Skip to main content
Version: Next

pfSense

We have created a pfSense package with a simple UI to configure the Security Engine and the Firewall Remediation Component (bouncer).

Three types of setup are supported:

Small (remediation only) - the pfSense machine receives blocklists from a CrowdSec Security Engine that you are running on a different machine. Incoming connections are blocked at the firewall by (configurable) pfSense rules.

Medium (small+log processor) - in addition to enforcing blocklists, the pfSense machine can detect attacks directed at the firewall itself, for example port scans. The data about the attacks is sent (for analysis and possibly sharing) to a Security Engine that you are running on a different machine.

Large (medium+LAPI) - deploy a fully autonomous Security Engine on the firewall system and allow other Log Processors to connect to it. Requires a persistent /var directory (no RAM disk) and a slightly larger pfSense machine, depending on the amount of data to be processed.

If you are not already a CrowdSec user, the Large setup is the easiest: just leave the default values to enable remediation, log processor and Local API.

info

The CrowdSec configuration is not transferred when you restore a pfSense backup, and you'll need to reconfigure it or backup separately.

Installing the package

We have submitted the package to the pfSense developers for review and inclusion in the official repositories.

If you find crowdsec under System/Package Manager, we recommend you to install it from there.

If the package is not available yet, you'll have to install it manually.

  • Choose the release you want to install, click Assets for the list of packages to install.

  • Open an ssh connection to your pfSense box and run the following commands in the right order. Do not activate or run the services, because pfSense will take care of it.

# setenv IGNORE_OSVERSION yes
# pkg add -f <link to abseil>
# pkg add -f <link to re2>
# pkg add -f <link to crowdsec-firewall-bouncer>
# pkg add -f <link to crowdsec>
# pkg add -f <link to pfSense-pkg-crowdsec>

Configuration

Once the package and its dependencies are installed, go to Service/CrowdSec. The options Remediation Component, Log Processor and Local API should be enabled. Click Save.

With the size analogy, the default is a "Large", autonomous installation. For a "Medium", disable Local API and fill the fields in the Remote LAPI section. For a "Small", disable Log Processor too.

CrowdSec on pfSense is fully functional from the command line but the web interface is read-only, with the exception of decision revocation (unban). Most actions require the shell or the CrowdSec Console. For simple things, Diagnostics/Command Prompt works as well as ssh. You are free to edit the files in /usr/local/etc/crowdsec, although some setting may be overwritten by the pfSense package if they are mandatory.

caution

Ram Disk: unless you disable Local API, ensure that you are not using a RAM disk for the /var directory. The persistent CrowdSec database and GeoIP tables are in /var/db. If you really need a RAM disk, you can still use the log processor and remediation but you will need to connect them to a remote CrowdSec instance.

Service Status

In the page Status/CrowdSec you can see

  • registered log processors and bouncers
  • installed hub items (collections, scenarios, parsers, postoverflows)
  • alerts and local decisions

All tables are read-only with an exception: you can delete single decisions, to unban an IP for example.

All hub objects are periodically upgraded with a cron job.

Detecting attacks

If a Log Processor is running, the following scenarios are enabled by default:

  • portscan
  • ssh brute-force
  • pfSense admin UI brute-force
  • HTTP vulnerability probing

These will trigger a ban on the attacking IP (4 hours by default) and report it to the CrowdSec Central API (meaning timestamp, scenario, attacking IP, for inclusion in the Community Blocklist.

You can add scenarios to detect other types of attack on the pfSense server, or connect several log processors to the same LAPI node. Other types of remediation are possible (ex. captcha test for scraping attempts).

If disk space is not an issue, you can increase the maximum size of log files before they are compressed and rotated. This will help us in case you report acquisition issues and we need to match the application behavior with the content of the acquired logs.

We recommend you to register to the Console, especially if you protect several machines.

Processing logs

If a collection, parser or scenario can be applied to a software that you are running on pfSense, you add it with cscli collections install ..., then you need to configure where CrowdSec will find the logs.

New acquisition files should go in /usr/local/etc/crowdsec/acquis.d/. See pfsense.yaml for an example. The option poll_without_inotify: true is required if the log sources are symlinks. Make sure to reload or restart CrowdSec when you add new data sources.

Diagnostics

Under Diagnostics/CrowdSec you can check if the logs are acquired and the events are triggered correctly. For real monitoring, you can fetch the same metrics with Prometheus (Grafana dashboard included) Telegraf or your favorite solution.

If you are not running a LAPI or a Log Processor, some metrics are always empty.

Logs

You can see the Security Engine logs in Status/System Logs/Packages/crowdsec.

Other logs not shown in the UI are in /var/log/crowdsec/crowdsec_api.log and crowdsec-firewall-bouncer.log.

Service Management

Both services, Security Engine (crowdsec) and Remediation (crowdsec-firewall-bouncer) can be controlled from Status/Services. The equivalent shell commands are service crowdsec.sh start/stop/restart and service crowdsec_firewall.sh start/stop/restart. Note the ending .sh!

Viewing blocked IPs

You can see the tables of the blocked IPs in Diagnostics/Tables or from the shell, with the commands pfctl -T show -t crowdsec_blacklists (IPv4) and pfctl -T show -t crowdsec6_blacklists (IPv6).

To show the same data with more context, use cscli decisions list -a.

Testing

A quick way to test that everything is working correctly end-to-end is to execute the following command.

Your ssh session should freeze and you should be kicked out from the firewall. You will not be able to connect to it (from the same IP address) for two minutes.

It might be a good idea to have a secondary IP from which you can connect, should anything go wrong.

# cscli decisions add -t ban -d 2m -i <your_ip_address>

You may have to disable the Anti-lockout rule in System/Advanced/Admin Access for the time of the test.

This is a more secure way to test than attempting to brute-force yourself: the default ban period is 4 hours, and CrowdSec reads the logs from the beginning, so it could ban you even if you failed ssh login 10 times in 30 seconds two hours before installing it.

It must be noted that the Login Protection service which is enabled by default on pfSense can be triggered - and block a brute force attacker - before CrowdSec does, because it's more sensitive. Still, some attacks that are not detected by Login Protection will be detected by CrowdSec and shared. If you need more CrowdSec tests you may want to temporarily disable Login Protection, depending on the scenario.

LAN / private networks whitelist

By default the FreeBSD version of CrowdSec does not install any whitelist. If you trust your 10.0.0.0/8, 192.168.0.0/16 and 172.16.0.0/12 networks, you can use cscli parsers install crowdsecurity/whitelists to whitelist them.

Uninstalling

In most cases, just remove the crowdsec package from System/Package Manager/Installed Packages. This won't remove the database or configuration files, just in case you want to reinstall CrowdSec later.

If you need to make sure you removed all traces of CrowdSec, you can run the following commands:

# pkg remove pfSense-pkg-crowdsec crowdsec crowdsec-firewall-bouncer
# rm -rf /usr/local/etc/crowdsec /usr/local/etc/rc.conf.d/crowdsec*
# rm -rf /var/db/crowdsec /var/log/crowdsec* /var/run/crowdsec*

For testing purposes, you may want to remove the <crowdsec> section from /conf/config.xml as well.

The blocklist mirror

Before releasing the official package, one way to integrate pfSense and CrowdSec was to install a blocklist mirror and connect it to pfBlockerNG. While this is still a viable solution, it has slower performance than the method described above, especially in terms of latency when receiving decision updates. It also required pfBlockerNG in addition to CrowdSec.

Enrolling your instance

The next step is to enroll your instance with the CrowdSec Console.

For the benefits, please visit the Console section.