1AUDITCTL(8) System Administration Utilities AUDITCTL(8)
2
6 auditctl - a utility to assist controlling the kernel's audit system
7
9 auditctl [options]
10
12 The auditctl program is used to configure kernel options related to au‐
13 diting, to see status of the configuration, and to load discretionary
14 audit rules.
15
17 -b backlog
18 Set max number (limit) of outstanding audit buffers allowed
19 (Kernel Default=64) If all buffers are full, the failure flag is
20 consulted by the kernel for action.
21 --backlog_wait_time wait_time
23 Set the time for the kernel to wait (Kernel Default 60*HZ) when
24 the backlog limit is reached before queuing more audit events to
25 be transferred to auditd. The number must be greater than or
26 equal to zero and less that 10 times the default value.
27 --reset_backlog_wait_time_actual
29 Reset the actual backlog wait time counter shown by the status
30 command.
31 -c Continue loading rules in spite of an error. This summarizes the
33 results of loading the rules. The exit code will not be success
34 if any rule fails to load.
35 -D Delete all rules and watches. This can take a key option (-k),
37 too.
38 -e [0..2]
40 Set enabled flag. When 0 is passed, this can be used to tempo‐
41 rarily disable auditing. When 1 is passed as an argument, it
42 will enable auditing. To lock the audit configuration so that it
43 can't be changed, pass a 2 as the argument. Locking the configu‐
44 ration is intended to be the last command in audit.rules for
45 anyone wishing this feature to be active. Any attempt to change
46 the configuration in this mode will be audited and denied. The
47 configuration can only be changed by rebooting the machine.
48 -f [0..2]
50 Set failure mode 0=silent 1=printk 2=panic. This option lets you
51 determine how you want the kernel to handle critical errors. Ex‐
52 ample conditions where this mode may have an effect includes:
53 transmission errors to userspace audit daemon, backlog limit ex‐
54 ceeded, out of kernel memory, and rate limit exceeded. The de‐
55 fault value is 1. Secure environments will probably want to set
56 this to 2.
57 -h Help
59 -i When given by itself, ignore errors when reading rules from a
61 file. This causes auditctl to always return a success exit code.
62 If passed as an argument to -s then it gives an interpretation
63 of the numbers to human readable words if possible.
64 --loginuid-immutable
66 This option tells the kernel to make loginuids unchangeable once
67 they are set. Changing loginuids requires CAP_AUDIT_CONTROL. So,
68 its not something that can be done by unprivileged users. Set‐
69 ting this makes loginuid tamper-proof, but can cause some prob‐
70 lems in certain kinds of containers.
71 -q mount-point,subtree
73 If you have an existing directory watch and bind or move mount
74 another subtree in the watched subtree, you need to tell the
75 kernel to make the subtree being mounted equivalent to the di‐
76 rectory being watched. If the subtree is already mounted at the
77 time the directory watch is issued, the subtree is automatically
78 tagged for watching. Please note the comma separating the two
79 values. Omitting it will cause errors.
80 -r rate
82 Set limit in messages/sec (0=none). If this rate is non-zero and
83 is exceeded, the failure flag is consulted by the kernel for ac‐
84 tion. The default value is 0.
85 --reset-lost
87 Reset the lost record counter shown by the status command.
88 -R file
90 Read rules from a file. The rules must be 1 per line and in the
91 order that they are to be executed in. The rule file must be
92 owned by root and not readable by other users or it will be re‐
93 jected. The rule file may have comments embedded by starting the
94 line with a '#' character. Rules that are read from a file are
95 identical to what you would type on a command line except they
96 are not preceded by auditctl (since auditctl is the one execut‐
97 ing the file) and you would not use shell escaping since au‐
98 ditctl is reading the file instead of bash.
99 --signal signal
101 Send a signal to the audit daemon. You must have privileges to
102 do this. Supported signals are TERM, HUP, USR1, USR2, CONT.
103 -t Trim the subtrees after a mount command.
105
107 -l List all rules 1 per line. Two more options may be given to this
108 command. You can give either a key option (-k) to list rules
109 that match a key or a (-i) to have a0 through a3 interpreted to
110 help determine the syscall argument values are correct .
111 -m text
113 Send a user space message into the audit system. This can only
114 be done if you have CAP_AUDIT_WRITE capability (normally the
115 root user has this). The resulting event will be the USER type.
116 -s Report the kernel's audit subsystem status. It will tell you the
118 in-kernel values that can be set by -e, -f, -r, and -b options.
