Keep watch for malicious command execution

Linux systems have a powerful auditing facility called auditd which can give a very detailed accounting of actions and changes in a system, but by default, no auditd rules are active so we tend to miss out on this detailed history. In this lab, we will configure auditd on a Linux machine to account for all commands executed by a given user (it should be "centos" if you built the lab following this guide), including commands run by this user in a sudo command or after sudo-ing to root. After causing some audit events to be generated, we will look them over in Kibana. Then we will set up several custom Wazuh rules to alert on especially suspicious command calls, making use of the CDB list lookup capability that allows rules to look up decoded field values in various lists and to use the results as part of the alert criteria.

The Linux auditd system is an extensive auditing tool, which we will only touch on here. Consider reading the Monitoring system calls section to get a broader picture of the ways you can take advantage of it.

Turn on program call auditing on linux-agent

  1. Having already sudo-ed to root on our linux-agent machine, append the following audit rules to /etc/audit/rules.d/audit.rules

    [root@linux-agent centos]# echo "-a exit,always -F auid=1000 -F egid!=994 -F auid!=-1 -F arch=b32 -S execve -k audit-wazuh-c" >> /etc/audit/rules.d/audit.rules
    [root@linux-agent centos]# echo "-a exit,always -F auid=1000 -F egid!=994 -F auid!=-1 -F arch=b64 -S execve -k audit-wazuh-c" >> /etc/audit/rules.d/audit.rules

    Where auid=1000 represents the user ID. If unsure, you may verify this value by running: grep centos /etc/passwd (replacing centos if you have a different user name).

  2. Then reload the rules and confirm they are in place:

    [root@linux-agent centos]# auditctl -R /etc/audit/rules.d/audit.rules
    [root@linux-agent centos]# auditctl -l
    -a always,exit -F arch=b32 -S execve -F auid=1000 -F egid!=994 -F auid!=-1 -F key=audit-wazuh-c
    -a always,exit -F arch=b64 -S execve -F auid=1000 -F egid!=994 -F auid!=-1 -F key=audit-wazuh-c

Trigger a few audit events

  1. Dropping from root back to the unpriviledged user, run a ping.

    [root@linux-agent centos]# exit
    [centos@linux-agent ~]$ ping -c1
    PING ( 56(84) bytes of data.
    64 bytes from icmp_seq=1 ttl=51 time=1.09 ms
    --- ping statistics ---
    1 packets transmitted, 1 received, 0% packet loss, time 0ms
    rtt min/avg/max/mdev = 1.093/1.093/1.093/0.000 ms
  2. While still centos, use sudo to run a privileged commands

    [centos@linux-agent ~]$ sudo cat /etc/shadow
  3. Now sudo back to root and run another commands

    [centos@linux-agent ~]$ sudo su -
    Last login: Thu Nov 14 12:27:12 UTC 2019 on pts/0
    [root@linux-agent ~]# df
    Filesystem     1K-blocks    Used Available Use% Mounted on
    /dev/xvda1       8377344 1616824   6760520  20% /
    devtmpfs          486604       0    486604   0% /dev
    tmpfs             507288       0    507288   0% /dev/shm
    tmpfs             507288   12956    494332   3% /run
    tmpfs             507288       0    507288   0% /sys/fs/cgroup
    tmpfs             101460       0    101460   0% /run/user/1000

Look over the audit events

  1. On the monitored linux machine, inspect the content of /var/log/audit/audit.log. Auditd writes events here, but it is not very readable. Thankfully the Linux Wazuh agents already monitors this file by default.

  2. Search Kibana for in the Kibana Discover area. That will catch all auditd command audit events.

  3. Pick the following Kibana fields for columnar display:

    • data.audit.command

    • data.audit.auid

    • data.audit.euid

    • full_log

  4. Explore the audit records, finding and examining your unprivileged ping, and your privileged cat and df calls. They will be mingled with other commands.

