AppArmor confinement is provided via profiles loaded into the kernel via apparmor_parser(8), typically through the apparmor.service systemd unit, which is used like this:
# systemctl start apparmor
# systemctl reload apparmor
AppArmor can operate in two modes: enforcement, and complain or learning:
Profiles are traditionally stored in files in /etc/apparmor.d/ under filenames with the convention of replacing the / in pathnames with . (except for the root /) so profiles are easier to manage (e.g. the /usr/sbin/nscd profile would be named usr.sbin.nscd).
Profiles are applied to a process at exec(3) time (as seen through the execve(2) system call): once a profile is loaded for a program, that program will be confined on the next exec(3). If a process is already running under a profile, when one replaces that profile in the kernel, the updated profile is applied immediately to that process. On the other hand, a process that is already running unconfined cannot be confined.
AppArmor supports the Linux kernel's securityfs filesystem, and makes available the list of the profiles currently loaded; to mount the filesystem:
# mount -tsecurityfs securityfs /sys/kernel/security
$ cat /sys/kernel/security/apparmor/profiles
/usr/bin/mutt
/usr/bin/gpg
...
Normally, the initscript will mount securityfs if it has not already been done.
AppArmor also restricts what privileged operations a confined process may execute, even if the process is running as root. A confined process cannot call the following system calls:
create_module(2) delete_module(2) init_module(2) ioperm(2)
iopl(2) ptrace(2) reboot(2) setdomainname(2)
sethostname(2) swapoff(2) swapon(2) sysctl(2)
Complain mode can be used to develop profiles incrementally as an application is exercised. The logged accesses can be added to the profile and then can the application further exercised to discover further additions that are needed. Because AppArmor allows the accesses the application will behave as it would if AppArmor was not confining it.
Warning complain mode does not provide any security, only auditing, while it is enabled. It should not be used in a hostile environment or bad behaviors may be logged and added to the profile as if they are resource accesses that should be used by the application.
Note complain mode can be very noisy with new or empty profiles, but with developed profiles might not log anything if the profile covers the application behavior well. See Audit Rate Limiting if complain mode is generating too many log messages.
To set a profile and any children or hat profiles the profile may contain into complain mode use
aa-complain /etc/apparmor.d/<the-application>
To manually set a specific profile in complain mode, add the "complain" flag, and then manually reload the profile:
profile foo flags=(complain) { ... }
Note that the "complain" flag must also be added manually to any hats or children profiles of the profile or they will continue to use the previous mode.
To enable complain mode globally, run:
echo -n complain > /sys/module/apparmor/parameters/mode
or to set it on boot add:
apparmor.mode=complain
as a kernel boot parameter.
Warning Setting complain mode globally disables all apparmor security protections. It can be useful during debugging or profile development, but setting it selectively on a per profile basis is safer.
audit(1386511672.612:238): apparmor="DENIED" operation="exec"
parent=7589 profile="/tmp/sh" name="/bin/uname" pid=7605
comm="sh" requested_mask="x" denied_mask="x" fsuid=0 ouid=0
audit(1386511672.613:239): apparmor="DENIED" operation="open"
parent=7589 profile="/tmp/sh" name="/bin/uname" pid=7605
comm="sh" requested_mask="r" denied_mask="r" fsuid=0 ouid=0
audit(1386511772.804:246): apparmor="DENIED" operation="capable"
parent=7246 profile="/tmp/sh" pid=7589 comm="sh" pid=7589
comm="sh" capability=2 capname="dac_override"
The permissions requested by the process are described in the operation= and denied_mask= (for files - capabilities etc. use a slightly different log format). The "name" and process id of the running program are reported, as well as the profile name including any "hat" that may be active, separated by "//". ("Name" is in quotes, because the process name is limited to 15 bytes; it is the same as reported through the Berkeley process accounting.)
For confined processes running under a profile that has been loaded in complain mode, enforcement will not take place and the log messages reported to audit will be of the form:
audit(1386512577.017:275): apparmor="ALLOWED" operation="open"
parent=8012 profile="/usr/bin/du" name="/etc/apparmor.d/tunables/"
pid=8049 comm="du" requested_mask="r" denied_mask="r" fsuid=1000 ouid=0
audit(1386512577.017:276): apparmor="ALLOWED" operation="open"
parent=8012 profile="/usr/bin/du" name="/etc/apparmor.d/tunables/"
pid=8049 comm="du" requested_mask="r" denied_mask="r" fsuid=1000 ouid=0
If the userland auditd is not running, the kernel will send audit events to klogd; klogd will send the messages to syslog, which will log the messages with the KERN facility. Thus, REJECTING and PERMITTING messages may go to either /var/log/audit/audit.log or /var/log/messages, depending upon local configuration.
To enable debug mode, run:
echo 1 > /sys/module/apparmor/parameters/debug
or to set it on boot add:
apparmor.debug=1
To turn off deny audit quieting, run:
echo -n noquiet >/sys/module/apparmor/parameters/audit
or to set it on boot add:
apparmor.audit=noquiet
Warning! Force audit mode can be extremely noisy even for a single profile, let alone when enabled globally.
To set a specific profile in force audit mode, add the "audit" flag:
profile foo flags=(audit) { ... }
To enable force audit mode globally, run:
echo -n all > /sys/module/apparmor/parameters/audit
or to set it on boot add:
apparmor.audit=all
as a kernel boot parameter.
Audit Rate Limiting
If auditd is not running, to avoid losing too many of the extra log messages, you will likely have to turn off rate limiting by doing:
echo 0 > /proc/sys/kernel/printk_ratelimit
But even then the kernel ring buffer may overflow and you might lose messages.
Else, if auditd is running, see auditd(8) and auditd.conf(5).