119 The pid value is the process number of the audit daemon. Note
120 that a pid of 0 indicates that the audit daemon is not running.
121 The lost entry will tell you how many event records that have
122 been discarded due to the kernel audit queue overflowing. The
123 backlog field tells how many event records are currently queued
124 waiting for auditd to read them. This option can be followed by
125 the -i to get a couple fields interpreted.
126 -v Print the version of auditctl.
128
131 -a [list,action|action,list]
132 Append rule to the end of list with action. Please note the
133 comma separating the two values. Omitting it will cause errors.
134 The fields may be in either order. It could be list,action or
135 action,list. The following describes the valid list names:
136 task Add a rule to the per task list. This rule list is
138 used only at the time a task is created -- when
139 fork() or clone() are called by the parent task.
140 When using this list, you should only use fields
141 that are known at task creation time, such as the
142 uid, gid, etc.
143 exit Add a rule to the syscall exit list. This list is
145 used upon exit from a system call to determine if an
146 audit event should be created.
147 user Add a rule to the user message filter list. This
149 list is used by the kernel to filter events origi‐
150 nating in user space before relaying them to the au‐
151 dit daemon. It should be noted that the only fields
152 that are valid are: uid, auid, gid, pid, subj_user,
153 subj_role, subj_type, subj_sen, subj_clr, msgtype,
154 and executable name. All other fields will be
155 treated as non-matching. It should be understood
156 that any event originating from user space from a
157 process that has CAP_AUDIT_WRITE will be recorded
158 into the audit trail. This means that the most
159 likely use for this filter is with rules that have
160 an action of never since nothing has to be done to
161 allow events to be recorded.
162 exclude Add a rule to the event type exclusion filter list.
164 This list is used to filter events that you do not
165 want to see. For example, if you do not want to see
166 any avc messages, you would using this list to
167 record that. Events can be excluded by process ID,
168 user ID, group ID, login user ID, message type, sub‐
169 ject context, or executable name. The action is ig‐
170 nored and uses its default of "never".
171 filesystem Add a rule that will be applied to a whole filesys‐
173 tem. The filesystem must be identified with a fstype
174 field. Normally this filter is used to exclude any
175 events for a whole filesystem such as tracefs or de‐
176 bugfs.
177 The following describes the valid actions for the rule:
179 never No audit records will be generated. This can be used
181 to suppress event generation. In general, you want
182 suppressions at the top of the list instead of the
183 bottom. This is because the event triggers on the
184 first matching rule.
185 always Allocate an audit context, always fill it in at
187 syscall entry time, and always write out a record at
188 syscall exit time.
189 -A list,action
191 Add rule to the beginning list with action.
192 -C [f=f | f!=f]
194 Build an inter-field comparison rule: field, operation, field.
195 You may pass multiple comparisons on a single command line. Each
196 one must start with -C. Each inter-field equation is anded with
197 each other as well as equations starting with -F to trigger an
198 audit record. There are 2 operators supported - equal, and not
199 equal. Valid fields are:
200 auid, uid, euid, suid, fsuid, obj_uid; and gid, egid, sgid, fs‐
202 gid, obj_gid
203 The two groups of uid and gid cannot be mixed. But any compari‐
205 son within the group can be made. The obj_uid/gid fields are
206 collected from the object of the event such as a file or direc‐
207 tory.
208 -d list,action
211 Delete rule from list with action. The rule is deleted only if
212 it exactly matches syscall name(s) and every field name and
213 value.
214 -F [n=v | n!=v | n<v | n>v | n<=v | n>=v | n&v | n&=v]
216 Build a rule field: name, operation, value. You may have up to
217 64 fields passed on a single command line. Each one must start
218 with -F. Each field equation is anded with each other (as well
219 as equations starting with -C) to trigger an audit record. There
220 are 8 operators supported - equal, not equal, less than, greater
221 than, less than or equal, and greater than or equal, bit mask,
222 and bit test respectively. Bit test will "and" the values and
223 check that they are equal, bit mask just "ands" the values.
224 Fields that take a user ID may instead have the user's name; the
225 program will convert the name to user ID. The same is true of
226 group names. Valid fields are:
227 a0, a1, a2, a3
229 Respectively, the first 4 arguments to a syscall.
230 Note that string arguments are not supported. This
231 is because the kernel is passed a pointer to the
232 string. Triggering on a pointer address value is not
233 likely to work. So, when using this, you should only
234 use on numeric values. This is most likely to be
235 used on platforms that multiplex socket or IPC oper‐
236 ations.
