Try in Splunk Security Cloud

Description

CVE-2021-44228 Log4Shell payloads can be injected via various methods, but on of the most common vectors injection is via Web calls. Many of the vulnerable java web applications that are using log4j have a web component to them are specially targets of this injection, specifically projects like Apache Struts, Flink, Druid, and Solr. The exploit is triggered by a LDAP lookup function in the log4j package, its invocation is similar to ${jndi:ldap://PAYLOAD_INJECTED}, when executed against vulnerable web applications the invocation can be seen in various part of web logs. Specifically it has been successfully exploited via headers like X-Forwarded-For, User-Agent, Referer, and X-Api-Version. In this detection we match the invocation function with a network connection to a malicious ip address.

  • Type: Anomaly
  • Product: Splunk Enterprise, Splunk Enterprise Security, Splunk Cloud
  • Datamodel: Network_Traffic, Web
  • Last Updated: 2021-12-13
  • Author: Jose Hernandez
  • ID: 69afee44-5c91-11ec-bf1f-497c9a704a72

Annotations

ATT&CK
ID Technique Tactic
T1190 Exploit Public-Facing Application Initial Access
Kill Chain Phase
  • Exploitation
NIST
  • DE.CM
CIS20
  • CIS 3
  • CIS 5
  • CIS 16
CVE
ID Summary CVSS
CVE-2021-44228 Apache Log4j2 2.0-beta9 through 2.15.0 (excluding security releases 2.12.2, 2.12.3, and 2.3.1) JNDI features used in configuration, log messages, and parameters do not protect against attacker controlled LDAP and other JNDI related endpoints. An attacker who can control log messages or log message parameters can execute arbitrary code loaded from LDAP servers when message lookup substitution is enabled. From log4j 2.15.0, this behavior has been disabled by default. From version 2.16.0 (along with 2.12.2, 2.12.3, and 2.3.1), this functionality has been completely removed. Note that this vulnerability is specific to log4j-core and does not affect log4net, log4cxx, or other Apache Logging Services projects. 9.3
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
| from datamodel Web.Web 
| rex field=_raw max_match=0 "[jJnNdDiI]{4}(\:
|\%3A
|\/
|\%2F)(?<proto>\w+)(\:\/\/
|\%3A\%2F\%2F)(\$\{.*?\}(\.)?)?(?<affected_host>[a-zA-Z0-9\.\-\_\$]+)" 
| join affected_host type=inner [
| tstats `security_content_summariesonly` count min(_time) as firstTime max(_time) as lastTime from datamodel=Network_Traffic.All_Traffic by All_Traffic.dest 
| `drop_dm_object_name(All_Traffic)` 
| `security_content_ctime(firstTime)` 
| `security_content_ctime(lastTime)` 
| rename dest AS affected_host] 
| fillnull 
| stats count by action, category, dest, dest_port, http_content_type, http_method, http_referrer, http_user_agent, site, src, url, url_domain, user 
| `log4shell_jndi_payload_injection_with_outbound_connection_filter`

Macros

The SPL above uses the following Macros:

:information_source: log4shell_jndi_payload_injection_with_outbound_connection_filter is a empty macro by default. It allows the user to filter out any results (false positives) without editing the SPL.

Required field

  • action
  • category
  • dest
  • dest_port
  • http_content_type
  • http_method
  • http_referrer
  • http_user_agent
  • site
  • src
  • url
  • url_domain
  • user

How To Implement

This detection requires the Web datamodel to be populated from a supported Technology Add-On like Splunk for Apache or Splunk for Nginx.

Known False Positives

If there is a vulnerablility scannner looking for log4shells this will trigger, otherwise likely to have low false positives.

Associated Analytic story

RBA

Risk Score Impact Confidence Message
15.0 50 30 CVE-2021-44228 Log4Shell triggered for host $dest$

:information_source: The Risk Score is calculated by the following formula: Risk Score = (Impact * Confidence/100). Initial Confidence and Impact is set by the analytic author.

Reference

Test Dataset

Replay any dataset to Splunk Enterprise by using our replay.py tool or the UI. Alternatively you can replay a dataset into a Splunk Attack Range

source | version: 1