Decisions, profiles, allowlists and AppSec

1. Decisions #

View, add, delete.

Day to day #

sudo cscli decisions list
sudo cscli decisions list --origin cscli
sudo cscli decisions add --ip 192.0.2.5 --duration 1h --reason "manual test"
sudo cscli decisions add --range 192.0.2.128/25 --duration 1h
sudo cscli decisions add --ip 192.0.2.6 --duration 1h --type captcha
sudo cscli decisions delete --ip 192.0.2.5

sudo cscli alerts list
sudo cscli alerts inspect 3

2. Profiles #

From alert to decision.

How it works #

The /etc/crowdsec/profiles.yaml file holds a series of profiles, separated by ---. The first one whose filter matches applies, and on_success: break stops reading. By default: a 4 h ban for an IP or range.

Banning repeat offenders longer #

name: progressive_ban
filters:
  - Alert.Remediation == true && Alert.GetScope() == "Ip"
decisions:
  - type: ban
    duration: 4h
# 4 h the first time, then 8 h, 12 h… for each repeat offense
duration_expr: Sprintf('%dh', (GetDecisionsCount(Alert.GetValue()) + 1) * 4)
on_success: break

Order matters. Place your custom profiles before the default ones, otherwise they will never be reached.

A captcha for the web #

name: captcha_http
filters:
  - Alert.Remediation == true && Alert.GetScope() == "Ip" && Alert.GetScenario() startsWith "crowdsecurity/http-"
decisions:
  - type: captcha
    duration: 1h
on_success: break

Check that your bouncer can enforce a captcha: a firewall bouncer cannot.

3. False positives #

Protecting trusted addresses.

Allowlists #

An allowlist declares addresses that must never be blocked. It is the preferred method and acts immediately: when added, existing decisions on these addresses are deleted.

sudo cscli allowlists create bureau -d "Trusted IPs"
sudo cscli allowlists add bureau 203.0.113.60 198.51.100.0/24 -d "VPN"
sudo cscli allowlists inspect bureau
sudo cscli allowlists check 198.51.100.7
# 198.51.100.7 is allowlisted by item 198.51.100.0/24 from bureau (VPN)
sudo cscli allowlists remove bureau 203.0.113.60

A manual decision on an allowed address is refused, unless using --bypass-allowlist.

Tutorials to correct. Some articles suggest cscli decisions add --type whitelist. This is not the right mechanism: a decision type describes the remediation to apply. Use an allowlist.

Other levers #

• Simulation, to observe a scenario without blocking.
• Parser whitelists, to ignore events based on their content, such as a monitoring tool.
• The firewall bouncer's scenarios_not_containing filter, to exclude some scenarios.

4. AppSec: the WAF #

Blocking a request before it arrives.

The principle #

Scenarios analyze logs: they react after requests. AppSec, on the other hand, receives each request from the bouncer before it reaches the application and answers allow or ban.

sudo cscli collections install crowdsecurity/appsec-virtual-patching crowdsecurity/appsec-generic-rules

Create /etc/crowdsec/acquis.d/appsec.yaml:

source: appsec
listen_addr: 127.0.0.1:7422
path: /
appsec_configs:
  - crowdsecurity/appsec-default
labels:
  type: appsec

What we observed #

Make no mistake. Virtual patching targets known flaws and specific cases. In our test, a generic SQL injection passed with these two collections. AppSec complements your application security, it does not replace it.

sudo cscli metrics show appsec

Sources #

• CrowdSec : Profiles
• CrowdSec : AppSec

Commands and results verified on CrowdSec 1.8.2.