1AUDIT.RULES(7) System Administration Utilities AUDIT.RULES(7)
2
6 audit.rules - a set of rules loaded in the kernel audit system
7
9 audit.rules is a file containing audit rules that will be loaded by the
10 audit daemon's init script whenever the daemon is started. The auditctl
11 program is used by the initscripts to perform this operation. The syn‐
12 tax for the rules is essentially the same as when typing in an auditctl
13 command at a shell prompt except you do not need to type the auditctl
14 command name since that is implied. The audit rules come in 3 vari‐
15 eties: control, file, and syscall.
16 Control
19 Control commands generally involve configuring the audit system rather
20 than telling it what to watch for. These commands typically include
21 deleting all rules, setting the size of the kernel's backlog queue,
22 setting the failure mode, setting the event rate limit, or to tell au‐
23 ditctl to ignore syntax errors in the rules and continue loading. Gen‐
24 erally, these rules are at the top of the rules file.
25 File System
28 File System rules are sometimes called watches. These rules are used to
29 audit access to particular files or directories that you may be inter‐
30 ested in. If the path given in a watch rule is a directory, then the
31 rule used is recursive to the bottom of the directory tree excluding
32 any directories that may be mount points. The syntax of these watch
33 rules generally follow this format:
34 -w path-to-file -p permissions -k keyname
36 where the permission are any one of the following:
38 r - read of the file
41 w - write to the file
43 x - execute the file
45 a - change in the file's attribute
47 Watches can also be created using the syscall format described below
49 which allow for greater flexibility and options. Using syscall rules
50 you can choose between path and dir which is against a specific inode
51 or directory tree respectively. It should also be noted that the recur‐
52 sive directory watch will stop if there is a mount point below the par‐
53 ent directory. There is an option to make the mounted subdirectory
54 equivalent by using a -q rule.
55 System Call
58 The system call rules are loaded into a matching engine that intercepts
59 each syscall that all programs on the system makes. Therefore it is
60 very important to only use syscall rules when you have to since these
61 affect performance. The more rules, the bigger the performance hit. You
62 can help the performance, though, by combining syscalls into one rule
63 whenever possible.
64 The Linux kernel has 6 rule matching lists or filters as they are some‐
66 times called. They are: task, exit, user, exclude, filesystem, and
67 io_uring. The task list is checked only during the fork or clone
68 syscalls. It is rarely used in practice.
69 The exit filter is the place where all syscall and file system audit
71 requests are evaluated.
72 The user filter is used to filter (remove) some events that originate
74 in user space. By default, any event originating in user space is al‐
75 lowed. So, if there are some events that you do not want to see, then
76 this is a place where some can be removed. See auditctl(8) for fields
77 that are valid.
78 The exclude filter is used to exclude certain events from being emit‐
80 ted. The msgtype and a number of subject attribute fields can be used
81 to tell the kernel which message types you do not want to record. This
82 filter can remove the event as a whole and is not selective about any
83 other attribute. The user and exit filters are better suited to selec‐
84 tively auditing events. The action is ignored for this filter, de‐
85 faulting to "never".
86 The io_uring filter is used to watch underlying syscalls performed by
88 io_uring operations.
89 Syscall rules take the general form of:
91 -a action,list -S syscall -F field=value -k keyname
93 The -a option tells the kernel's rule matching engine that we want to
95 append a rule at the end of the rule list. But we need to specify which
96 rule list it goes on and what action to take when it triggers. Valid
97 actions are:
98 always - always create an event
101 never - never create an event
103 The action and list are separated by a comma but no space in between.
105 Valid lists are: task, exit, user, exclude, filesystem, and io_uring.
106 Their meaning was explained earlier.
107 Next in the rule would normally be the -S option. This field can either
109 be the syscall name or number. For readability, the name is almost al‐
110 ways used. You may give more than one syscall in a rule by specifying
111 another -S option. When sent into the kernel, all syscall fields are
112 put into a mask so that one compare can determine if the syscall is of
113 interest. So, adding multiple syscalls in one rule is very efficient.
114 When you specify a syscall name, auditctl will look up the name and get
115 its syscall number. This leads to some problems on bi-arch machines.
116 The 32 and 64 bit syscall numbers sometimes, but not always, line up.
117 So, to solve this problem, you would generally need to break the rule
118 into 2 with one specifying -F arch=b32 and the other specifying -F
119 arch=b64. This needs to go in front of the -S option so that auditctl
120 looks at the right lookup table when returning the number.
121 After the syscall is specified, you would normally have one or more -F
123 options that fine tune what to match against. Rather than list all the
124 valid field types here, the reader should look at the auditctl man page
125 which has a full listing of each field and what it means. But it's
126 worth mentioning a couple things.
127 The audit system considers uids to be unsigned numbers. The audit sys‐
129 tem uses the number -1 to indicate that a loginuid is not set. This
130 means that when it's printed out, it looks like 4294967295. But when
131 you write rules, you can use either "unset" which is easy to remember,
132 or -1, or 4294967295. They are all equivalent. If you write a rule that
133 you wanted try to get the valid users of the system, you need to look
134 in /etc/login.defs to see where user accounts start. For example, if
135 UID_MIN is 1000, then you would also need to take into account that the
136 unsigned representation of -1 is higher than 500. So you would address
137 this with the following piece of a rule:
138 -F auid>=1000 -F auid!=unset
140 These individual checks are "anded" and both have to be true.
142 The last thing to know about syscall rules is that you can add a key
144 field which is a free form text string that you want inserted into the
145 event to help identify its meaning. This is discussed in more detail in
146 the NOTES section.
