You're Invited:Meet the Socket Team at BlackHat and DEF CON in Las Vegas, Aug 7-8.RSVP
Sign inDemoInstall


Package Overview
File Explorer

Advanced tools

Install Socket

Detect and block malicious and high-risk dependencies



enforce various types of limits on remote MTAs

Version published



[1.2.5] - 2024-07-16

  • fix: remove expired rate_conn hash keys (#66)




Tests Code Climate


Apply many types of limits to SMTP connections:

- concurrent connections
- max recipients
- max unrecognized commands
- max SMTP errors
- outbound concurrency
- rate limits
    - max connections / period
    - max recipients / period
        - by host
        - by sender
    - max null recipients / period


cd /etc/haraka
npm i haraka-plugin-limit
grep -q ^redis config/plugins || echo 'redis' >> config/plugins
echo 'limit' >> config/plugins


Each limit type has values that can be defined in limit.ini. See the default limit.ini in this packages config directory.

Each limit type is disabled until enabled=true is set within its block in limit.ini.

Haraka's config loader loads the defaults from limit.ini within this plugins installed config directory and applies any overrides found in the limit.ini within your Haraka install/config directory.


  • tarpit_delay = seconds (optional)

Set this to the length in seconds that you want to delay every SMTP response to a remote client that has exceeded the rate limits.


Redis is the cluster-safe storage backend for maintaining the counters necessary to impose limits reliably.

  • host (default:
  • port (default: 6379)
  • db (default: 0)

If this [redis] section or any values are missing, the defaults from redis.ini are used.


When [concurrency]max is defined, it limits the maximum number of simultaneous connections per IP address. Connection attempts in excess of the limit are optionally delayed before being disconnected.

This works well in conjunction with a history / reputation database, so that one can assign very low concurrency (1) to bad or unknown senders and higher limits for reputable mail servers.


History: when enabled, the history setting is the name of a plugin that stores IP history / reputation results. The result store must have a positive value for good connections and negative integers for poor / undesirable connections. Karma is one such plugin.


When [recipients]max is defined, each connection is limited to that number of recipients. The limit is imposed against all recipient attempts. Attempts in excess of the limit are issued a temporary failure.


When [unrecognized_commands]max is set, a connection that exceeeds the limit is disconnected.

Unrecognized commands are normally SMTP verbs invalidly issued by the client. Examples:

  • issuing AUTH when we didn't advertise AUTH extension
  • issuing STARTTLS when we didn't advertise STARTTLS
  • invalid SMTP verbs


The unrecognized_command hook is used by the tls and auth plugins, so running this plugin before those would result in valid operations getting counted against that connections limits. The solution is simple: list limit in config/plugins after those.


When [errors]max is set, a connection that exceeeds the limit is disconnected. Errors that count against this limit include:

  • issuing commands out of turn (MAIL before EHLO, RCPT before MAIL, etc)
  • attempting MAIL on port 465/587 without AUTH
  • MAIL or RCPT addresses that fail to parse

Rate Limits

By default DENYSOFT will be returned when rate limits are exceeded. You can also tarpit the connection delaying every response.

Missing sections disable that particular test.

They all use a common configuration format:

  • <lookup> = <limit>[/time[unit]] (optional)

'lookup' is based upon the limit being enforced and is either an IP address, rDNS name, sender address or recipient address either in full or part.

The lookup order is as follows and the first match in this order is returned and is used as the record key in Redis (except for 'default' which always uses the full lookup for that test as the record key):

IPv4/IPv6 address or rDNS hostname:


Sender or Recipient address:

In all tests 'default' is used to specify a default limit if nothing else has matched.

'limit' specifies the limit for this lookup. Specify 0 (zero) to disable limits on a matching lookup.

'time' is optional and if missing defaults to 60 seconds. You can optionally specify the following time units (case-insensitive):

  • s (seconds)
  • m (minutes)
  • h (hours)
  • d (days)


This section limits the number of connections per interval from a given host or set of hosts.

IP and rDNS names are looked up by this test.


This section limits the number of recipients per interval from a given host or set of hosts.

IP and rDNS names are looked up by this test.


This section limits the number of recipients per interval from a sender or sender domain.

The sender is looked up by this test.


This section limits the rate which a recipient or recipient domain can receive messages over an interval.

Each recipient is looked up by this test.


This section limits the rate at which a recipient can receive messages from a null sender (e.g. DSN, MDN etc.) over an interval.

Each recipient is looked up by this test.


enabled=true ; delay=30 ;

The number after the domain is the maximum concurrency limit for that domain.

Delay is the number of seconds to wait before retrying this message. Outbound concurrency is checked on every attempt to deliver.


Applying strict connection and rate limits is an effective way to reduce spam delivery. It's also an effective way to inflict a stampeding herd on your mail server. When spam/malware is delivered by MTAs that have queue retries, if you disconnect early (which the rate limits do) with a 400 series code (a sane default), the remote is likely to try again. And again. And again. And again. This can cause an obscene rise in the number of connections your mail server handles. Plan a strategy for handling that.


  • Don't enforce limits early. I use karma and wait until DATA before disconnecting. By then, the score of the connection is determinate and I can return a 500 series code telling the remote not to try again.
  • enforce rate limits with your firewall instead



Package last updated on 17 Jul 2024

Did you know?


Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.


Related posts

SocketSocket SOC 2 Logo


  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog


Stay in touch

Get open source security insights delivered straight into your inbox.

  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc