Rules

• Format
• Performance / Important notes
• Best practices

---

Format

Rules are stored as JSON files inside the -rule-path folder, in the simplest case a rule looks like this:

{
   "created": "2018-04-07T14:13:27.903996051+02:00",
   "updated": "2018-04-07T14:13:27.904060088+02:00",
   "name": "deny-simple-www-google-analytics-l-google-com",
   "enabled": true,
   "precedence": false,
   "action": "deny",
   "duration": "always",
   "operator": {
     "type": "simple",
     "sensitive": false,
     "operand": "dest.host",
     "data": "www-google-analytics.l.google.com"
   }
}

Field	Description
created	UTC date and time of creation.
update	UTC date and time of the last update.
name	The name of the rule.
enabled	Use to temporarily disable and enable rules without moving their files.
precedence	true or false. Sets if a rule take precedence (>= v1.2.0)
action	Can be deny, reject or allow.
duration	For rules persisting on disk, this value is default to always.
operator.type	Can be simple, in which case a simple == comparison will be performed, regexp if the data field is a regular expression to match, network which will match a network range (127.0.0.1/8), lists which will look for matches on lists of something (domains, IPs, etc), or list, which is a combination of all of the types.
operator.operand	What element of the connection to compare, can be one of:
	* true (will always match)
	* process.path (the path of the executable)
	* process.id PID
	* process.command (full command line, including path and arguments)
	* provess.env.ENV_VAR_NAME (use the value of an environment variable of the process given its name)
	* user.id (UID)
	* protocol
	* dest.ip
	* dest.host
	* dest.network (>= v1.3.0)
	* dest.port
	* lists.domains (>= 1.4.0) lists of domains in hosts format read more
	* lists.domains_regexp (>= 1.5.0) list of domains with regular expressions (.*\.example\.com) read more
	* lists.ips (>= 1.5.0) list of IPs read more
	* lists.nets (>= 1.5.0) list of network ranges read more
operator.data	The data to compare the operand to, can be a regular expression if type is regexp, or a path to a directory with list of IPs/domains in the case of lists.

Some considerations

• All the fields you select when defining a rule will be used to match connections, for example:

• Rule: allow -> port 443 -> Dst IP 1.1.1.1 -> Protocol TCP -> Host www.site.test

  • This rule will match connections to port 443 AND IP 1.1.1.1 AND protocol TCP AND host www.site.test
  • connections to IP 2.2.2.2 won't match, connections to port 80 won't match, etc...

• Rule: allow -> port 53 -> [x] domains list -> [x] network ranges list

  • This rule will match connections to port 53 AND domains in the list AND IPs in the network ranges list

• Rule: allow -> port ^(53|80|443)$ -> UID 1000 -> Path /app/bin/test -> [x] domains list

  • This rule will match connections to ports (53 OR 80 OR 443) AND UID 1000 AND Path /app/bin/test AND domains in the specified.

• If you select multiple lists on the same rule, bear in mind that the connections you want to match must Read this disccussion to learn more

• By default Deny rules take precedence over the rest of the rules. If a connection match a Deny rule, opensnitch won't continue evaluating rules.

• Since v1.2.0, rules are sorted and checked in alphabetical order. You can name them this way to prioritize Deny rules, for example:

000-allow-chrome-to-specific-domains
001-allow-not-so-important-rule
001-deny-chrome

• Also since v1.2.0, you can configure a rule as Important ([x] Priority) to take precedence over the rest of the rules. If you set this flag and name the rule as mentioned above, you can also prioritize Allow rules:

000-allow-chrome-to-specific-domains [x] Priority <-- if the connection matches this rule, it'll allow this rule and won't continue evaluating the rest of rules.
001-allow-not-so-important-rule
001-deny-chrome

This way you can not only prioritize critical connections (like VPNs), but also gain performance.

More on rules performance

As already mentioned, the order of the rule is critical. If you prioritize Firefox the web navegation will be faster.

But the type of rule also impacts the rules performance. regexp and list types are slower than simple, in the end, regexp and list types check multiple parameters while simple rules check just one.

