Our website uses cookies to enhance your browsing experience.
Your message has been sent. We will email you at
If you haven't received our response, please do the following:
check your Spam/Junk folder and click the
"Not Spam" button for our message.
This way, you won't miss messages from our team in the future.
How can simple XML files processing turn into a security weakness? How can a blog deployed on your machine cause a data leak? Today we'll find answers to these questions, learn what XXE is and how it looks like.
Before we begin, note that there are several types of vulnerabilities related to XML processing. The most popular vulnerabilities are XXE, XEE, and XPath injection. In this article we inspect XXE. If you're interested in the essence of an XEE attack, you can read this article: "How Visual Studio 2022 ate up 100 GB of memory and what XML bombs had to do with it". We'll get to XPath injection some time later. :)
XXE (XML eXternal Entities) is an application security weakness. The possible source of this attack — compromised data processed by an insecurely configured XML parser. This attack can result in disclosure of data from the target machine or server-side request forgery (SSRF).
XML files may contain the document type definition (DTD), which describes the structure of an XML file. DTD allows us to define and use XML entities.
It can look like this:
<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE order [
<!ENTITY myEntity "lol">
]>
<order>&myEntity;</order>In this XML, we declare myEntity and use it further — &myEntity;. In this case, the entity is internal and is defined as literal. If an XML parser expands this entity, it substitutes &myEntity; with the actual value — lol. Besides, some internal entities can expand through others. XML bombs can be created this way and perform XEE attacks.
However, entities can be external. They can refer to some local files or access external resources:
<!ENTITY myExternalEntity SYSTEM "https://test.com/target.txt">Here's an example of an XML file where an external entity refers to a local file:
<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE order [
<!ENTITY myExternalEntity SYSTEM "file:///D:/HelloWorld.cs">
]>
<order>&myExternalEntity;</order>In this case, an XML parser substitutes myExternalEntity with the contents of the file along path D:/HelloWorld.cs. If it's properly configured, of course.
XXE attack exploits the feature above.
Here's an example. Let's assume that there's an application that accepts queries as XML files and processes items with the corresponding ID.
The application works with the following XML file format:
<?xml version="1.0" encoding="utf-8" ?>
<order>
<itemID>62</itemID>
</order>Simplified C# code:
static void ProcessItemWithID(XmlReader reader, String pathToXmlFile)
{
....
while (reader.Read())
{
if (reader.Name == "itemID")
{
var itemIdStr = reader.ReadElementContentAsString();
if (long.TryParse(itemIdStr, out var itemIdValue))
{
// Process item with the 'itemIdValue' value
Console.WriteLine(
$"An item with the '{itemIdValue}' ID was processed.");
}
else
{
Console.WriteLine($"{itemIdStr} is not valid 'itemID' value.");
}
}
}
}The logic is simple:
Thus, for the XML file above, the application will display the following line:
An item with the '62' ID was processed.If we insert something else in the ID instead of the number ("Hello world", for example), the application reports an error:
"Hello world" is not valid 'itemID' value.If an XML parser (reader) processes external entities, this is a security flaw. Below is an XML file that can be used to compromise the application:
<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE order [
<!ENTITY xxe SYSTEM "file:///D:/MySecrets.txt">
]>
<order>
<itemID>&xxe;</itemID>
</order>The xxe external entity is declared in this file. When an XML parser process this file, it substitutes &xxe; with the contents of the file along path D:/MySecrets.txt. For example, "This is an XXE attack target.". As a result, the application will display the following:
"This is an XXE attack target." is not valid 'itemID' value.Thus, an application will be vulnerable to XXE attacks, if:
If an attacker can obtain the value of the entity, they can get the file contents from the compromised device. This is already dangerous. Besides, an attacker can get more data about the system as a whole and find other security weaknesses.
XXE can also lead to an SSRF attack. The hacker may not have access to some resources (access restricted for external users), but the exploited application may have it. Since XXE allows to make requests over the network, a compromised application is a breach in the resource protection.
Speaking about the importance and danger of XXE — this security weakness is often mentioned in various standards, tops, and enumerations.
CWE
The Common Weakness Enumeration has a separate entry for XXE: CWE-611: Improper Restriction of XML External Entity Reference.
