title | section | header | footer |
---|---|---|---|
laurel - Advice for writing Audit rulesets for use with `laurel(8)` |
7 |
System Administration Utilities |
laurel 0.6.4 |
laurel-audit-rules - Advice for writing Audit rulesets for use with laurel(8)
This page contains suggestions for Linux Audit rulesets that are useful to aid in detecting common attacker's tactics.
It is not possible for /auditctl(8)/ ro load file watches for files or directories that are not present. Depending on the rule set, it will spam possibly lots of error messages to standard error. The specific file watches are not installed, but those error messages can be ignored otherwise.
-w /etc/group -p wa -k wr_group
-w /etc/passwd -p wa -k wr_passwd
-w /etc/shadow -p wa -k wr_passwd
-w /etc/pam.conf -p wa -k wr_pam
-w /etc/pam.d/ -p wa -k wr_pam
-w /etc/ssh/sshd_config -p wa -k wr_sshd
-w /etc/sudoers -p wa -k wr_sudo
-w /etc/sudoers.d -p wa -k wr_sudo
-w /etc/crontab -p wa -k wr_cron
-w /etc/cron.d/ -p wa -k wr_cron
-w /etc/cron.daily/ -p wa -k wr_cron
-w /etc/cron.hourly/ -p wa -k wr_cron
-w /etc/cron.monthly/ -p wa -k wr_cron
-w /etc/cron.weekly/ -p wa -k wr_cron
-w /etc/cron.yearly/ -p wa -k wr_cron
-w /etc/cron.allow -p wa -k wr_cron
-w /etc/cron.deny -p wa -k wr_cron
-w /var/spool/cron/crontabs/ -p wa -k wr_cron
-w /etc/at.allow -p wa -k wr_cron
-w /etc/at.deny -p wa -k wr_cron
-w /var/spool/cron/atjobs/ -p wa -k wr_cron
Systemd also has cron-like timer mechanism. udev triggers have also
been abused for persistence. Note that watching the files in /etc
is
not sufficient.
-w /etc/systemd -p wa -k wr_systemd
-w /lib/systemd -p wa -k wr_systemd
-w /usr/lib/systemd -p wa -k wr_systemd
-w /etc/udev -p wa -k wr_systemd
-w /lib/udev -p wa -k wr_systemd
-w /usr/lib/udev -p wa -k wr_systemd
-w /lib/ld-linux.so.2 -p wa -k wr_ldso
-w /lib64/ld-linux-x86-64.so.2 -p wa -k wr_ldso
-w /lib/ld-musl-x86_64.so.1 -p wa -k wr_ldso
-w /lib/ld-musl-i386.so.1 -p wa -k wr_ldso
-w /etc/ld.so.conf -p wa -k wr_ldso
-w /etc/ld.so.conf.d -p wa -k wr_ldso
-w /etc/ld.so.preload -p wa -k wr_ldso
-w /etc/selinux -p wa -k wr_selinux
-w /usr/share/selinux -p wa -k wr_selinux
-w /usr/libexec/selinux -p wa -k wr_selinux
-w /etc/apparmor.d -p wa -k wr_apparmor
-w /usr/lib/apparmor -p wa -k wr_apparmor
-w /usr/share/apparmor -p wa -k wr_apparmor
-w /usr/share/apparmor-features -p wa -k wr_apparmor
-w /etc/modprobe.conf -p wa -k wr_modules
-w /etc/modprobe.d/ -p wa -k wr_modules
-w /lib/modules/ -p wa -k wr_modules
-w /etc/audit/ -p wa -k wr_audit_config
-w /etc/libaudit.conf -p wa -k wr_audit_config
-w /etc/audisp/ -p wa -k wr_audit_config
-w /etc/laurel/ -p wa -k wr_laurel_confg
-w /sbin/auditctl -p x -k wr_audit_tools
-w /sbin/auditd -p x -k wr_audit_tools
-w /usr/sbin/laurel -p x -k wr_audit_tools
Adding context to system service activities is useful because together
with Laurel's process labels (label-process.label-keys
,
label-process.propagate-labels
), it enables more accurate detection
rules that can help recognize benign system management activity.
