events: 0 Number of process IDs: 2382 Number of events: 14424

Run a summary report for failed events and check the “les” record for the number of failed le access events:

aureport --failed

Failed Summary Report ====================== Range of time: 04/19/2007 13:42:43.280 - 05/02/2007 14:07:41.609 Number of changes in configuration: 0 Number of changes to accounts, groups, or roles: 0 Number of logins: 0 Number of failed logins: 16 Number of users: 1 Number of terminals: 7 Number of host names: 4 Number of executables: 7 Number of files: 102 Number of AVC denials: 0

42 The Linux Audit Framework

Number of MAC events: 0 Number of failed syscalls: 799 Number of anomaly events: 0 Number of responses to anomaly events: 0 Number of crypto events: 0 Number of process IDs: 40 Number of events: 841

To list the les that could not be accessed, run a summary report of failed le events:

aureport -f -i --failed --summary

Failed File Summary Report =========================== total file =========================== 21 /usr/lib/locale/locale-archive 13 /usr/lib/locale/en_US.UTF-8/LC_IDENTIFICATION 13 /usr/lib/locale/en_US.UTF-8/LC_MEASUREMENT 13 /usr/lib/locale/en_US.UTF-8/LC_TELEPHONE 13 /usr/lib/locale/en_US.UTF-8/LC_ADDRESS 13 /usr/lib/locale/en_US.UTF-8/LC_NAME 13 /usr/lib/locale/en_US.UTF-8/LC_PAPER 13 /usr/lib/locale/en_US.UTF-8/LC_MESSAGES 13 /usr/lib/locale/en_US.UTF-8/LC_MONETARY 13 /usr/lib/locale/en_US.UTF-8/LC_COLLATE 13 /usr/lib/locale/en_US.UTF-8/LC_TIME 13 /usr/lib/locale/en_US.UTF-8/LC_NUMERIC 13 /usr/lib/locale/en_US.UTF-8/LC_CTYPE 13 /etc/ld.so.preload ...

To focus this summary report on a few les or directories of interest only, such as /etc/auditd.conf, /etc/pam.d, and /etc/sysconfig, use a command similar to the following:

aureport -f -i --failed --summary |grep -e "/etc/auditd.conf" -e "/etc/pam.d/" -e "/etc/sysconfig"

1 /etc/sysconfig/displaymanager

From the summary report, then proceed to isolate these items of interest from the log and nd out their event IDs for further analysis:

aureport -f -i --failed |grep -e "/etc/auditd.conf" -e "/etc/pam.d/" -e "/etc/sysconfig"

Setting Up the Linux Audit Framework 43

68. 04/27/2007 04:45:00 PM /etc/sysconfig/displaymanager getxattr no /bin/vim root 1451

Use the event ID to get a detailed record for each item of interest:

ausearch -a 1451 -i

type=PATH msg=audit(04/27/2007 16:45:00.736:1451) : item=0 name=/etc/sysconfig/displaymanager inode=440645 dev=03:01 mode=file,644 ouid=root ogid=root rdev=00:00 type=CWD msg=audit(04/27/2007 16:45:00.736:1451) : cwd=/root type=SYSCALL msg=audit(04/27/2007 16:45:00.736:1451) : arch=i386 syscall=getxattr success=no exit=-61(No data available) a0=8175bf8 a1=b7f00c4f a2=bf88d950 a3=84 items=1 ppid=7591 pid=7593 auid=root uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=pts0 comm=vi exe=/bin/vim subj=unconstrained key=(null)

TIP: Focusing on a Certain Time Frame

If you are interested in events during a particular period of time, trim down the reports by using start and end dates and times with your aureport commands (-ts and -te). For more information, refer to Section 1.5.2, “Generating

Custom Audit Reports” (page 20).

All steps except for the last one can be run automatically and would easily be scriptable and congured as cron jobs. Any of the --failed --summary reports could be transformed easily into a bar chart that plots les versus failed access attempts. For more information about visualizing audit report data, refer to Section 2.7, “Conguring

Log Visualization” (page 44).