237 arch The CPU architecture of the syscall. The arch can be
239 found doing 'uname -m'. If you do not know the arch
240 of your machine but you want to use the 32 bit
241 syscall table and your machine supports 32 bit, you
242 can also use b32 for the arch. The same applies to
243 the 64 bit syscall table, you can use b64. In this
244 way, you can write rules that are somewhat arch in‐
245 dependent because the family type will be auto de‐
246 tected. However, syscalls can be arch specific and
247 what is available on x86_64, may not be available on
248 ppc. The arch directive should precede the -S option
249 so that auditctl knows which internal table to use
250 to look up the syscall numbers.
251 auid The original ID the user logged in with. Its an ab‐
253 breviation of audit uid. Sometimes its referred to
254 as loginuid. Either the user account text or number
255 may be used.
256 devmajor Device Major Number
258 devminor Device Minor Number
260 dir Full Path of Directory to watch. This will place a
262 recursive watch on the directory and its whole sub‐
263 tree. It can only be used on exit list. See "-w".
264 egid Effective Group ID. May be numeric or the groups
266 name.
267 euid Effective User ID. May be numeric or the user ac‐
269 count name.
270 exe Absolute path to application that while executing
272 this rule will apply to. It supports = and != opera‐
273 tors. Note that you can only use this once for each
274 rule.
275 exit Exit value from a syscall. If the exit code is an
277 errno, you may use the text representation, too.
278 fsgid Filesystem Group ID. May be numeric or the groups
280 name.
281 fsuid Filesystem User ID. May be numeric or the user ac‐
283 count name.
284 filetype The target file's type. Can be either file, dir,
286 socket, link, character, block, or fifo.
287 gid Group ID. May be numeric or the groups name.
289 inode Inode Number
291 key This is another way of setting a filter key. See
293 discussion above for -k option.
294 msgtype This is used to match the event's record type. It
296 should only be used on the exclude or user filter
297 lists.
298 obj_uid Object's UID
300 obj_gid Object's GID
302 obj_user Resource's SE Linux User
304 obj_role Resource's SE Linux Role
306 obj_type Resource's SE Linux Type
308 obj_lev_low Resource's SE Linux Low Level
310 obj_lev_high
312 Resource's SE Linux High Level
313 path Full Path of File to watch. It can only be used on
315 exit list.
316 perm Permission filter for file operations. See "-p". It
318 can only be used on exit list. You can use this
319 without specifying a syscall and the kernel will se‐
320 lect the syscalls that satisfy the permissions being
321 requested.
322 pers OS Personality Number
324 pid Process ID
326 ppid Parent's Process ID
328 saddr_fam Address family number as found in /usr/in‐
330 clude/bits/socket.h. For example, IPv4 would be 2
331 and IPv6 would be 10.
332 sessionid User's login session ID
334 subj_user Program's SE Linux User
336 subj_role Program's SE Linux Role
338 subj_type Program's SE Linux Type
340 subj_sen Program's SE Linux Sensitivity
342 subj_clr Program's SE Linux Clearance
344 sgid Saved Group ID. See getresgid(2) man page.
346 success If the exit value is >= 0 this is true/yes otherwise
348 its false/no. When writing a rule, use a 1 for
349 true/yes and a 0 for false/no
350 suid Saved User ID. See getresuid(2) man page.
352 uid User ID. May be numeric or the user account name.
354 -k key Set a filter key on an audit rule. The filter key is an arbi‐
356 trary string of text that can be up to 31 bytes long. It can
357 uniquely identify the audit records produced by a rule. Typical
358 use is for when you have several rules that together satisfy a
359 security requirement. The key value can be searched on with
360 ausearch so that no matter which rule triggered the event, you
361 can find its results. The key can also be used on delete all
362 (-D) and list rules (-l) to select rules with a specific key.
363 You may have more than one key on a rule if you want to be able
364 to search logged events in multiple ways or if you have an au‐
365 ditd plugin that uses a key to aid its analysis.
366 -p [r|w|x|a]
368 Describe the permission access type that a file system watch
369 will trigger on. r=read, w=write, x=execute, a=attribute change.
370 These permissions are not the standard file permissions, but
371 rather the kind of syscall that would do this kind of thing. The
372 read & write syscalls are omitted from this set since they would
373 overwhelm the logs. But rather for reads or writes, the open
374 flags are looked at to see what permission was requested.