CWE Top 25
Every year 25 most common and dangerous weaknesses are selected from the CWE list to compile the CWE Top 25.
In 2021, XXE lost 4 positions compared to 2020, but remained in the top on the 23d place.
OWASP ASVS
OWASP ASVS (Application Security Verification Standard) contains requirements for secure development. It also has an entry about XXE: OWASP ASVS 4.0.3 (ID 5.5.2): Verify that the application correctly restricts XML parsers to only use the most restrictive configuration possible and to ensure that unsafe features such as resolving external entities are disabled to prevent XML eXternal Entity (XXE) attacks.
OWASP Top 10
The OWASP Top 10 2017 had a separate category for XXE: A4:2017-XML External Entities (XXE). In the OWASP Top 10 2021 a separate category for XXE was eliminated. XXE now belongs to A05:2021-Security Misconfiguration.
As I mentioned above, XXE needs at least two components: an insecurely configured parser and data from the attacker that this parser processes.
Everything is quite simple here. The application has several places where it accepts external data. It has to be processed carefully — not all people use an application for its intended purpose.
Such application places are console application arguments, various form fields, query data, etc. The first thing that comes to mind is console input.
var taintedVar = Console.ReadLine();We don't know what's inside taintedVar. This variable can contain data in the expected format or a string to compromise the system. We can't trust it.
You can read more about it in the "Taint sources" section of "OWASP, vulnerabilities, and taint analysis in PVS-Studio for C#. Stir, but don't shake. You should also be suspicious of public access parameters. The data in those methods may be safe or not. You can read about it here.
An XML parser is vulnerable to XXE, if:
If an XML parser does not set a limit on the entities' maximum size (or the size is large), this may worsen the attack, since the attacker will be able to extract larger amounts of data.
The desired behavior is set with the following properties:
Some XML parsers have all these options, others — don't. Their semantic meaning does not change from type to type.
ProhibitDtd
The ProhibitDtd property has the Obsolete attribute. Now the DtdProcessing property is used instead of ProhibitDtd. Still, it can be used in the old code. The true value prohibits DTD processing, false — allows it.
DtdProcessing
The DtdProcessing property has the System.Xml.DtdProcessing type and can take the Prohibit, Ignore and Parse values:
You probably have a question now, and I'll answer it. If the ProhibitDtd and DtdProcessing properties occur together in code (for example, in XmlReaderSettings), they are related to each other. So, if you prohibit DTD in one property and allow in another, only the last option set would be applied. :)
XmlResolver
The XmlResolver property is responsible for the object used to process external entities. The safest option — absence of resolver at all (null value). In this case, even if DTD processing is enabled, external entities won't expand.
MaxCharactersFromEntities
Another option of interest for us. MaxCharactersFromEntities is responsible for the maximum allowable size of entities. The bigger the value, the potentially more information will be extracted during an XXE attack.
The most common standard types to work with XML are XmlReader, XmlTextReader, XmlDocument. Note that the list is not limited to them.
Once again, the configuration of a parser is dangerous, if:
XmlReader
The XmlReaderSettings object, created explicitly or implicitly, configures behavior of the XmlReader. The XmlReaderSettings type has all the settings listed earlier.
A parser with a dangerous configuration may look like this:
var settings = new XmlReaderSettings()
{
DtdProcessing = DtdProcessing.Parse,
XmlResolver = new XmlUrlResolver(),
MaxCharactersFromEntities = 0
};
using (var xmlReader = XmlReader.Create(xmlFileStringReader, settings))
....Here the developer explicitly allowed DTD processing, set a resolver for external entities, and removed the limitations on their size.
XmlTextReader
In this case, we are dealing with the same properties: ProhibitDtd, DtdProcessing, XmlResolver.
An example of a dangerously configurated parser:
using (var xmlTextReader = new XmlTextReader(xmlFileStringReader))
{
xmlTextReader.XmlResolver = new XmlUrlResolver();
xmlTextReader.DtdProcessing = DtdProcessing.Parse;
....
}XmlDocument
In the XmlDocument type, we are interested in the XmlResolver property. In this case, a dangerously configurated parser can look like this:
XmlDocument xmlDoc = new XmlDocument();
xmlDoc.XmlResolver = new XmlUrlResolver();xmlDoc in this configuration expands external entities and can be considered dangerous.