2.7 Conguring Log Visualization

Using the scripts in Appendix B, Creating Bar Charts from the Audit Statistics (page 65) and Appendix A, Creating Flow Graphs from the Audit Statistics (page 61), you can illustrate your audit statistics with various graphs and charts. Follow the installation instructions for the visualization scripts and determine which plots you need to run. As with any other aureport command, the plotting commands are scriptable and can easily be congured to run as cron jobs.

44 The Linux Audit Framework

To plot summary reports, such as the ones discussed in Section 2.6, “Conguring Audit

Reports” (page 41), use the script introduced in Appendix B, Creating Bar Charts from

the Audit Statistics (page 65). Some example commands could look like the following:

Create a Summary of Events

aureport -e -i --summary | ./mkbar events

Create a Summary of File Events

aureport -f -i --summary | ./mkbar files

Create a Summary of Login Events

aureport -l -i --summary | ./mkbar login

Create a Summary of User Events

aureport -u -i --summary | ./mkbar users

Create a Summary of System Call Events

aureport -s -i --summary | ./mkbar syscalls

To create a summary chart of failed events of any of the above event types, just add the --failed option to the respective aureport command. To cover a certain period of time only, use the -ts and -te options on aureport. Any of these commands can be tweaked further by narrowing down its scope using grep or egrep and regular expressions. See the comments in Appendix B, Creating Bar Charts from the Audit

Statistics (page 65) for an example. Any of the above commands produces a PNG le

containing a bar chart of the requested data.

To illustrate the relationship between different kinds of audit objects, such as users and system calls, use the script introduced in Appendix A, Creating Flow Graphs from the

Audit Statistics (page 61). Some example commands could look like the following:

Users versus Executables

LC_ALL=C aureport -u -i | awk '/^[0-9]/ { printf "%s %s\n",

Setting Up the Linux Audit Framework 45

$4, $7 }' | sort | uniq | ./mkgraph

Users versus Files

LC_ALL=C aureport -f -i | awk '/^[0-9]/ { printf "%s %s\n",

$8, $4 }' | sort | uniq | ./mkgraph

System Calls versus Commands

LC_ALL=C aureport -s -i | awk '/^[0-9]/ { printf "%s %s\n",

$4, $6 }' | sort | uniq | ./mkgraph

System Calls versus Files

LC_ALL=C aureport -s -i | awk '/^[0-9]/ { printf "%s %s\n",

$5, $4 }' | sort | uniq | ./mkgraph

Graphs can also be combined to illustrate complex relationships. See the comments in

Appendix A, Creating Flow Graphs from the Audit Statistics (page 61) for further in-

formation and an example. The graphs produced by this script are created in PostScript format by default, but you can change the output format by changing the EXT variable from ps to png or jpg. To cover a certain period of time only, use the -ts and -te options on aureport.

46 The Linux Audit Framework

Introducing an Audit Rule Set

The following example conguration illustrates how audit can be used to monitor your system. It highlights the most important items that need to be audited to cover the list of auditable events specied by Controlled Access Protection Prole (CAPP).

The example rule set is divided into the following sections:

• Basic audit conguration (see Section 3.1, “Adding Basic Audit Conguration

Parameters” (page 48))

• Watches on audit log les and conguration les (see Section 3.2, “Adding

Watches on Audit Log Files and Conguration Files” (page 49))

• Monitoring operations on le system objects (see Section 3.3, “Monitoring File

System Objects” (page 50))

• Monitoring security databases (see Section 3.4, “Monitoring Security Conguration

Files and Databases” (page 51))

• Monitoring miscellaneous system calls (Section 3.5, “Monitoring Miscellaneous

System Calls” (page 54))

• Filtering system call arguments (see Section 3.6, “Filtering System Call Arguments” (page 54))

To transform this example into a conguration le to use in your live setup, proceed as follows:

Choose the appropriate settings for your setup and adjust them.

Introducing an Audit Rule Set 47

Save all examples to one single le called /etc/audit.rules.

Adjust the le permissions to make sure that the rules le is owned by root and by the root group and its permissions are set to read-write for root and read-only for the group and access is denied to any other user or group.

NOTE: Adjusting the Level of Audit Logging

Do not copy the example below into your audit setup without adjusting it to your needs. Determine what and to what extent to audit.

The entire audit.rules is just a collection of auditctl commands. Every line in this le expands to a full auditctl command line. The syntax used in the rule set is the same as that of the auditctl command.

