npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2024 – Pkg Stats / Ryan Hefner

haraka-plugin-karma

v2.1.6

Published

A heuristics scoring and reputation engine for SMTP connections

Downloads

5,495

Readme

Build Status Code Climate NPM

Karma - A heuristics based reputation engine for the Haraka MTA

Karma is a heuristic scoring engine that uses connection metadata and other Haraka plugin data as inputs. Connections scoring in excess of specified thresholds are rewarded or penalized in proportionate ways.

Description

Haraka includes some excellent plugins that detect message or sender patterns that are indicative of spam. Some plugins have accuracy rates above 95%. Detecting messages incorrectly 5% of the time means that some portion of ham will get blocked and some portion of spam will get delivered. The traditional solution is to permit users to specify their filtering aggressiveness, and then suffer the consequences (blocked ham -vs- too much spam).

By scoring and weighting several "95+" plugins, accuracy rates above 99% are attainable. With a half dozen such plugins, 99.99% accuracy is attainable. With karma, good senders with good history can occasionally fail tests (false positives) and still deliver their mail. Senders with poor history will have a much harder time.

Karma Prereqs

One challenge for mail filtering is that filters (or plugins, in Haraka's case) are usually called in sequence. A message passes from one filter to the next in a pipeline and each filter gets to choose "yay or nay." In order for karma to transcend this limit, karma requires that every filter passes every message. Nearly all Haraka plugins have a reject or deny option for this reason.

In order to score a plugins results, plugins must save their results to the Result Store. Karma will see that and apply the awards specified in karma.ini.

How Karma Works

Karma takes a holistic view of connections. During the connection, karma collects these results and applies the result_awards defined in karma.ini. Once a connection/message exceeds the threshold.negative score (default: -8), karma rejects it at the next [deny]hook.

The scoring mechanism is not dissimilar to SpamAssassin, but Karma has some particular advantages:

* Runs entirely in Node, so it's very fast
* Async and very scalable.
* Builds sender and network reputation databases
* Has access to connection properties (relaying, port, auth attempts, etc..)
* Access to raw SMTP commands (data + formatting inspection)
* Can reject connections before DATA (save lots of bandwidth)

Karma is not a replacement for content filters. Karma focuses on the quality of the connection. Content filters (bayes*) focus on the content of the message. Karma works best with content filters.

CONFIG

See config/karma.ini for options and inline documentation.

AWARDS

Karma allows the site administrator to control how much weight to assign to plugin results, providing a great deal of control over what results are worth rejecting for.

Karma begins scoring the connection when the first packet arrives. The IP reputation, sender OS, GeoIP location, DNSBL listing, and FCrDNS are often a sufficient basis for rejecting a connection without ever blocking a ham.

Karma performs checks early and often, maximizing the penality it can exact upon bad mailers.

Penalties

Deny / Reject

When connections become worse than [thresholds]negative, they are denied during the next [deny]hook.

History

Karma history is computed as the number of good - bad connections.

When each connection ends, karma records the result. When a sufficient history has been built for an IP or ASN, future connections from that address(es) will get a positive or negative karma award.

The reward is purposefully small, to permit good senders in bad neighborhoods to still send.

Connection Delays

Connection delays (aka tarpitting, teergrubing, early talker) slow down a SMTP conversation by inserting artificial delays. Karma increases connection delays adaptively as connection quality changes.

Karma's delay goals:

1. Don't delay valid senders
2. Penalize senders in proportion to their karma score
3. Dampen bruteforce AUTH attacks.
4. Since the only *cost* we can exact from miscreants is time, and connections are cheap to maintain, keep miscreants online as long as possible.

There are three tarpit options:

[tarpit]
* delay=0   (seconds to delay,      default: 0)
* max=5     (max seconds to delay,  default: 5)

When set to zero, the value of the delay is adaptive, calculated proportional to the karma score of the connection. Connections with good karma will see no delay and bad ones will see long delays.

When delay is non-zero, each SMTP response will be delayed by that many seconds.

In practice, most naughty senders abandon the connection when forced to wait more than a handful of seconds. max sets the maximum delay between responses.

When using karma, do not use Haraka's tarpit plugin.

Included Tests

Connection data that karma considers:

The data from these tests are helpful but the real power of karma is scoring the results of other plugins. See karma.ini for a rich set of examples.

IP Reputation

Karma records the number of good, bad, and total connections. The results are accessible to other plugins as well.

var karma = connection.results.get('karma');

The karma result object contains at least the following:

score: 0,         <- score for this connection
history: 0,       <- score for all connections
   good: 0,       <- qty of past 'good' connections
   bad: 0,        <- qty of past 'bad' connections
connections: 0,   <- total past connections
pass: [],         <- tests that added positive karma
fail: [],         <- tests that added negative karma

If an IP has at least 5 connections and all are good or bad, and all_good or all_bad result will be added to the pass or fail test list. This are very good indicators of future connection quality and are scored in karma.ini.

ASN / Network Neighborhood Reputation

[asn]
enable=true    (default: true)
report_as

When [asn]enable is true, karma records the number of good and bad connections from each ASN.

ASNs with less than 5 karma points in either direction are ignored.

report_as

Store the ASN results as another plugin. Example: I set report_as=asn, so that karma history for an ASN is reported with the ASN plugin data. A practical consequence of changing report_as is that the award location in karma.ini would need to change from:

NNN karma | pass | equals | asn_all_good |  2
NNN karma | fail | equals | asn_all_bad  | -3

to:

NNN asn | pass | equals | asn_all_good |  2
NNN asn | fail | equals | asn_all_bad  | -3

Malformed Envelope Addresses

Very old versions of Outlook Express and some malware senders don't bother complying with the RFC (5321, 2821, 821) address format. Karma checks the envelope from and to addresses for a common RFC ignorant pattern that is highly correlated with malware.

Unrecognized SMTP verbs/commands

Certain bruteforce password hacking tools have a pre-programmed SMTP path that ignores SMTP responses. After EHLO, they attempt AUTH,LOGIN with a valid username. To bruteforce a password often requires millions of attempts so each bot sprays a couple dozen connections at the target server. Better quality MTAs like Haraka have built-in auth protection that inserts timeouts between successive auth attempts. The bots work around that by dropping the connection after each failure and reconnecting. The attempts are distributed so IP blocking is of limited effectiveness.

To combat these bruteforce attacks several strategies are called for:

1. Impose [connection delays](#delay)
2. Disable SMTP-AUTH when not encrypted. The bots rarely use STARTTLS.
   Besides preventing user passwords from transiting the internet in clear
   text, requiring TLS encryption also means AUTH is not available to
   poorly written bots.
3. Having done #2, bot AUTH attempts show up as unrecognized commands.
   Penalizing these with tarpitting and rate limiting will almost never
   harm a legit sender but it will make it take much much longer for
   attackers to bruteforce passwords.

Disabling karma for certain connections

Karma plugin is disabled for all private IP addresses according to RFC1918. Optionally you can prevent any plugin operations by setting connection.notes.disable_karma = true at connection_init hook.

LIMITATIONS

Karma is most effective at filtering mail delivered by bots and rogue servers. Spam delivered by servers with good reputations normally pass karma's checks. Expect to use karma with content filters.