Introducing a tool for exfiltrating all of the things from a Google Drive
I am a Master’s of Information Security Engineering student at SANS Technology Institute (STI). I was doing my reading the other night for my SEC 530: Defensible Security Architecture and Engineering course and got into the topics of Data Classification and Data Loss Prevention (DLP). This material covered the many different methods of classifying data and of implementing DLP from a detections perspective. I am professional Red Teamer, and thus began thinking about the many different times that I had hacked my way into a victim’s Google Drive (GD) and exfiltrated data and was never detected. I then started thinking of ways that I could test DLP detections on data leaving GD. How could I best show the impact of lack of DLP detections on data leaving a GD? Most of the time I just manually searched and pillaged through a GD when looking for sensitive data to exfiltrate. What if I could develop an app that would allow me to programmatically, and easily exfiltrate an entire drive, or even just exfiltrate files that contain sensitive data? This would allow me to test DLP detections an exfiltrate data at a large enough scale to show maximum impact to the organization.
This is what I did. Enter GD-Thief.
What is GD-Thief?
Have you ever been on a Red Team Operation where you have access to a victim’s G-Suite, you are rifling through the victim’s GD, and you wished you had a way to quickly download all of the files? Well that’s what GD-Thief does. GD-Thief is a script I wrote in python that allows the Red Teamer to download every file in the victim’s drive. But, wait there’s more!
GD-Thief not only downloads all of the files in the victim’s drive, but it also downloads all of the files from drives shared with the victim, and all of the drives in the victim’s corporate domain that the victim has access to. So, you are getting it all! But, wait there’s more!
GD-Thief allows for the Red Teamer to only search for files that contain specific strings via a wordlist. So, if you want to look for only files that contain the words, “password,” “username,” or “pastrami” just add the words to your wordlist/dictionary and GD-Thief will search all of the files for these keywords within the file’s entire content, and download only those files of interest. But, wait there’s more!
You think searching text file content for words of interest is cool, what if I told you GD-Thief performs Optical Character Recognition (OCR) to search for the same specified strings in image files and pictures. Yeah, so if the victim took a picture of the menu at their local Capriotti’s sandwich shop and stored it in their GD, GD-Thief will find the “pastrami.”
How does it work?
GD-Thief uses the Google Drive API to make calls to the victim’s GD. The Red Teamer will need to create an external, testing, Google App, get an API key for the App, give the App permissions to access the required GD APIs, and will need to add the victim’s Gmail address to the list of application testers. I will explain how all of this is done in the tutorial below.
Now, adding testers to the application’s tester list does not inform the individual that you add that they have been added to the list. They do not need to give permission to be, acknowledge that they are, nor authorize that they be added to the tester list. I find this to be a security issue, and disclosed this to Google on April 23rd, 2021. Google responded on April 26th, 2021 and stated that this is did not meet their criteria for abuse, it is intended behavior and that it would not be fixed.
Yes, the attacker must already have access to the victim’s Google account in order to use this application to do harm, but as an extra layer of security, for detection and alerting purposes, or for an early-warning purpose this just might allow the victim to become aware that they are under attack. Especially if it required the target to authorize being added to the tester list using two-factor authentication (2FA).
A Very Effective Attack Path
Getting access to a target’s Google account can be accomplished fairly easy. I have accomplished this on MANY Red Team operations via a Machine-in-the-Middle (MitM) attack using EvilGinx2 to create a fake Okta page for the target organization. EvilGinx2 basically proxies the authentication traffic sent to the fake web application page (Okta in this case) and forwards it to the real web application page using the credentials provided by the victim. It logs the victim in to the real web application and sends the credentials to the attacker controlled, EvilGinx2 server. This even bypasses 2FA. Let us discuss how this works in greater detail:
- Phishing Email Sent to target via a Red Team controlled phishing server (GoPhish). Phishing email pretext can be something to do with the target needing to make a change on Workday.
- Victim clicks a malicious link and is sent to a fake Okta, login page on a Red Team controlled server (EvilGinx2)
- Victim inputs their Okta credentials and EvilGinx2 proxies the credentials to the victim’s real Okta login portal.
- The real Okta sends the 2FA push notification to the victim as normal.
- Victim accepts the push and is then forwarded by EvilGinx2 to a dummy web page of the Red Team’s choosing. Sticking to the phishing pretext described in step 1, this dummy page could be the victim’s Workday site.