3.1 Adding Basic Audit Conguration Parameters

-D❶

-b 8192❷

-f 2❸

❶

Delete any preexisting rules before starting to dene new ones.

❷

Set the number of buffers to take the audit messages. Depending on the level of audit logging on your system, increase or decrease this gure.

❸

Set the failure ag to use when the kernel needs to handle critical errors. Possible values are 0 (silent), 1 (printk, print a failure message), and 2 (panic, halt the system).

By emptying the rule queue with the -D option, you make sure that audit does not use any other rule set than what you are offering it by means of this le. Choosing an appropriate buffer number (-b) is vital to avoid having your system fail because of too high an audit load. Choosing the panic failure ag -f 2 ensures that your audit records are complete even if the system is encountering critical errors. By shutting down the system on a critical error, audit makes sure that no process escapes from its control as it otherwise might if level 1 (printk) were chosen.

48 The Linux Audit Framework

IMPORTANT: Choosing the Failure Flag

Before using your audit rule set on a live system, make sure that the setup has been thoroughly evaluated on test systems using the worst case production workload. It is even more critical that you do this when specifying the -f 2 ag, because this instructs the kernel to panic (perform an immediate halt without ushing pending data to disk) if any thresholds are exceeded. Consider the use of the -f 2 ag for only the most security-conscious environments.

3.2 Adding Watches on Audit Log Files and Conguration Files

Adding watches on your audit conguration les and the log les themselves ensures that you can track any attempt to tamper with the conguration les or detect any attempted accesses to the log les.

NOTE: Creating Directory and File Watches

Creating watches on a directory is not necessarily sufcient if you need events for le access. Events on directory access are only triggered when the directory's inode is updated with metadata changes. To trigger events on le access, add watches for each individual le to monitor.

-w /var/log/audit/ ❶

-w /var/log/audit/audit.log

#-w /var/log/audit/audit_log.1 #-w /var/log/audit/audit_log.2 #-w /var/log/audit/audit_log.3 #-w /var/log/audit/audit_log.4

-w /etc/auditd.conf -p wa❷

-w /etc/audit.rules -p wa

-w /etc/libaudit.conf -p wa

-w /etc/sysconfig/auditd -p wa

Introducing an Audit Rule Set 49

❶

Set a watch on the directory where the audit log is located. Trigger an event for any type of access attempt to this directory. If you are using log rotation, add watches for the rotated logs as well.

❷

Set a watch on an audit conguration le. Log all write and attribute change attempts to this le.

3.3 Monitoring File System Objects

Auditing system calls helps track your system's activity well beyond the application level. By tracking le system–related system calls, get an idea of how your applications are using these system calls and determine whether that use is appropriate. By tracking mount and umount operations, track the use of external resources (removable media, remote le systems, etc.).

IMPORTANT: Auditing System Calls

Auditing system calls results in a high logging activity. This activity, in turn, puts a heavy load on the kernel. With a kernel less responsive than usual, the system's backlog and rate limits might be exceeded. Carefully evaluate which system calls to include in your audit rule set and adjust the log settings accordingly. See Section 1.2, “Conguring the Audit Daemon” (page 5) for details on how to tweak the relevant settings.

-a entry,always -S chmod -S fchmod -S chown -S chown32 -S fchown -S fchown32

-S lchown -S lchown32❶

-a entry,always -S creat -S open -S truncate -S truncate64 -S ftruncate -S ftruncate64❷

-a entry,always -S mkdir -S rmdir❸

-a entry,always -S unlink -S rename -S link -S symlink❹

-a entry,always -S setxattr❺

-a entry,always -S lsetxattr

-a entry,always -S fsetxattr

-a entry,always -S removexattr

-a entry,always -S lremovexattr

-a entry,always -S fremovexattr

-a entry,always -S mknod❻

50 The Linux Audit Framework

-a entry,always -S mount -S umount -S umount2❼

❶

Enable an audit context for system calls related to changing le ownership and permissions. Depending on the hardware architecture of your system, enable or disable the *32 rules. 64-bit systems, like x86_64 and ia64, require the *32 rules to be removed.