Above we looked at examples where XML parsers were configured explicitly. However, all the listed types have some default settings, and there's a couple of interesting things about them.
Firstly, these settings are different for different .NET versions.
Secondly, the settings vary from type to type. For example, the DTD processing can be enabled or disabled by default.
In some cases, an XML parser can have a dangerous configuration by default, even if dangerous settings were not set explicitly.
As a result, we have to remember different types of parsers, different default settings in different types and .NET versions. It's a good amount of information that can be difficult to keep in mind (especially at first).
So, sometimes we can't say if an XML parser is XXE-resistant by only looking at code. For example, here:
XmlDocument doc = new XmlDocument();
doc.Load(xmlReader);It's unclear whether doc can process external entities or not — we need to know the framework version first.
The values of the 'dangerous' settings changed between .NET Framework 4.5.1 and .NET Framework 4.5.2. Below is the table that shows in which .NET versions parsers with default settings are XXE-resistant by default, and in which they're not.
|
Instances of types |
.NET Framework 4.5.1 and lower |
.NET Framework 4.5.2 and higher (including .NET Core and .NET) |
|---|---|---|
|
XmlReader (XmlReaderSettings) |
Safe |
Safe |
|
XmlTextReader |
Vulnerable |
Safe |
|
XmlDocument |
Vulnerable |
Safe |
Yes, XmlReader (created via XmlReaderSettings) is safe in .NET Framework 4.5.1 and lower because DTD processing is disabled in it.
Even though in the new framework versions parsers are configured securely by default, the best option is to explicitly configure the necessary settings. Yes, there'll be a lot more code. At the same time, it'll be more obvious and stable when you port it between different .NET Framework versions.
Done with the theory. Next let's look at the real vulnerability. Make yourself a cup of coffee and let's go!
Above, we analyzed the theoretical component of XXE, talked a little more specifically about these security weaknesses in .NET, looked at what the insecure components of the vulnerability look like from the point of view of the code. Now it's time for practice. BlogEngine.NET is here to help.
Description from the project's website: BlogEngine is an open source blogging platform since 2007. Easily customizable. Many free built-in Themes, Widgets, and Plugins.
The project's source code is available on GitHub.
For us, this project is interesting because 3 XXE vulnerabilities were found there. They were fixed in BlogEngine.NET v3.3.8.0. This means we'll take the previous version for the experiment – v3.3.7.0. If you want, you can easily reproduce the described steps and see the real XXE yourself.
First, we download the desired version — v3.3.7.0. There should be no problems with building the project — it's very simple. I built the project with Visual Studio 2022.
After the project is built, we run it. If everything is successful, we'll see the site of the following type:
If the website is not available for other machines on the same network by default, I highly recommend you make it. A bit of configuring makes 'playing' with XXE more interesting.
When searching for vulnerabilities, you may have different inputs. For example, the system may represent a black box for you. Then you'll have to collect information about the system, search for influence points on it, and so on. If the system represents a white box, it changes the approach and the tools used to achieve the goal (or at least expands their list).
Here's an interesting thing about open-source projects. Seems like every person can work with the code and contribute to its quality / security. However, there are some drawbacks. On the other hand, hackers would have more ways to investigate the code — since they have access to the sources, they will easily find vulnerabilities. Would these vulnerabilities be reported?
There's no answer to this question. Let's get back to our business.
Since the project is open-source, we'll take advantage of this. To search for vulnerabilities, in addition to our own knowledge, we use PVS-Studio — a solution that searches for errors and security weaknesses. We need a group of security-related diagnostics — OWASP. You can read about turning on the corresponding warnings here.
In Visual Studio you need to set "Show All" for the OWASP group on the "Detectable Errors (C#)" tab: Extensions > PVS-Studio > Options > Detectable Errors (C#).
After that make sure that you enabled the display of the corresponding warnings. In this case we're interested in the 'OWASP' group of the 'High' certainty level. Thus, you need to click on the necessary buttons — they'll be framed.
Then, run the solution analysis (Extensions > PVS-Studio > Check > Solution) and wait for the results.