  5. The centos user has uid 1000. User root has uid 0. Notice the auid (audited user identity) always traces back to the centos user, even though the euid effective user identity is sometimes 0 and sometimes 1000 depending on whether privileges were escalated. This allows you to see who actually ran the command with sudo or while sudo-ed to root.

Look over the relevant Wazuh rule

  1. Here is Wazuh rule 80792:

    <rule id="80792" level="3">
        <list field="audit.key" lookup="match_key_value" check_value="command">etc/lists/audit-keys</list>
        <description>Audit: Command: $(audit.exe)</description>

    Parent rule 80700 catches all auditd events, while this rule focuses on auditd command events. Notice how the <list> line in this rule takes the decoded audit.key value which all our auditd rules set to "audit-wazuh-c" presently, and looks this up in a CDB list called audit-keys to see if the audit.key value is listed with a value of "command".

  2. Look over the key-value pairs in the lookup file. The file is /var/ossec/etc/lists/audit-keys.


    This CDB list contains keys and values separated colons. Some lists only contain keys, in which case each key exists on a line of its own and is directly followed by a colon.

  3. Notice that in addition to the text file /var/ossec/etc/lists/audit-keys, there is also a binary /var/ossec/etc/lists/audit-keys.cdb file that Wazuh uses for actual lookups.

Create a list of commands that Wazuh will watch for

Wazuh allows you to maintain flat file CDB lists (key only or key:value) which are compiled into a special binary format to facilitate high-performance lookups in Wazuh rules. Such lists must be created as files, added to the Wazuh configuration, and then compiled. After that, rules can be built to look up decoded fields in those CDB lists as part of their match criteria. Right now, we want to create a list of commands that Wazuh will use to give us special alerts when executed.

  1. On wazuh-manager, create /var/ossec/etc/lists/suspicious-programs with this content:

  2. On wazuh-manager, add this to the <ruleset> section of ossec.configuration in /var/ossec/etc/ossec.conf:



    Before Wazuh v3.11.0 it was necessary to run /var/ossec/bin/ossec-makelists after changing CDB lists. After v3.11.0 the lists are already compiled when the manager is started.

  3. Now let's add a new rule that uses this list as part of its criteria to do so add the following to /var/ossec/etc/rules/local_rules.xml on the Wazuh Manager.

    <group name="audit">
      <rule id="100200" level="8">
          <list field="audit.command" lookup="match_key">etc/lists/suspicious-programs</list>
          <description>Audit: Suspicious Command: $(audit.exe)</description>

    In this case we are simply checking to see if the decoded audit.command value appears in our new CDB lists at all, without checking its value.

  4. Compile the CDB list (if your version is inferior to v3.11.0):

    [root@wazuh-manager centos]# /var/ossec/bin/ossec-makelists
  5. Restart the Wazuh manager:

    1. For Systemd:

      # systemctl restart wazuh-manager
    2. For SysV Init:

      # service wazuh-manager restart
  6. On the linux computer monitored agent, install and run tcpdump to trip our new rule:

    [root@linux-agent ~]# yum -y install tcpdump
    [root@linux-agent ~]# tcpdump --version
  7. Search Kibana for data.audit.command:tcpdump and expand the record, where you should see a of 100200.

Make a smarter list and rule

Let's make this list a little smarter by including values that indicate how alarmed we should be about a given program being run.

  1. On The manager, replace the content of /var/ossec/etc/lists/suspicious-programs with the following:

  2. Now that our suspicious-programs list is more granular, let's create a higher severity rule to fire specifically on instances when a "red" program is executed.