❷

Enable an audit context for system calls related to le content modication. Depending on the hardware architecture of your system, enable or disable the *64 rules. 64-bit systems, like x86_64 and ia64, require the *64 rules to be removed.

❸

Enable an audit context for any directory operation, like creating or removing a directory.

❹

Enable an audit context for any linking operation, such as symlink, link, unlink, or rename.

❺

Enable an audit context for any operation related to extended le system attributes.

❻

Enable an audit context for the mknod system call, which creates special (device) les.

❼

Enable an audit context for any mount or umount operation. For the x64_64 architecture, disable the umount rule. For the ia64 architecture, disable the umount2 rule.

3.4 Monitoring Security Conguration Files and Databases

To make sure that your system is not made to do undesired things, track any attempts to change the cron and at congurations or the lists of scheduled jobs. Tracking any write access to the user, group, password and login databases and logs helps you identify any attempts to manipulate your system's user database.

Tracking changes to your system conguration (kernel, services, time, etc.) helps you spot any attempts of others to manipulate essential functionality of your system. Changes to the PAM conguration should also be monitored in a secure environment, because changes in the authentication stack should not be made by anyone other than the administrator and it should be logged which applications are using PAM and how it is used.

Introducing an Audit Rule Set 51

The same applies to any other conguration les related to secure authentication and communication.

❶

-w /var/spool/at

-w /etc/at.allow

-w /etc/at.deny

-w /etc/cron.allow -p wa

-w /etc/cron.deny -p wa

-w /etc/cron.d/ -p wa

-w /etc/cron.daily/ -p wa

-w /etc/cron.hourly/ -p wa

-w /etc/cron.monthly/ -p wa

-w /etc/cron.weekly/ -p wa

-w /etc/crontab -p wa

-w /var/spool/cron/root

❷

-w /etc/group -p wa

-w /etc/passwd -p wa

-w /etc/shadow

-w /etc/login.defs -p wa

-w /etc/securetty

-w /var/log/faillog

-w /var/log/lastlog

❸

-w /etc/hosts -p wa

-w /etc/sysconfig/

-w /etc/inittab -p wa

-w /etc/init.d/

-w /etc/init.d/auditd -p wa

-w /etc/ld.so.conf -p wa

-w /etc/localtime -p wa

-w /etc/sysctl.conf -p wa

-w /etc/modprobe.conf.d/

-w /etc/modprobe.conf.local -p wa

-w /etc/modprobe.conf -p wa

❹

-w /etc/pam.d/

❺

-w /etc/aliases -p wa

52 The Linux Audit Framework

-w /etc/postfix/ -p wa

❻

-w /etc/ssh/sshd_config

-w /etc/stunnel/stunnel.conf

-w /etc/stunnel/stunnel.pem

-w /etc/vsftpd.ftpusers

-w /etc/vsftpd/vsftpd.conf

❼

-a exit,always -S sethostname

-w /etc/issue -p wa

-w /etc/issue.net -p wa

❶

Set watches on the at and cron conguration and the scheduled jobs and assign labels to these events.

❷

Set watches on the user, group, password, and login databases and logs and set labels to better identify any login-related events, such as failed login attempts.

❸

Set a watch and a label on the static hostname conguration in /etc/hosts. Track changes to the system conguration directory, /etc/sysconfig. Enable per-le watches if you are interested in le events. Set watches and labels for changes to the boot conguration in /etc/inittab and the /etc/init.d directory. Enable per-le watches if you are interested in le events. Set watches and labels for any changes to the linker conguration in /etc/ld.so.conf. Set watches and a label for /etc/localtime. Set watches and labels for the kernel conguration les /etc/sysctl.conf, /etc/modprobe.d/, /etc/ modprobe.conf.local, and /etc/modprobe.conf.

❹

Set watches on the PAM conguration directory. If you are interested in particular les below the directory level, add explicit watches to these les as well.

❺

Set watches to the postx conguration to log any write attempt or attribute change and use labels for better tracking in the logs.

❻

Set watches and labels on the ssh, stunnel, and vsftpd conguration les.

❼

Perform an audit of the sethostname system call and set watches and labels on the system identication conguration in /etc/issue and /etc/issue .net.