147
150 The purpose of auditing is to be able to do an investigation periodi‐
151 cally or whenever an incident occurs. A few simple steps in planning up
152 front will make this job easier. The best advice is to use keys in both
153 the watches and system call rules to give the rule a meaning. If rules
154 are related or together meet a specific requirement, then give them a
155 common key name. You can use this during your investigation to select
156 only results with a specific meaning.
157 When doing an investigation, you would normally start off with the main
159 aureport output to just get an idea about what is happening on the sys‐
160 tem. This report mostly tells you about events that are hard coded by
161 the audit system such as login/out, uses of authentication, system
162 anomalies, how many users have been on the machine, and if SE Linux has
163 detected any AVCs.
164 aureport --start this-week
166 After looking at the report, you probably want to get a second view
168 about what rules you loaded that have been triggering. This is where
169 keys become important. You would generally run the key summary report
170 like this:
171 aureport --start this-week --key --summary
173 This will give an ordered listing of the keys associated with rules
175 that have been triggering. If, for example, you had a syscall audit
176 rule that triggered on the failure to open files with EPERM that had a
177 key field of access like this:
178 -a always,exit -F arch=b64 -S open -S openat -S openat2 -F exit=-EPERM -k access
180 Then you can isolate these failures with ausearch and pipe the results
182 to aureport for display. Suppose your investigation noticed a lot of
183 the access denied events. If you wanted to see the files that unautho‐
184 rized access has been attempted, you could run the following command:
185 ausearch --start this-week -k access --raw | aureport --file --summary
187 This will give an ordered list showing which files are being accessed
189 with the EPERM failure. Suppose you wanted to see which users might be
190 having failed access, you would run the following command:
191 ausearch --start this-week -k access --raw | aureport --user --summary
193 If your investigation showed a lot of failed accesses to a particular
195 file, you could run the following report to see who is doing it:
196 ausearch --start this-week -k access -f /path-to/file --raw | aureport
198 --user -i
199 This report will give you the individual access attempts by person. If
201 you needed to see the actual audit event that is being reported, you
202 would look at the date, time, and event columns. Assuming the event was
203 822 and it occurred at 2:30 on 09/01/2009 and you use the en_US.utf8
204 locale, the command would look something like this:
205 ausearch --start 09/01/2009 02:30 -a 822 -i --just-one
207 This will select the first event from that day and time with the match‐
209 ing event id and interpret the numeric values into human readable val‐
210 ues.
211 The most important step in being able to do this kind of analysis is
213 setting up key fields when the rules were originally written. It should
214 also be pointed out that you can have more than one key field associ‐
215 ated with any given rule.
216
219 If you are not getting events on syscall rules that you think you
220 should, try running a test program under strace so that you can see the
221 syscalls. There is a chance that you might have identified the wrong
222 syscall.
223 If you get a warning from auditctl saying, "32/64 bit syscall mismatch
225 in line XX, you should specify an arch". This means that you specified
226 a syscall rule on a bi-arch system where the syscall has a different
227 syscall number for the 32 and 64 bit interfaces. This means that on one
228 of those interfaces you are likely auditing the wrong syscall. To solve
229 the problem, re-write the rule as two rules specifying the intended
230 arch for each rule. For example,
231 -a always,exit -S openat -k access
233 would be rewritten as
235 -a always,exit -F arch=b32 -S openat -k access
237 -a always,exit -F arch=b64 -S openat -k access
238 If you get a warning that says, "entry rules deprecated, changing to
240 exit rule". This means that you have a rule intended for the entry fil‐
241 ter, but that filter is no longer available. Auditctl moved your rule
242 to the exit filter so that it's not lost. But to solve this so that you
243 do not get the warning any more, you need to change the offending rule
244 from entry to exit.
245
248 The following rule shows how to audit failed access to files due to
249 permission problems. Note that it takes two rules for each arch ABI to
250 audit this since file access can fail with two different failure codes
251 indicating permission problems.
252 -a always,exit -F arch=b32 -S open -S openat -S openat2 -F exit=-EACCES -k access
254 -a always,exit -F arch=b32 -S open -S openat -S openat2 -F exit=-EPERM -k access
255 -a always,exit -F arch=b64 -S open -S openat -S openat2 -F exit=-EACCES -k access
256 -a always,exit -F arch=b64 -S open -S openat -S openat2 -F exit=-EPERM -k access
257
260 Io_uring rules do not take an arch field. It is implicit in the speci‐
261 fication of the filter. The following example rule watches for file
262 opens through the io_uring subsystem:
263 -a always,io_uring -S openat -S openat2 -F key=access
265
268 If auditing is enabled, then you can get any event that is not caused
269 by syscall or file watch rules (because you don't have any rules
270 loaded). So, that means, any event from 1100-1299, 1326, 1328, 1331 and
271 higher can be emitted. The reason that there are a number of events
272 that are hardwired is because they are required by regulatory compli‐
273 ance and are sent automatically as a convenience. (For example, lo‐
274 gon/logoff is a mandatory event in all security guidance.) If you don't
275 want this, you can use the exclude filter to drop events that you do
276 not want.
277 -a always,exclude -F msgtype=CRED_REFR
279
282 auditctl(8), auditd(8).
283
286 Steve Grubb
287Red Hat Feb 2023 AUDIT.RULES(7)