Rspamd user settings


Rspamd can apply different settings for messages scanned. Each setting can define a set of custom metric weights, symbols or actions scores and enable or disable certain checks. Rspamd settings can be loaded as dynamic maps and updated automatically if a corresponding file or URL has changed since its last update.

To load settings as a dynamic map, you can set ‘settings’ to a map string:

settings = "http://host/url"

If you don’t want dynamic updates then you can define settings as an object:

settings {
	setting1 = {
	setting2 = {

To define static settings, you might want to edit local.d/settings.conf file (from Rspamd 1.8). If you want to use dynamic map for settings, it might be better to define it in the override file: rspamd.conf.override:

settings = "http://host/url"

Alternatively, settings apply part (see later) could be passed to Rspamd by a client by query parameter:

POST /scanv2?settings="{symbol1 = 10.0}" HTTP/1.0

or HTTP header

POST /scanv2 HTTP/1.0
Settings: {symbol1 = 10.0}

Settings could also be indexed by ID, allowing to select a specific setting without checking for its conditions. For example, this feature could be used to split inbound and outbound mail flows by specifying different rules set from the MTA side. Another use case of settings id option is to create a dedicated lightweight checks for certain conditions, for example DKIM checks.

Important note: using settings ID is optimal from the terms of performance.

Let’s assume that we have the following settings in the configuration that have id dkim:

# local.d/settings.conf
dkim {
	id = "dkim";
	apply {
		groups_enabled = ["dkim"];

Afterwards, if we send a request with this settings id using HTTP protocol:

POST /scanv2 HTTP/1.0
Settings-ID: dkim

then Rspamd won’t check all rules but DKIM ones. Alternatively, you could check this setup using rspamc command:

rspamc --header="settings-id=dkim" message.eml

Settings structure

The settings file should contain a single section called “settings”:

# local.d/settings.conf
some_users {
	id = "some_users";
	priority = high;
	from = "";
	rcpt = "admin";
	rcpt = "/user.*/";
	ip = "";
	user = "";
	request_header = {
		"MTA-Tag" = "\.example\.net$";
	apply {
		symbol1 = 10.0;
		symbol2 = 0.0;
		actions {
			reject = 100.0;
			greylist = null; # Disable greylisting (from 1.8.1)
			"add header" = 5.0; # Please note the space, NOT an underscore
	# Always add these symbols when settings rule has matched
	symbols [
		"symbol2", "symbol4"
whitelist {
	priority = low;
	rcpt = "";
	want_spam = yes;
# Disable some checks for authenticated users
authenticated {
	priority = high;
	authenticated = yes;
	apply {
		groups_disabled = ["rbl", "spf"];

So each setting has the following attributes:

  • name - section name that identifies this specific setting (e.g. some_users)
  • priority - high (3), medium (2), low (1) or any positive integer value (default priority is low). Rules with greater priorities are matched first. From version 1.4 Rspamd checks rules with equal priorities in alphabetical order. Once a rule matches only that rule is applied and the rest are ignored.
  • match list - list of rules which this rule matches:
    • from - match SMTP sender
    • from_mime - match MIME sender
    • rcpt - match SMTP recipient
    • rcpt_mime - match MIME recipient
    • ip - match source IP address
    • hostname - match the source hostname (regexp supported)
    • user - matches authenticated user ID of message sender if any
    • authenticated - matches any authenticated user
    • local - matches any local IP
    • request_header - collection of request header names and regexes to match them against (condition is satisfied if any match)
    • header - collection of MIME message header names and regexes to match them against (condition is satisfied if any match), available since Rspamd 1.7
    • selector - apply the specific selector to check if we need to apply these settings. If selector returns non-nil, then the settings are applied (selector’s value is ignored so far). Available since Rspamd 1.8.
  • apply - list of applied rules
    • symbol - modify weight of a symbol
    • actions - defines actions
    • symbols_enabled - array of symbols that should be checked (all other rules are disabled)
    • groups_enabled - array of rules groups that should be checked (all other rules are disabled)
    • symbols_disabled - array of disabled checks by symbol name (all other rules are enabled)
    • groups_disabled - array of disabled checks by group name (all other rules are enabled)
    • subject - set subject based on the new pattern: %s is replaced with the existing subject, %d is replaced with the message’s spam score (e.g. subject = "SPAM: %s (%d)")
  • symbols - add symbols from the list if a rule has matched
  • inverse - inverse match (e.g. it will NOT match when all elements are matched and vice-versa)

If symbols_enabled or groups_enabled are found in apply element, then Rspamd disables all checks with the exception of the enabled ones. When enabled and disabled options are both presented, then the precedence of operations is the following:

  1. Disable all symbols
  2. Enable symbols from symbols_enabled and groups_enabled
  3. Disable symbols from symbols_disabled and groups_disabled

Some rules, such as metadata exporter, history redis or clickhouse are marked as explicit_disable. It means that even if you set some specific symbols in symbols_enabled these rules will still be executed. This is normally what is expected: enabling specific checks should not interfere with the data exporting/history.

Important notice: This is NOT applicable to want_spam option. This option disable ALL Rspamd rules, even history or data exporting. Actually, it is a full bypass of all Rspamd processing.

Settings match

The match section performs AND operation on different matches: for example, if you have from and rcpt in the same rule, then the rule matches only when from AND rcpt match. For similar matches, the OR rule applies: if you have multiple rcpt matches, then any of these will trigger the rule. If a rule is triggered then no more rules are matched.

By default, regular expressions are case-sensitive. This can be changed with the i flag. Regexp rules can be slow and should not be used extensively.

In order to make matching case-insensitive, string comparisons convert input strings to lowercase. Thus, strings in the match lists should always be in lowercase.

The picture below describes the architecture of settings matching.

Redis settings

Storing settings in Redis provides a very flexible way to apply settings & avoids the need to reload a map.

To use settings in Redis we write one or more handlers in Lua, each of which might return a key. If a key is returned, and it exists in Redis, the value of the key is used as settings. This value should be formatted as in the contents of the apply block or settings posted in headers.

Let’s presume that we want to base our settings on the domain of the first SMTP recipient.

We could set our keys as follows> SET "" "{symbol1 = 5000;}"

Where “setting:” is a prefix we have chosen for our settings and “” is the recipient domain we want to apply settings to and the value of the key contains our desired settings.

We would then define configuration as follows in /etc/rspamd/rspamd.conf.override:

# Redis settings are configured in a "settings_redis" block
settings_redis {
  # Here we will define our Lua functions
  handlers = {
    # Everything in here is a Lua function with an arbitrary name
my_check_rcpt_domain = <<EOD
return function(task)
  local rcpt = task:get_recipients('smtp')
  -- Return nothing if we can't find domain of first SMTP recipient
  if not (rcpt and rcpt[1] and rcpt[1]['domain']) then return end
  -- Return "setting:" concatenated with the domain
  local key = 'setting:' .. rcpt[1]['domain']
  return key
  -- From Rspamd 1.6.3 this function can return a list of keys to check.
  -- Use this if you need to check for settings according to priority:
  return {key, 'setting:global'}

Redis servers are configured as per usual - see here for details.