Introducing an Audit Rule Set 53

3.5 Monitoring Miscellaneous System Calls

As well as auditing le system related system calls, as described in Section 3.3,

“Monitoring File System Objects” (page 50), you can also track various other system

calls. Tracking task creation helps you understand your applications' behavior. Auditing the umask system call lets you track how processes modify permissions. Tracking any attempts to change the system time helps you identify anyone or any process trying to manipulate the system time.

❶

-a entry,always -S clone -S fork -S vfork

## For ia64 architecture, disable fork and vfork rules above, and ## enable the following: #-a entry,always -S clone2

❷

-a entry,always -S umask

❸

-a entry,always -S adjtimex -S settimeofday

❶

Track task creation. To enable task tracking on the ia64 architecture, comment the rst rule and enable the second one.

❷

Add an audit context to the umask system call.

❸

Track attempts to change the system time. adjtimex can be used to skew the time. settimeofday sets the absolute time.

3.6 Filtering System Call Arguments

In addition to the system call auditing introduced in Section 3.3, “Monitoring File

System Objects” (page 50) and Section 3.5, “Monitoring Miscellaneous System Calls”

(page 54), you can track application behavior to an even higher degree. Applying lters helps you focus audit on areas of primary interest to you. This section introduces ltering system call arguments for nonmultiplexed system calls like access and for multiplexed ones like socketcall or ipc. Whether system calls are multiplexed depends on the hardware architecture used. Both socketcall and ipc are not multiplexed on 64-bit architectures, such as x86_64 and ia64.

54 The Linux Audit Framework

IMPORTANT: Auditing System Calls

Auditing system calls results in a high logging activity, which in turn puts a heavy load on the kernel. With a kernel less responsive than usual, the system's backlog and rate limits might well be exceeded. Carefully evaluate which system calls to include in your audit rule set and adjust the log settings accordingly. See Section 1.2, “Conguring the Audit Daemon” (page 5) for details on how to tweak the relevant settings.

The access system call checks whether a process would be allowed to read, write or test for the existence of a le or le system object. Using the -F lter ag, build rules matching specic access calls in the format-F a1=access_mode. Check /usr/

include/fcntl.h for a list of possible arguments to the access system call.

-a entry,always -S access -F a1=4❶

-a entry,always -S access -F a1=6❷

-a entry,always -S access -F a1=7❸

❶

Audit the access system call, but only if the second argument of the system call (mode) is 4 (R_OK). This rule lters for all access calls testing for sufcient write permissions to a le or le system object accessed by a user or process.

❷

Audit the access system call, but only if the second argument of the system call (mode) is 6, meaning 4 OR 2, which translates to R_OK OR W_OK. This rule lters for access calls testing for sufcient read and write permissions.

❸

Audit the access system call, but only if the second argument of the system call (mode) is 7, meaning 4 OR 2 OR 1, which translates to R_OK OR W_OK OR X_OK. This rule lters for access calls testing for sufcient read, write, and execute permissions.

The socketcall system call is a multiplexed system call. Multiplexed means that there is only one system call for all possible calls and that libc passes the actual system call to use as the rst argument (a0). Check the manual page of socketcall for possible system calls and refer to /usr/include/linux/net.h for a list of possible argument values and system call names. Audit supports ltering for specic system calls using a -F a0=syscall_number.

-a entry, always -S socketcall -F a0=1 -F a1=10❶

-a entry, always -S socketcall -F a0=5❷

Introducing an Audit Rule Set 55

❶

Audit the socket(PF_INET6) system call. The -F a0=1 lter matches all socket system calls and the -F a1=10 lter narrows the matches down to socket system calls carrying the IPv6 protocol family domain parameter (PF_INET6). Check

/usr/include/linux/net.h for the rst argument (a0) and /usr/ include/linux/socket.h for the second parameter (a1).

❷

Audit the socketcall system call. The lter ag is set to lter for a0=5 as the rst argument to socketcall, which translates to the accept system call if you check /usr/include/linux/net.h.

The ipc system call is another example of multiplexed system calls. The actual call to invoke is determined by the rst argument passed to the ipc system call. Filtering for these arguments helps you focus on those IPC calls of interest to you. Check /usr/ include/asm-generic/ipc.h for possible argument values.