With the CWE filter (remember that XXE corresponds to CWE-611) or OWASP ASVS ID (OWASP ASVS 5.5.2) it is easy to find what we are interested in – 3 warnings V5614.
From the point of view of code, these errors are similar. We will analyze the most interesting one (located in several methods), and for the rest I will just provide basic information.
XMLRPCRequest.cs
Warning: V5614 [CWE-611, OWASP-5.5.2] Potential XXE vulnerability inside method. Insecure XML parser is used to process potentially tainted data from the first argument: 'inputXml'. BlogEngine.Core XMLRPCRequest.cs 41
In fact, the analyzer points at 3 lines to make the warning more understandable: a 'dangerous' method call, taint source, and a place where the tainted data is used by a dangerously configured parser.
public XMLRPCRequest(HttpContext input)
{
var inputXml = ParseRequest(input);
// LogMetaWeblogCall(inputXml);
this.LoadXmlRequest(inputXml); // Loads Method Call
// and Associated Variables
}According to the message, inputXml may contain tainted data (see taint checking) which is used by an insecurely configured parser inside the LoadXmlRequest method. Thus, it's a rather complex interprocedural case: data comes from one method (ParseRequest) and then is passed to another (LoadXmlRequest) where it's used.
Let's start with data — we need the ParseRequest method's code.
private static string ParseRequest(HttpContext context)
{
var buffer = new byte[context.Request.InputStream.Length];
context.Request.InputStream.Position = 0;
context.Request.InputStream.Read(buffer, 0, buffer.Length);
return Encoding.UTF8.GetString(buffer);
}Let's accompany the code with the taint distribution route, to make clear what we're talking about.
It all starts with the context.Request property that has the HttpRequest type. The analyzer considers it a taint source, since data received as a query may be compromised.
There are several ways to extract the data and working with a stream (the InputStream property) is one of them. Thus, the tainted data is passed to InputStream
Next, we call the System.IO.Stream.Read method for this stream. This method reads data from InputStream into the byte array (buffer). As a result, now buffer can also contain tainted data.
After that the Encoding.UTF8.GetString method is called. It constructs a string from the byte array (buffer). Since the source data for creating a string is tainted, the string is also tainted. After the construction, the string returns from the method.
So, the attackers may compromise the value returned by the ParseRequest method. At least in theory.
Let's go back to the original method:
public XMLRPCRequest(HttpContext input)
{
var inputXml = ParseRequest(input);
// LogMetaWeblogCall(inputXml);
this.LoadXmlRequest(inputXml); // Loads Method Call
// and Associated Variables
}Done with ParseRequest. Suppose that the inputXml variable can contain tainted data. Next step — analyze the LoadXmlRequest method that takes inputXml as an argument.
The method is long (100+ lines), so here's the shortened version. The fragment that triggered the analyzer is marked.
private void LoadXmlRequest(string xml)
{
var request = new XmlDocument();
try
{
if (!(xml.StartsWith("<?xml") || xml.StartsWith("<method")))
{
xml = xml.Substring(xml.IndexOf("<?xml"));
}
request.LoadXml(xml); // <=
}
catch (Exception ex)
{
throw new MetaWeblogException("01",
$"Invalid XMLRPC Request. ({ex.Message})");
}
....
}As we see, the argument is processed by an XML parser: request.LoadXml(xml). PVS-Studio thinks that request is vulnerable to XXE. Our job is to prove it. Or refute. Then this warning will be marked as false positive. Here we need the theory described in the beginning of this article.
The object type that the request reference points to is XmlDocument. The parser has default settings, which means we need to find out the .NET version. You can find it in the project's properties.
Now let's look at the table at the beginning of the article. We see that in applications on .NET Framework 4.5.1 and lower instances of the XmlDocument type are vulnerable to XXE by default.
It looks like we got all conditions for potential XXE:
Theoretically, this is an XXE, but it's still a potential vulnerability. We have to prove that the attack is possible. To do this we need to dig into the code a bit more.
We started our analysis with the constructor of the XMLRPCRequest type. It's called in one place:
internal class MetaWeblogHandler : IHttpHandler
{
....
public void ProcessRequest(HttpContext context)
{
try
{
var rootUrl = Utils.AbsoluteWebRoot.ToString();
// context.Request.Url.ToString().Substring(0,
// context.Request.Url.ToString().IndexOf("metaweblog.axd"));
var input = new XMLRPCRequest(context); // <=
....
