CookieInterceptor in Apache Struts 2.x before 2.3.20, when a wildcard cookiesName value is used, does not properly restrict access to the getClass method, which allows remote attackers to "manipulate" the ClassLoader and modify session state via a crafted request. NOTE: this vulnerability exists because of an incomplete fix for CVE-2014-0113.
It was found that the Struts 2 ParametersInterceptor allows access to the 'class' parameter, which is directly mapped to the getClass() method. A remote attacker could use this flaw to manipulate the ClassLoader used by the application server running Struts 2. This could lead to arbitrary remote code execution under certain conditions.
This flaw is reported to affect Struts 2.0.0 through to Struts 2.3.16. It is corrected in 2.3.16.1.
External References:
https://cwiki.apache.org/confluence/display/WW/S2-020
Apache Struts 2.0.0 through 2.3.4 allows remote attackers to cause a denial of service (CPU consumption) via a long parameter name, which is processed as an OGNL expression.
The token check mechanism in Apache Struts 2.0.0 through 2.3.4 does not properly validate the token name configuration parameter, which allows remote attackers to perform cross-site request forgery (CSRF) attacks by setting the token name configuration parameter to a session attribute.
Multiple cross-site scripting (XSS) vulnerabilities in component handlers in the javatemplates (aka Java Templates) plugin in Apache Struts 2.x before 2.2.3 allow remote attackers to inject arbitrary web script or HTML via an arbitrary parameter value to a .action URI, related to improper handling of value attributes in (1) FileHandler.java, (2) HiddenHandler.java, (3) PasswordHandler.java, (4) RadioHandler.java, (5) ResetHandler.java, (6) SelectHandler.java, (7) SubmitHandler.java, and (8) TextFieldHandler.java.
ParametersInterceptor in OpenSymphony XWork 2.0.x before 2.0.6 and 2.1.x before 2.1.2, as used in Apache Struts and other products, does not properly restrict # (pound sign) references to context objects, which allows remote attackers to execute Object-Graph Navigation Language (OGNL) statements and modify server-side context objects, as demonstrated by use of a \u0023 representation for the # character.