❶

## msgctl

-a entry,always -S ipc -F a0=14

## msgget

-a entry,always -S ipc -F a0=13

## Enable if you are interested in these events (x86_64, ia64) #-a entry,always -S msgctl #-a entry,always -S msgget

❷

## semctl

-a entry,always -S ipc -F a0=3

## semget

-a entry,always -S ipc -F a0=2

## semop

-a entry,always -S ipc -F a0=1

## semtimedop

-a entry,always -S ipc -F a0=4

## Enable if you are interested in these events (x86_64, ia64) #-a entry,always -S semctl #-a entry,always -S semget #-a entry,always -S semop #-a entry,always -S semtimedop

❸

## shmctl

-a entry,always -S ipc -F a0=24

## shmget

-a entry,always -S ipc -F a0=23

## Enable if you are interested in these events (x86_64, ia64) #-a entry,always -S shmctl #-a entry,always -S shmget

56 The Linux Audit Framework

❶

Audit system calls related to IPC SYSV message queues. In this case, the a0 values specify that auditing is added for the msgctl and msgget system calls (14 and 13). 64-bit platforms, like x86_64 and ia64, do not use multiplexing on ipc system calls. For these platforms, comment the rst two rules and add the plain system call rules without argument ltering.

❷

Audit system calls related to IPC SYSV message semaphores. In this case, the a0 values specify that auditing is added for the semctl, semget, semop, and semtimedop system calls (3, 2, 1, and 4). 64-bit platforms, like x86_64 and ia64, do not use multiplexing on ipc system calls. For these platforms, comment the rst four rules and add the plain system call rules without argument ltering.

❸

Audit system calls related to IPC SYSV shared memory. In this case, the a0 values specify that auditing is added for the shmctl and shmget system calls (24, 23). 64-bit platforms, like x86_64 and ia64, do not use multiplexing on ipc system calls. For these platforms, comment the rst two rules and add the plain system call rules without argument ltering.

3.7 Managing Audit Event Records Using Keys

After conguring a few rules generating events and populating the logs, you need to nd a way to tell one event from the others. Using the ausearch command, you can lter the logs for various criteria. Using ausearch -m message_type, you can at least lter for events of a certain type. However, to be able to lter for events related to a particular rule, you need to add a key to this rule in the /etc/audit.rules le. This key is then added to the event record every time the rule logs an event. To retrieve these log entries, simply run ausearch -k your_key to get a list of records related to the rule carrying this particular key.

As an example, assume you have added the following rule to your rule le:

-w /etc/audit.rules -p wa

Without a key assigned to it, you would probably have to lter for SYSCALL or PATH events then use grep or similar tools to isolate any events related to the above rule. Now, add a key to the above rule, using the -k option:

Introducing an Audit Rule Set 57

-w /etc/audit.rules -p wa -k CFG_audit.rules

You can specify any text string as key. Distinguish watches related to different types of les (conguration les or log les) from one another using different key prexes (CFG, LOG, etc.) followed by the lename. Finding any records related to the above rule now comes down to the following:

ausearch -k CFG_audit.rules

time->Thu Apr 26 14:56:25 2007 type=PATH msg=audit(1177592185.922:52): item=0 name="/etc/audit.rules" inode=444083 dev=03:01 mode=0100644 ouid=0 ogid=0 rdev=00:00 type=CWD msg=audit(1177592185.922:52): cwd="/root" type=SYSCALL msg=audit(1177592185.922:52): arch=40000003 syscall=226 success=yes exit=0 a0=8175be8 a1=b7e8dc4f a2=81b6de8 a3=1c items=1 ppid=3765 pid=3839 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=pts0 comm="vi" exe="/bin/vim" subj=unconstrained key="CFG_audit.rules"

58 The Linux Audit Framework

Useful Resources

There are other resources available containing valuable information about the Linux audit framework:

The Audit Manual Pages

There are several man pages installed along with the audit tools that provide valuable and very detailed information:

• auditd(8)

• auditd.conf(8)

• auditctl(8)

• autrace(8)

• ausearch(8)

• aureport(8)

http://people.redhat.com/sgrubb/audit/index.html

The home page of the Linux audit project. This site contains several specications relating to different aspects of Linux audit as well as a short FAQ.

