Contextualize Alerts

Introduction

CrowdSec doesn't store raw log information once it has finished processing them. However, you can send additional data within an alert with a feature we call alert context:

Whilst some collections already include a context configuration, you must enable the context option within your Security Engine console configuration.

This can be achieved by running:

sudo cscli console enable context

Check current config

info

These commands have to be run on the CrowdSec Local API

You can view the current status of your console options with:

$ sudo cscli console status
╭────────────────────┬───────────┬───────────────────────────────────────────────────╮
│ Option Name        │ Activated │ Description                                       │
├────────────────────┼───────────┼───────────────────────────────────────────────────┤
│ custom             │ ✅        │ Send alerts from custom scenarios to the console  │
│ manual             │ ❌        │ Send manual decisions to the console              │
│ tainted            │ ✅        │ Send alerts from tainted scenarios to the console │
│ context            │ ✅        │ Send context with alerts to the console           │
│ console_management │ ❌        │ Receive decisions from console                    │
╰────────────────────┴───────────┴───────────────────────────────────────────────────╯

You can enable alert context with:

sudo cscli console enable context

You can as well inspect and control the enabled alert context configuration:

$ sudo cscli lapi context status
method:
- evt.Meta.http_verb
status:
- evt.Meta.http_status
target_uri:
- evt.Meta.http_path
target_user:
- evt.Meta.target_user
user_agent:
- evt.Meta.http_user_agent

note

The context configuration can as well be found in /etc/crowdsec/contexts/ as yaml files.

Vizualise alert context

The alert context will be display when inspecting an alert:

$ sudo cscli alerts inspect 7
################################################################################################
 - ID         : 7
 - Date       : 2022-12-12T18:46:44Z
 - Machine    : c620f52cedf9432e969f26afee12d651
 - Simulation : false
 - Reason     : crowdsecurity/http-bad-user-agent
 - Events Count : 2
 - Scope:Value: Ip:127.0.0.1
 - Country    :
 - AS         :
 - Begin      : 2022-12-12 18:46:43.339960525 +0000 UTC
 - End        : 2022-12-12 18:46:43.341210351 +0000 UTC
 - Active Decisions  :
╭───────┬──────────────┬────────┬────────────────────┬──────────────────────╮
│  ID   │ scope:value  │ action │     expiration     │      created_at      │
├───────┼──────────────┼────────┼────────────────────┼──────────────────────┤
│ 15006 │ Ip:127.0.0.1 │ ban    │ 3h59m50.949157024s │ 2022-12-12T18:46:44Z │
╰───────┴──────────────┴────────┴────────────────────┴──────────────────────╯
 - Context  :
╭─────────────┬──────────────────────────────────────────────────────────────╮
│     Key     │                            Value                             │
├─────────────┼──────────────────────────────────────────────────────────────┤
│ target_fqdn │ www.crowdsec-test.net                                        │
│ target_fqdn │ www.crowdsec-test-1.net                                      │
│ user_agent  │ Mozilla/5.00 (Nikto/2.1.5) (Evasions:None) (Test:Port Check) │
│ user_agent  │ Mozilla/5.00 (Nikto/2.1.5) (Evasions:None) (Test:getinfo)    │
╰─────────────┴──────────────────────────────────────────────────────────────╯

And we can see that the target_fqdn and the user_agent are now displayed in the context of the alert.

Adding custom alert context

You can get context values from:

• Part of a parsed log line (with evt.Parsed)
• Meta values set by parsers (with evt.Meta)
• Hardcoded strings in the context configuration (with "my_value")
• More generally, anything available in evt (eg, evt.Unmarshaled with some parsers)
• From expr helpers (all expr helpers are available in the context, allowing for example CrowdsecCTI(evt.Meta.source_ip).GetMaliciousnessScore())

You can add your custom alert context by adding a yaml file in /etc/crowdsec/contexts/ :

cat > /etc/crowdsec/contexts/example.yaml << EOF
context:
  example_value:
  - '"something"'
  http_extra_status:
  - evt.Meta.http_status
EOF

Let's now trigger a http scenario by trying to exploit a well known vulnerability:

$ curl 'target.com/%2E%2E/%2E%2E'

In crowdsec.log, we're seeing our alert:

time="2024-01-24T18:08:34+01:00" level=info msg="Ip x.x.x.x performed 'crowdsecurity/http-cve-2021-41773' (1 events over 336ns) at 2024-01-24 17:08:34.026228434 +0000 UTC"

It as well appears when we're inspecting our alert:

$ cscli  alerts list
╭──────┬────────────────────┬────────────────────────────────────────────┬─────────┬─────────────────────────────────────────────────────────┬───────────┬─────────────────────────────────────────╮
│  ID  │       value        │                   reason                   │ country │                           as                            │ decisions │               created_at                │
├──────┼────────────────────┼────────────────────────────────────────────┼─────────┼─────────────────────────────────────────────────────────┼───────────┼─────────────────────────────────────────┤
│ 6545 │ Ip:xxx.xx.xx.xx    │ crowdsecurity/http-cve-2021-41773          │ FR      │ 5410 Bouygues Telecom SA                                │ ban:1     │ 2024-01-24 17:08:34.026228866 +0000 UTC │
...
$  cscli  alerts inspect -d 6545
...
 - Context  :
╭───────────────────┬────────────────╮
│        Key        │     Value      │
├───────────────────┼────────────────┤
│ example_value     │ something      │
│ http_extra_status │ 400            │
│ method            │ HEAD           │
│ status            │ 400            │
│ target_uri        │ /%2E%2E/%2E%2E │
│ user_agent        │ -              │
╰───────────────────┴────────────────╯
...

And in the console:

More information here for managing the context from the Hub with cscli.

Automagically detect possible alert context

It is possible to detect all the possible fields that a given parser can output (or all the installed parsers with --all flag):

$ sudo cscli lapi context detect crowdsecurity/nginx-logs
Acquisition :

  - evt.Line.Module
  - evt.Line.Raw
  - evt.Line.Src

crowdsecurity/nginx-logs :

  - evt.Meta.http_path
  - evt.Meta.http_status
  - evt.Meta.http_user_agent
  - evt.Meta.http_verb
  - evt.Meta.log_type
  - evt.Meta.service
  - evt.Meta.source_ip
  - evt.Meta.target_fqdn
  - evt.Parsed.body_bytes_sent
  - evt.Parsed.cid
  - evt.Parsed.http_referer
  - evt.Parsed.http_user_agent
  - evt.Parsed.http_version
  - evt.Parsed.loglevel
  - evt.Parsed.message
  - evt.Parsed.pid
  - evt.Parsed.proxy_alternative_upstream_name
  - evt.Parsed.proxy_upstream_name
  - evt.Parsed.remote_addr
  - evt.Parsed.remote_user
  - evt.Parsed.request
  - evt.Parsed.request_length
  - evt.Parsed.request_time
  - evt.Parsed.status
  - evt.Parsed.target_fqdn
  - evt.Parsed.tid
  - evt.Parsed.time
  - evt.Parsed.time_local
  - evt.Parsed.verb
  - evt.StrTime

Delete a context

Delete the yaml file containing your custom alert context, and reload crowdsec.

rm /etc/crowdsec/contexts/example.yaml
systemctl restart crowdsec