Examining Apache Struts remote code execution vulnerabilities

On Sept. 7 Equifax announced a major data breach affecting 143 million customers. The breach was the result of a Remote Code Execution (RCE) attack that exploited a vulnerability in Apache Struts software—an open source framework for creating web applications. On the same day as the breach announcement, Apache Struts published details of CVE- 2017-12611–the fourth in a series of critical RCE vulnerabilities exposed in the Struts platform since March this year alone.

According to an Equifax statement, the data breach was discovered on July 29, but admitted that sensitive data was accessible to the attacker from mid-May. This means that Equifax systems had been compromised for at least 40 days before discovery, raising serious concerns about Equifax’s security practices.

Apache Struts responded quickly in a statement to clarify their position on the breach. Describing their policy of due diligence on securing their software quickly against reported vulnerabilities, Apache’s statement firmly rebuts reports that vulnerabilities published in either July or September could have been targeted unless the attacker had a zero-day exploit. In other words, the attacker would have needed knowledge of the vulnerabilities before Apache released their advisories and fixes.

The timeline of related events makes it clear that fixed versions of Struts were available at or before the security advisories were published, and that known exploits were not available in the wild beforehand. The timeline also bears witness to Apache’s assertions of consistent good practice and tells us that the attack was likely to be a product of poor security practices on the part of Equifax.

CVE-2017-5638 CVSS 2 Base Score 8.7

CVE-2017-5636 was initially described in the Apache advisory S2-045. Originally, the HTTP header ‘Content-Type’ was identified as the attack vector. However, a subsequent Apache advisory S2-046 provided additional vectors associated with the same vulnerability HTTP Headers ‘Content-Disposition’ & ‘Content-Length’.

Due to the additional technical details being released, it is advised to verify that you have received updated rule sets in order to validate existing mitigation strategies.  

The attacker can leverage these conditions to execute OGNL commands. OGNL is an expressive interpreted language that enables full command execution under the web user system privileges. Likewise, Content-Disposition and Content-Length HTTP headers are viable attack vectors.

Workaround: For users of affected versions, several immediate mitigation measures were available including the option to switch to a different implementation of the parser.

CVE-2017-9791 CVSS 2 Base Score 9.0

An attacker can achieve remote code execution in the Struts Showcase app in the Struts 1 plugin example in Struts 2.3.x. A malicious `OGNL` expression that executes arbitrary system commands can be supplied as a field value when using the Struts 2 Struts 1 plugin.

When the relevant field is associated with a Struts 1 action and the malicious input is part of the error message used by the ActionMessage class, the ‘OGNL’ commands will get executed as the web server user.

Workaround: Always use resource keys instead of passing a raw message to the ActionMessage and never pass a raw value directly.

CVE-2017-9805 CVSS 2 Base Score 8.3

Apache Struts Allows Remote Code Execution when using the Struts REST plugin with `XStream` handler to deserialise XML requests. This vulnerability could manifest itself by providing crafted user-controlled data as an XML payload to the web application. Successful exploitation of the flaw could enable a hacker to gain full control of the affected server, finally letting the attacker infiltrate into other systems on the same network.

Workaround: Various workaround options are available with the best being to remove the Struts REST plugin if or when it’s not being used.