fapolicyd.rules - Man Page
compiled fapolicyd rules to determine access rights fapolicyd.rules — deprecated fapolicyd rules to determine access rights
compiled.rules is a file that is compiled by fagenrules which contains the rules that fapolicyd uses to make decisions about access rights. The rules follow a simple format of:
decision perm subject : object
They are evaluated from top to bottom with the first rule to match being used for the access control decision. The colon is mandatory to separate subject and object since they share keywords.
The decision is either allow, deny, allow_audit, deny_audit, allow_syslog, deny_syslog, allow_log, or deny_log. If the rule triggers, this is the access decision that fapolicyd will tell the kernel. If the decision is one of the audit variety, then the decision will trigger a FANOTIFY audit event with all relevant information. You must have at least one audit rule loaded to generate an audit event. If the decision is one of the syslog variety, then the decision will trigger writing an event into syslog. If the decision is of one the log variety, then it will create an audit event and a syslog event. Regardless of the notification, any rule with a deny in the keyword will deny access and any with an allow in the keyword will allow access.
Perm describes what kind permission is being asked for. The permission is either open, execute, or any. If none are given, then open is assumed.
The subject is the process that is performing actions on system resources. The fields in the rule that describe the subject are written in a name=value format. There can be one or more subject fields. Each field is and'ed with others to decide if a rule triggers. The name values can be any of the following:
This matches against any subject. When used, this must be the only subject in the rule.
This is the login uid that the audit system assigns users when they log in to the system. Daemons have a value of -1. The given value may be numeric or the account name.
This is the user id that the program is running under. The given value may be numeric or the account name.
This is the group id that the program is running under. The given value may be numeric or the group name.
This is the numeric session id that the audit system assigns to users when they log in. Daemons have a value of -1.
This is the numeric process id that a program has.
This is the numeric process id of the program's parent. Note that programs that are orphaned or started directly from systemd have a ppid value of 1. Kernel threads have a ppid value of 2.
This is a boolean describing whether it is required for the subject to be in the trust database or not. A value of 1 means its required while 0 means its not. Trust checking is extended by the integrity setting in fapolicyd.conf. When trust is used on the subject, it could be a daemon. If that daemon gets updated on disk, the trustdb will be updated to the new SHA256 hash. If the integrity setting is not none, the running daemon is not likely to be trusted unless it gets restarted. The default rules are not written in a way that this would happen. But this needs to be highlighted as it may not be obvious when writing a new rule.
This is the shortened command name. When an interpreter starts a program, it usually renames the program to the script rather than the interpreter.
This is the full path to the executable. Globbing is not supported. You may also use the special keyword untrusted to match on the subject not being listed in the rpm database.
If you wish to match a directory, then use this by giving the full path to the directory. Its recommended to end with the / to ensure it matches a directory. There are 3 keywords that dir supports: execdirs, systemdirs, untrusted.
The execdirs option will match against the following list of directories:
/bin/ /sbin/ /lib/ /lib64/ /usr/libexec/
The systemdirs option will match against the same list as execdirs but also includes /etc/.
The untrusted option will look up the current executable's full path in the rpm database to see if the executable is known to the system. The rule will trigger if the file in question is not in the trust database. This option is deprecated in favor of using obj_trust with execute permission when writing rules.
This option takes the mime type of a file as an argument. If you wish to check the mime type of a file while writing rules, run the following command:
fapolicyd-cli --ftype /path-to-file
This option will match against the device that the executable resides on. To use it, start with /dev/ and add the target device name.
There are various ways that an attacker may try to execute code that may reveal itself in the pattern of file accesses made during program startup. This rule can take one of several options depending on which access patterns is wished to be blocked. Fapolicyd is able to detect these different access patterns and provide the access decision as soon as it identifies the pattern. The pattern type can be any of:
This matches against any ELF program that is dynamically linked.
This matches against access patterns that indicate that the program is being started directly by the runtime linker.