---

An example with a regular expression:

{
   "created": "2018-04-07T14:13:27.903996051+02:00",
   "updated": "2018-04-07T14:13:27.904060088+02:00",
   "name": "deny-any-google-analytics",
   "enabled": true,
   "precedence": false,
   "action": "deny",
   "duration": "always",
   "operator": {
     "type": "regexp",
     "sensitive": false,
     "operand": "dest.host",
     "data": "(?i)
   }
}

An example whitelisting a process path:

{
   "created": "2018-04-07T15:00:48.156737519+02:00",
   "updated": "2018-04-07T15:00:48.156772601+02:00",
   "name": "allow-simple-opt-google-chrome-chrome",
   "enabled": true,
   "precedence": false,
   "action": "allow",
   "duration": "always",
   "operator": {
     "type": "simple",
     "sensitive": false,
     "operand": "process.path",
     "data": "/opt/google/chrome/chrome"
   }
 }

Example of a complex rule using the operator list, saved from the GUI (Note: version v1.2.0):

{
  "created": "2020-02-07T14:16:20.550255152+01:00",
  "updated": "2020-02-07T14:16:20.729849966+01:00",
  "name": "deny-list-type-simple-operand-destip-data-1101-type-simple-operand-destport-data-23-type-simple-operand-userid-data-1000-type-simple-operand-processpath-data-usrbintelnetnetkit",
  "enabled": true,
  "precedence": false,
  "action": "deny",
  "duration": "always",
  "operator": {
    "type": "list",
    "operand": "list",
    "data": "[{\"type\": \"simple\", \"operand\": \"dest.ip\", \"data\": \"1.1.0.1\"}, {\"type\": \"simple\", \"operand\": \"dest.port\", \"data\": \"23\"}, {\"type\": \"simple\", \"operand\": \"user.id\", \"data\": \"1000\"}, {\"type\": \"simple\", \"operand\": \"process.path\", \"data\": \"/usr/bin/telnet.netkit\"}]",
    "list": [
      {
        "type": "simple",
        "operand": "dest.ip",
        "sensitive": false,
        "data": "1.1.0.1",
        "list": null
      },
      {
        "type": "simple",
        "operand": "dest.port",
        "sensitive": false,
        "data": "23",
        "list": null
      },
      {
        "type": "simple",
        "operand": "user.id",
        "sensitive": false,
        "data": "1000",
        "list": null
      },
      {
        "type": "simple",
        "operand": "process.path",
        "sensitive": false,
        "data": "/usr/bin/telnet.netkit",
        "list": null
      }
    ]
  }
}

Best practices

• Allow DNS queries only to your configured DNS nameservers:

  ⚠️ DNS protocol can be used to exfiltrate information from local networks.

  • Allow systemd-resolved, dnsmasq, dnscrypt-proxy, etc, connect only to your DNS nameservers + port 53 + UID.
  • Besides allowing connections to remote DNS servers (9.9.9.9 for example), you may need to allow connections to localhost IPs (127.0.0.1, etc)
  • If you already allowed these stub resolvers, the easiest way would we to delete the existing rule, let it ask you again to allow/deny it, click on the [+] button and then select from the pop-up from this command line AND to IP x.x.x.x AND_ to port xxx

• Limit what an application can do as much as possible:

  • Filter by executable + command line: You don't want to allow curl or wget system wide. Instead, allow only a particular command line, for example:

    command launched: $ wget https://mirror.karneval.cz/pub/linux/fedora/linux/releases/34/Workstation/x86_64/iso/Fedora-Workstation-Live-x86_64-34-1.2.iso

    Instead of allowing from this executable: wget, use allow from this executable + from this command line

    You can narrow it further, by allowing from this command line + from this User ID + to this IP + to this port

• Don't allow python3, perl or ruby binaries system-wide:

  • As explained above, filter by executable + command line + (... more parameters ...) If you allow python3for example, you'll allow ANY python3 script, so be careful.

    https://github.com/evilsocket/opensnitch/wiki/Rules-examples#filtering-python-scripts-applicable-to-java-and-others-interpreters

• Disable unprivileged namespaces to prevent rules bypass

  If /proc/sys/kernel/unprivileged_userns_clone is set to 1, change it to 0. Until we obtain the checksum of a binary, it's better to set it to 0.

• Don't allow connections opened by binaries located under certain directories: /dev/shm, /tmp, /var/tmp

  Why? If someone gets access to your system, usually these directories are the only ones where they can write files, thus it's usually used to drop malicious files, that download remote binaries to escalate privileges, etc.

  There're ton of examples [0] [1] (more common on servers than on the desktop): https://github.com/timb-machine/linux-malware

  [0]. https://www.akamai.com/blog/security-research/kmdsbot-the-attack-and-mine-malware

  • https://www.akamai.com/site/en/images/blog/2022/kmsdbot1.jpg [1]. https://www.elastic.co/guide/en/security/master/binary-executed-from-shared-memory-directory.html

  (*) Deny
  [x] From this executable: ^(/tmp/|/var/tmp/|/dev/shm/|/var/run|/var/lock).*
  

  Note that the default policy should be deny everything unless explicitely allowed. But by creating a rule to deny specifically these directories, you can have a place where to monitor these executions.