    Add this new rule to /var/ossec/etc/rules/local_rules.xml on wazuh-manager, directly after rule 100200 and before the closing </group> tag:

    <rule id="100210" level="12">
        <list field="audit.command" lookup="match_key_value" check_value="red">etc/lists/suspicious-programs</list>
        <description>Audit: Highly Suspicious Command executed: $(audit.exe)</description>
  3. Compile the CDB list (if your version is inferior to v3.11.0):

    [root@wazuh-manager centos]# /var/ossec/bin/ossec-makelists
  4. Restart the Wazuh manager:

    1. For Systemd:

      # systemctl restart wazuh-manager
    2. For SysV Init:

      # service wazuh-manager restart
  5. On the monitored linux agent install and run a "red" program (netcat):

    [root@linux-agent ~]# yum -y install nmap-ncat
    [root@linux-agent ~]# nc -v
  6. Search Kibana for data.audit.command:nc and expand the record, noting especially the rule.description of "Audit: Highly Suspicious Command executed: /usr/bin/ncat"

Make an exception

You have ping in your CDB list, but perhaps you have several systems that routinely ping as a connectivity check and you don't want these events to be logged. Another child rule of 80297, with a level of "0" can provide such an exception.

  1. Add this new rule to /var/ossec/etc/rules/local_rules.xml on wazuh-manager, directly after rule 100210 and before the closing </group> tag.:

    <rule id="100220" level="0">
        <description>Ignore pings of</description>
        <field name="audit.command">^ping$</field>

    This rule does not do a lookup, it just checks any auditd command records in which the ping command is called and the target IP address is mentioned.

  2. Restart the Wazuh manager:

    1. For Systemd:

      # systemctl restart wazuh-manager
    2. For SysV Init:

      # service wazuh-manager restart
  3. In you linux-agent, test the rule by pinging both and

    [root@linux-agent ~]# yum -y install tcpdump
    [root@linux-agent ~]# ping -c1
    [root@linux-agent ~]# ping -c1
  4. Search Kibana for data.audit.command:ping. Notice that only the ping event involving shows up, because the other one was ignored by this exception rule.

Observe the order in which our child rules are evaluated

  1. On your linux-agent, run a mundane command not listed in our CDB.

    [root@linux-agent ~]# sleep 1
  2. Search Kibana for data.audit.command:sleep to find the resulting event. Copy the full_log value.

  3. Run /var/ossec/bin/ossec-logtest -v on the Wazuh Manager and paste in the full_log value from above.

  4. Carefully note the order in which child rules of "80792 - Audit: Command" were evaluated:

    Trying rule: 80789 - Audit: Watch - Execute access: $(
    Trying rule: 80792 - Audit: Command: $(audit.exe)
       *Rule 80792 matched.
       *Trying child rules.
    Trying rule: 100220 - Ignore pings of
    Trying rule: 100210 - Audit: Highly Suspicious Command: $(audit.exe)
    Trying rule: 100200 - Audit: Suspicious Command: $(audit.exe)
  5. Remember that when a rule matches, if it has multiple child rules, they are not evaluated in order of ID nor in the order they appear in the rule file. Instead, child rules of level "0" are checked first since they are for making exceptions. Then any remaining child rules are checked in the order of highest severity to lowest severity. Keep this in mind as you build child rules of your own.


Why does my new rule never fire?

Sometimes a new rule never matches anything because of a flaw in its criteria. Other times it never matches because it is never even evaluated. Remember, ossec-logtest -v is your friend. Use it to see if your rule is being evaluated at all, and if not, what rule might be overshadowing it.

Remember to set your settings back to normal

When testing different things, it is recommendable that you reverse the changes to keep your testing Lab clean. So new tests don't interfere with previous ones.

You would need to delete the line we wrote in the <ruleset> section of Wazuh manager's configuration, ossec.conf:


In the linux-agent, delete the two lines we added to /etc/audit/rules.d/audit.rules:

-a always,exit -F arch=b32 -S execve -F auid=1000 -F egid!=994 -F auid!=-1 -F key=audit-wazuh-c
-a always,exit -F arch=b64 -S execve -F auid=1000 -F egid!=994 -F auid!=-1 -F key=audit-wazuh-c

Now you would need to reload the auditd rules and restart the manager for changes to take effect:

[root@linux-agent centos]# auditctl -R /etc/audit/rules.d/audit.rules
[root@wazuh-manager centos]# systemctl restart wazuh-manager