This matches against access patterns that indicate that the program is being started with either LD_PRELOAD or LD_AUDIT present in the environment. Note that even without this rule, you have protection against LD_PRELOAD of unknown binaries when the rules are written such that trust is used to determine if a library should be opened. In that case, the preloaded library would be denied but the application will still execute. This rule makes it so that even trusted libraries can be denied and the application will not execute.
This matches against ELF files that are statically linked.
The object is the file that the subject is interacting with. The fields in the rule that describe the object are written in a name=value format. There can be one or more object fields. Each field is and'ed with others to decide if a rule triggers. The name values can be any of the following:
This matches against any obbject. When used, this must be the only object in the rule.
This is the full path to the file that will be accessed. Globbing is not supported. You may also use the special keyword untrusted to match on the object not being listed in the rpm database.
If you wish to match on access to any file in a directory, then use this by giving the full path to the directory. Its recommended to end with the / to ensure it matches a directory. There are 3 keywords that dir supports: execdirs, systemdirs, untrusted. See the dir option under Subject for an explanation of these keywords.
This option will match against the device that the file being accessed resides on. To use it, start with /dev/ and add the target device name.
This option matches against the mime type of the file being accessed. See ftype under Subject for more information on determining the mime type.
This is a boolean describing whether it is required for the object to be in the trust database or not. A value of 1 means its required while 0 means its not. Trust checking is extended by the integrity setting in fapolicyd.conf.
This option matches against the sha256 hash of the file being accessed. The hash in the rules should be all lowercase letters and do NOT start with 0x. Lowercase is the default output of sha256sum.
Set is a named group of values of the same type. Fapolicyd internally distinguishes between INT and STRING set types. You can define your own set and use it as a value for a specific rule attribute. The definition is in key=value syntax and starts with a set name. The set name has to start with '%' and the rest is alphanumeric or '_'. The value is a comma separated list. The set type is inherited from the first item in the list. If that can be turned into number then whole list is expected to carry numbers. One can use these sets as a value for subject and object attributes. It is also possible to use a plain list as an attribute value without previous definition. The assigned set has to match the attribute type. It is not possible set groups for TRUST and PATTERN attributes.
# definition %python=/usr/bin/python2.7,/usr/bin/python3.6 allow exe=%python : all trust=1 # # definition # number set %uuids=0,1000 allow uid=%uuids : all
When writing rules, you should keep them focused to one goal and store them in one file. These rule files are kept in the /etc/fapolicyd/rules.d directory. During daemon startup, fagenrules will run and compile all these component files into one master file, compiled.rules. See the fagenrules man page for more information.
When you are writing a rule for the execute permission, remember that the file to be executed is an object. For example, you type ssh into the shell. The shell calls execve on /usr/bin/ssh. At that instant in time, ssh is the object that bash is working on. However, if you are blocking execution from a specific program, then you would normally state the program on the subject side and use all for the object side.
If you are writing rules that use patterns, just select any as the permission to be clear that this applies to anything. In reality, pattern matching ignores the permission but the suggestion is for documentation purposes.
Some interpreters do not immediately read all lines of input. Rather, they read content as needed until they get to end of file. This means that if they do stuff like networking or sleeping or anything that takes time, someone with the privileges to modify the file can add to it after the file's integrity has been checked. This is not unique to fapolicyd, it's simply how things work. Make sure that trusted file permissions are not excessive so that no unexpected file content modifications can occur.
The following rules illustrate the rule syntax.
deny_audit perm=open exe=/usr/bin/wget : dir=/tmp allow perm=open exe=/usr/bin/python3.7 : ftype=text/x-python trust=1 deny_audit perm=any pattern ld_so : all deny perm=any all : all
fapolicyd(8), fagenrules (8), fapolicyd-cli(8), and fapolicyd.conf(5)
fagenrules(8), fapolicy-analyzer(8), fapolicyd(8), fapolicyd-cli(8).