/usr/share/doc/packages/audit

The audit package itself contains a README with basic design information along with some Red Hat–specic instructions and a sample.rules le demonstrating the basic capabilities of audit.

Useful Resources 59

http://www.commoncriteriaportal.org/

The ofcial Web site of the Common Criteria project. Learn all about the Common Criteria security certication initiative and which role audit plays in this framework.

60 The Linux Audit Framework

Creating Flow Graphs from the Audit Statistics

The following script to convert aureport audit statistics into ow graphs was created by Steve Grubb at Red Hat. It is available from http://people.redhat.com/

sgrubb/audit/visualize/mkgraph.

Because the current version of audit in SUSE Linux Enterprise does not ship with this script, proceed as follows to make it available on your system:

Download the script from http://people.redhat.com/sgrubb/

audit/visualize/mkgraph or copy the text below into a le called

mkgraph. Each line containing commented aureport commands is meant to

be written on one continuous line. Adjust this if you use the copy method.

Move the mkgraph le to root's home directory.

Adjust the le permissions to read, write, and execute for root.

#!/bin/sh # # Copyright 2005 Red Hat Inc., Durham, North Carolina. # All Rights Reserved. # This software may be freely redistributed and/or modified under the # terms of the GNU General Public License as published by the Free # Software Foundation; either version 2, or (at your option) any # later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details.

# # You should have received a copy of the GNU General Public License # along with this program; see the file COPYING. If not, write to the # Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. # # Author: Steve Grubb <sgrubb@redhat.com> # ######## # # This program will take stdin and produce a graph for it. The input # should be 2 objects per line separated by a space. # Some interesting uses: # # See what syscalls a program makes # aureport -s -i | awk '/^[0-9]/ { printf "%s %s\n", $6, $4 }' | sort | uniq | ./mkgraph # # See avc denied subject to object map # aureport -a --failed -i | awk '/^[0-9]/ { printf "%s %s\n", $5, $8 }' | sort

| uniq | ./mkgraph # # See who is accessing files #aureport -f -i | awk '/^[0-9]/ { printf "%s %s\n", $8, $4 }' | sort | uniq | ./mkgraph # # See what account is running which exes # aureport -u -i | awk '/^[0-9]/ { printf "%s %s\n", $4, $7 }' | sort | uniq | ./mkgraph # # See what accounts are being used by remote hosts #aureport -h -i | awk '/^[0-9]/ { printf "%s %s\n", $4, $6 }' | sort | uniq | ./mkgraph # # Graphs can be combined, too. For example, to see what host people logged in # from and the commands they ran: #aureport -h -i | awk '/^[0-9]/ { printf "%s %s\n", $4, $6 }' | sort | uniq > tmp.rpt #aureport -u -i | awk '/^[0-9]/ { printf "%s %s\n", $4, $7 }' | sort | uniq >> tmp.rpt #cat tmp.rpt | ./mkgraph

if [ x"$1" != "x" ] ; then OUT="$1" else OUT="gr" fi DOT_CMD=`which dot 2>/dev/null` DOT_FILE="./$OUT.dot" IDX_FILE="./$OUT.index" # use png, ps, or jpg EXT="ps" if [ x"$DOT_CMD" = "x" ] ; then

62 The Linux Audit Framework

echo "graphviz is not installed. Exiting." exit 1 fi echo "digraph G {" > $DOT_FILE # Some options you may want to set #echo -e "\torientation=landscape" >> $DOT_FILE #echo -e "\tsize=\"60,18\"" >> $DOT_FILE #echo -e "\tranksep=\"1.25\"" >> $DOT_FILE #echo -e "\tratio=fill" >> $DOT_FILE #echo -e "\tpage=\"8.5,11\";" >> $DOT_FILE

while [ 1 ] do read -t 5 line 2>/dev/null if [ $? -ne 0 ] ; then break fi if [ x"$line" != "x" ] ; then echo $line | awk '{ printf("\t\"%s\" -> \"%s\";\n", $1, $2); }' >> $DOT_FILE

fi done echo "}" >> $DOT_FILE echo " " >> $DOT_FILE

$DOT_CMD -T$EXT -o ./$OUT.$EXT $DOT_FILE 1>&2 2>/dev/null if [ $? -ne 0 ] ; then echo "Error rendering" rm -f $DOT_FILE exit 1 fi rm -f $DOT_FILE if [ "$EXT" = "ps" ] ; then echo "Gzipping graph..." rm -f ./$OUT.ps.gz 2>/dev/null gzip --best ./$OUT.ps echo "Graph was written to $OUT.$EXT.gz" else echo "Graph was written to $OUT.$EXT" fi exit 0