}
....
}
....
}Yeah, we came across an HTTP handler. Here's an entry for it in the config:
<add name="MetaWeblog"
verb="*"
path="metaweblog.axd"
type="BlogEngine.Core.API.MetaWeblog.MetaWeblogHandler, BlogEngine.Core"
resourceType="Unspecified"
requireAccess="Script"
preCondition="integratedMode" />Now we know the address to send a request to and make the desired handler work. Let's try to reproduce the attack.
First, we need an XML file with which we'll steal data from the machine where the blog is deployed:
<?xml version="1.0"?>
<!DOCTYPE xxe [
<!ENTITY externalEntity SYSTEM
"file:///C:/Windows/System32/drivers/etc/hosts">
]>
<xxe>&externalEntity;</xxe>If an XML parser processes external entities, then instead of &externalEntity; it should paste the contents of the hosts file.
We make a request, send XML, and see how our handler will work. For convenience, it makes sense to save XML to a file (in this example - xxe.xml), so, if necessary, you can easily change its contents without changing the query command itself.
curl -d "@xxe.xml" -X POST http://vasiliev-pc:8081/metaweblog.axdSo, the handler caught our request and called the XMLRPCRequest constructor, which we inspected earlier.
Go inside the constructor and check the data in the inputXml variable.
Everything goes according to plan — the data is tainted, as we assumed (and wanted), and is passed to the LoadXmlRequest method as an argument. Let's observe further.
Due to the dangerous default settings, the parser worked exactly as we expected – it loaded the contents of the hosts file. Then the following code fragment is executed:
// Method name is always first
if (request.DocumentElement != null)
{
this.MethodName = request.DocumentElement.ChildNodes[0].InnerText;
}Luckily (for the hacker :)) the contents of the hosts file will be written to the MethodName property — exactly what we need. The next code fragment we need is large switch, where certain actions are performed depending on the method name:
switch (this.MethodName)
{
case "metaWeblog.newPost":
....
break;
case "metaWeblog.editPost":
....
break;
case "metaWeblog.getPost":
....
break;
....
default:
throw new MetaWeblogException("02", $"Unknown Method. ({MethodName})");
}Here we need the default branch to where execution will go since there's no suitable method. In this branch an exception is thrown. The exception's message will have the name of the method for which the mapping failed. In our case, the method's name is the contents of the hosts file.
When an exception is thrown, we return to the handler and get to the catch section where an unknown method is reported:
As a result, to our initial request:
curl -d "@xxe.xml" -X POST http://vasiliev-pc:8081/metaweblog.axdWe get the following answer:
So, we managed to obtain the contents of the hosts file, using an XXE attack. We got it on the machine with a deployed blog. If we know the location of other files, we can try to get their contents as well. And not only from the attacked machine, but also from other machines of the network to which we have access. Here, in the context of network requests, we can also talk about SSRF.
So, we have just seen XXE both from the point of view of the application (code) and from the point of view of the user (attacker). This is a real vulnerability – CVE-2018-14485 (here is the entry in the NVD).
What should we do with vulnerabilities? That's right, fix it. The commit can be found here. After that, the XML parser's configuration was changed, so now it can't process external entities. To do this, it is enough to set the value of the XmlResolver property to null:
var request = new XmlDocument() { XmlResolver = null };Now if we try to get the same hosts file, it won't get into the output.
By the way, PVS-Studio knows that the parser with this configuration (XmlResolver – null) won't process external entities. Thus, the analyzer won't issue a warning for the fixed code.
Two other warnings that we've seen before also point to vulnerabilities. We are not going to analyze them (the code is similar), but below is basic information about them.
CVE-2019-10718
CVE-2019-11392
Now you are a little more savvy in security and XXE issues, and also know that even a simple blog deployed on your machine can become a source of vulnerabilities.
In fact, the XXE theme is more serious and, of course, there is still a lot to dig into. But at least just knowing about this security flaw and understanding it at a basic level will already be useful.
Praemonitus, praemunitus.
As always, I invite you to subscribe to my Twitter so as not to miss anything interesting.
0
0
0
0