375 -S [Syscall name or number|all]
377 Any syscall name or number may be used. The word 'all' may also
378 be used. If the given syscall is made by a program, then start
379 an audit record. If a field rule is given and no syscall is
380 specified, it will default to all syscalls. You may also specify
381 multiple syscalls in the same rule by using multiple -S options
382 in the same rule. Doing so improves performance since fewer
383 rules need to be evaluated. Alternatively, you may pass a comma
384 separated list of syscall names. If you are on a bi-arch system,
385 like x86_64, you should be aware that auditctl simply takes the
386 text, looks it up for the native arch (in this case b64) and
387 sends that rule to the kernel. If there are no additional arch
388 directives, IT WILL APPLY TO BOTH 32 & 64 BIT SYSCALLS. This can
389 have undesirable effects since there is no guarantee that any
390 syscall has the same number on both 32 and 64 bit interfaces.
391 You will likely want to control this and write 2 rules, one with
392 arch equal to b32 and one with b64 to make sure the kernel finds
393 the events that you intend. See the arch field discussion for
394 more info.
395 -w path
397 Insert a watch for the file system object at path. You cannot
398 insert a watch to the top level directory. This is prohibited by
399 the kernel. Wildcards are not supported either and will generate
400 a warning. The way that watches work is by tracking the inode
401 internally. If you place a watch on a file, its the same as us‐
402 ing the -F path option on a syscall rule. If you place a watch
403 on a directory, its the same as using the -F dir option on a
404 syscall rule. The -w form of writing watches is for backwards
405 compatibility and the syscall based form is more expressive. Un‐
406 like most syscall auditing rules, watches do not impact perfor‐
407 mance based on the number of rules sent to the kernel. The only
408 valid options when using a watch are the -p and -k. If you need
409 to do anything fancy like audit a specific user accessing a
410 file, then use the syscall auditing form with the path or dir
411 fields. See the EXAMPLES section for an example of converting
412 one form to another.
413 -W path
415 Remove a watch for the file system object at path. The rule must
416 match exactly. See -d discussion for more info.
417
419 Syscall rules get evaluated for each syscall for every program. If you
420 have 10 syscall rules, every program on your system will delay during a
421 syscall while the audit system evaluates each rule. Too many syscall
422 rules will hurt performance. Try to combine as many as you can whenever
423 the filter, action, key, and fields are identical. For example:
424 auditctl -a always,exit -F arch=b64 -S openat -F success=0
426 auditctl -a always,exit -F arch=b64 -S truncate -F success=0
427 could be re-written as one rule:
429 auditctl -a always,exit -F arch=b64 -S openat -S truncate -F success=0
431 Also, try to use file system auditing wherever practical. This improves
433 performance. For example, if you were wanting to capture all failed
434 opens & truncates like above, but were only concerned about files in
435 /etc and didn't care about /usr or /sbin, its possible to use this
436 rule:
437 auditctl -a always,exit -S openat -S truncate -F dir=/etc -F success=0
439 This will be higher performance since the kernel will not evaluate it
441 each and every syscall. It will be handled by the filesystem auditing
442 code and only checked on filesystem related syscalls.
443
445 To see all syscalls made by a specific program:
446 # By pid:
448 auditctl -a always,exit -S all -F pid=1005
449 # By executable path
450 auditctl -a always,exit -S all -F exe=/usr/bin/ls
451 To see files opened by a specific user:
453 auditctl -a always,exit -S openat -F auid=510
455 To see unsuccessful openat calls:
457 auditctl -a always,exit -S openat -F success=0
459 To watch a file for changes (2 ways to express):
461 auditctl -w /etc/shadow -p wa
463 auditctl -a always,exit -F path=/etc/shadow -F perm=wa
464 To recursively watch a directory for changes (2 ways to express):
466 auditctl -w /etc/ -p wa
468 auditctl -a always,exit -F dir=/etc/ -F perm=wa
469 To see if an admin is accessing other user's files:
471 auditctl -a always,exit -F dir=/home/ -F uid=0 -C auid!=obj_uid
473
476 On many systems auditd is configured to install an -a never,task rule
477 by default. This rule causes every new process to skip all audit rule
478 processing. This is usually done to avoid a small performance overhead
479 imposed by syscall auditing. If you want to use auditd, you need to re‐
480 move that rule by deleting 10-no-audit.rules and adding 10-base-con‐
481 fig.rules to the audit rules directory.
482 If you have defined audit rules that are not matching when they should,
484 check auditctl -l to make sure there is no never,task rule there.
485
488 /etc/audit/audit.rules /etc/audit/audit-stop.rules
489
492 audit.rules(7), ausearch(8), aureport(8), auditd(8).
493
496 Steve Grubb
497Red Hat July 2021 AUDITCTL(8)