Creating Flow Graphs from the Audit Statistics 63

Creating Bar Charts from the Audit Statistics

The following script to convert aureport audit statistics into bar charts was created by Steve Grubb at Red Hat. It is available from http://people.redhat.com/

sgrubb/audit/visualize/mkbar.

Because the current version of audit in SUSE Linux Enterprise does not ship with this script, proceed as follows to make it available on your system:

Download the script from http://people.redhat.com/sgrubb/

audit/visualize/mkbar or copy the text below into a le called mkbar.

Each line ending in \ is meant to be written on one continuous line. Adjust this by removing the trailing \ and merging the two lines if you use the copy method.

Move the mkbar le to root's home directory.

Adjust the le permissions to read, write, and execute for root.

#!/bin/sh # Copyright 2005 Red Hat Inc., Durham, North Carolina. # All Rights Reserved. # This software may be freely redistributed and/or modified under the # terms of the GNU General Public License as published by the Free # Software Foundation; either version 2, or (at your option) any # later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License

# along with this program; see the file COPYING. If not, write to the # Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. # # Author: Steve Grubb <sgrubb@redhat.com> # ####### # This program will take data returned by aureport suumaries and # produce a bar chart of it. You can optionally pass a parameter # that names the file to create # # To see most often failed file access: # aureport -f -i --summary --failed | mkbar failed-access # # To see syscalls: # aureport -s -i --summary | ./mkbar syscall # # To see events: # aureport -e -i --summary | ./mkbar events # # To see all events except syscall & config change: # aureport -e -i --summary | egrep -vi '(syscall|change)' | ./mkbar events2 ###

if [ x"$1" != "x" ] ; then OUT="$1" else OUT="chart" fi

EXT="png" gpcommand="plot-script" gpdata="$OUT.dat" gpout="$OUT.$EXT" plotcommand=`which gnuplot`

if [ x"$plotcommand" = "x" ] ; then echo "gnuplot is not installed" exit 1 fi

# create gnuplot command file echo "set terminal $EXT small xfdf5e6 x000000 x404040 x0000ff x00ff00" > \ $gpcommand echo "set grid ytics" >> $gpcommand echo "set nokey" >> $gpcommand echo "set nolabel" >> $gpcommand echo "set data style lines" >> $gpcommand echo "set noxzeroaxis" >> $gpcommand echo "set noyzeroaxis" >> $gpcommand echo "set boxwidth 0.9 relative" >> $gpcommand echo "set style fill solid 1.0" >> $gpcommand echo 'set output "'$gpout'"' >> $gpcommand # This is to be able to start with a comma as we read input.

66 The Linux Audit Framework

echo -n "set xtics rotate (\"-1\" -1" >> $gpcommand

# make sure we don't append to pre-existing file rm -f $gpdata

# read input i=0 while [ 1 ] do read -t 5 line 2>/dev/null if [ $? -ne 0 ] ; then break fi if [ x"$line" != "x" ] ; then i=`expr $i + 1` echo $line | awk '/^[0-9]/ { printf ", \"%s\" %d", $2, 1+num }' "num=$i" \ >> $gpcommand echo $line | awk '/^[0-9]/ { printf "%d %s\n", 1+num, $1 }' "num=$i" >> \ $gpdata fi done echo -e ')\n' >> $gpcommand echo 'plot "'$gpdata'" with boxes' >> $gpcommand

# Create the chart gnuplot $gpcommand

# Cleanup rm -f $gpcommand $gpdata

# output results if [ -e $gpout ] ; then echo "Wrote $gpout" exit 0 fi exit 1

Creating Bar Charts from the Audit Statistics 67

Linux SUSE Linux Enterprise User Manual

Specifications and Main Features

Frequently Asked Questions

User Manual