-w /usr/sbin/sshd -p x -k sshd
-w /usr/bin/yum -p x -k pkg_mgmt
-w /usr/bin/rpm -p x -k pkg_mgmt
-w /usr/bin/dnf -p x -k pkg_mgmt
-w /usr/bin/dpkg -p x -k pkg_mgmt
-w /usr/bin/apt -p x -k pkg_mgmt
-w /usr/bin/apt-get -p x -k pkg_mgmt
-w /usr/bin/apt-key -p x -k pkg_mgmt
-w /usr/bin/apt-add-repository -p x -k pkg_mgmt
-w /usr/bin/aptitude -p x -k pkg_mgmt
-w /usr/bin/aptitude-curses -p x -k pkg_mgmt
-w /usr/bin/wajig -p x -k pkg_mgmt
-w /usr/bin/snap -p x -k pkg_mgmt
-w /usr/sbin/yast2 -p x -k pkg_mgmt
-w /usr/bin/zypper -p x -k pkg_mgmt
-w /usr/bin/containerd -p x -k container
-w /usr/bin/podman -p x -k container
-w /usr/bin/runc -p x -k container
-w /usr/bin/dockerd -p x -k container
-w /usr/bin/docker -p x -k container
-w /usr/bin/docker-containerd -p x -k container
-w /usr/bin/docker-runc -p x -k container
-w /usr/sbin/cron -p x -k sched_task
-w /usr/sbin/atd -p x -k sched_task
-w /usr/sbin/httpd -p x -k apache-httpd
-w /usr/local/apache2/bin/httpd -p x -k apache-httpd
-w /usr/sbin/nginx -p x -k nginx
-w /usr/local/nginx/sbin/nginx -p x -k nginx
-w /usr/local/openresty/nginx/sbin/nginx -p x -k nginx
For reliable process tracking that is required for assigning and
propagating process labels, it is useful to have the Linux Audit
subsystem produce events for all fork
/exec
style syscalls.
## Ignore clone( flags=CLONE_VM|… ), log other process-creating calls
-a never,exit -F arch=b32 -S clone -F a2&0x100
-a never,exit -F arch=b64 -S clone -F a2&0x100
-a always,exit -F arch=b32 -S fork,vfork,clone,clone3 -k fork
-a always,exit -F arch=b64 -S fork,vfork,clone,clone3 -k fork
-a always,exit -F arch=b32 -S execve,execveat
-a always,exit -F arch=b64 -S execve,execveat
It is only important that Laurel gets to observe these events. To reduce log volume, Laurel's filtering settings should be used, e.g.:
[filter]
filter-keys = ["fork"]
filter-action = drop
keep-first-per-process = true
We are interested in ptrce usage, but not in every transaction (PEEK
, POKE
, CONT
)
-a never,exit -F arch=b32 -S ptrace -F a0>=1 -F a0<=7
-a never,exit -F arch=b64 -S ptrace -F a0>=1 -F a0<=7
-a always,exit -F arch=b32 -S ptrace
-a always,exit -F arch=b64 -S ptrace
Usage of BPF should be restricted to few processes; log everything except data transfer operations because they would put too much load on the system.
-a never,exit -F arch=b32 -S bpf -F a0>=1 -F a0<=4
-a never,exit -F arch=b64 -S bpf -F a0>=1 -F a0<=4
-a never,exit -F arch=b32 -S bpf -F a0>=0xb -F a0<=0xf
-a never,exit -F arch=b64 -S bpf -F a0>=0xb -F a0<=0xf
-a never,exit -F arch=b32 -S bpf -F a0=0x13
-a never,exit -F arch=b64 -S bpf -F a0=0x13
-a always,exit -F arch=b32 -S bpf -F success=1
-a always,exit -F arch=b64 -S bpf -F success=1
-a always,exit -F arch=b32 -S init_module,finit_module,delete_module -k module
-a always,exit -F arch=b64 -S init_module,finit_module,delete_module -k module
audit.rules(7)
, laurel(8)
- Hilko Bengen <[email protected]>