- EvilGinx2 gives the Red Team the victim’s credentials as well as the victim’s Okta session token. The Red Team now has an interactive session for the victim’s Okta account.
Now that you have an Okta session, you are almost guaranteed to have a G-Suite session, and thus a GD session as long as the victim organization hosts a Google, Okta tile in their Okta portal. Now you are good to go exfiltrating files from the victim’s GD.
A Tutorial for Using GD-Thief
Before you ever get a session on a victim’s Google account as described above, you need to do a few things. First you’ll need Python3. Next you’ll need to download GD-Thief from GitHub:
git clone https://github.com/antman1p/GD-Thief.git
Next you’ll need to get the Google API, Python libraries:
pip install — upgrade google-api-python-client google-auth-httplib2 google-auth-oauthlib
Now you will need to follow the steps below to give GD-Thief access to the GD API.
Create a new Google Cloud Platform (GCP) project
Steps to get the Google API Access Token needed for connecting to the API
- Create a burner Gmail/google account. You will need to provide a credit card number to start your free Google Cloud account (Get a prepaid CC)
- Login to said burner Google account
- Navigate to the Google Cloud Console
- Next to “Google Cloud Platform,” click the “Select a project”
Down arrow. A dialog listing current projects appears.
New Project. The New Project screen appears.
6. In the
Project Name field, enter a descriptive name for your project.
7. (Optional) To edit the
Project ID, click
Edit. The project ID can't be changed after the project is created, so choose an ID that meets your needs for the lifetime of the project.
Create. The console navigates to the Dashboard page and your project is created within a few minutes.
Enable a Google Workspace API
- Next to “Google Cloud Platform,” click the
Down arrowand select the project you just created from the dropdown list.
- In the top-left corner, click
APIs & Services.
Enable APIs and Services. The "Welcome to API Library" page appears.
4. In the
search field, enter "Google Drive".
5. Click the Google Drive API to. The API page appears.
Enable. The Overview page appears.
Configure OAuth Consent Screen
- On the left side of the Overview page click
Credentials. The credential page for your project appears.
Configure Consent Screen. The "OAuth consent screen" screen appears.
3. Click the
External user type for your app.
Create. A second "OAuth consent screen" screen appears.
5. Fill out the form: Enter an Application Name in the
App name field. Enter your burner email address in the
User support email field. Enter your burner email address in the
Developer contact information field.
Save and Continue. The "Scopes" page appears.
Add or Remove Scopes. The "Update selected scopes" page appears.
8. Check all of the
Google Drive the scopes to use in the app. GD scopes cover 2 pages, so click the next page and ensure that you check them all.
Update. A list of scopes for your app appears.
Save and Continue. The "Edit app registration" page appears.
Save and Continue. The "OAuth consent screen" appears.
Create a credential
Create Credentialsand select
OAuth client ID. The "Create OAuth client ID" page appears.
2. Click the Application type drop-down list and select
3. In the
name field, type a name for the credential. This name is only shown in the Cloud Console.
Create. The OAuth client created screen appears. This screen shows the
Client ID and
OK. The newly created credential appears under "OAuth 2.0 Client IDs."
6. Click the
download button to the right of the newly-created OAuth 2.0 Client ID. This copies a client secret JSON file to your desktop. Note the location of this file.
7. Rename the client secret JSON file to “credentials.json” and move it to the
Add the victim’s Google account to the Application’s Test Users
In order to be able to run this script against the victim, you will need to add their Google account to the Test Users list for the App you just created.
This is the part that I reported to Google since it does not inform the user you add to this list that they have been added to an application for testing, nor does it require their authorization.
- On the Left side of the screen click
OAuth consent screen. You "OAuth Consent Screen" page appears.
Test Usersclick the
3. Enter the victim’s Gmail address in the
email address field.
4. Click the
First Time running gd_thief
Upon gaining access to a Target’s Google account, you can run gd_thief
- The first time running gd_thief, the script opens a new window prompting you to authorize access to your data:
- If you are signed in to multiple Google accounts, you are asked to select one account to use for the authorization. Make sure you select the victim’s Google account
Usage for GD-Thief is pretty simple. I won’t rewrite the usage here since it is fully described on the GD-Thief, GitHub README and in the application help, but just make sure you are using Python3.
python3 gd-thief.py …
Hopefully this tool comes in handy for you on your next Red Team operation, or for a DLP penetration test. Next, I would like to add functionality for a search option that uses Snort’s Sensitive Data Processor regex to find sensitive data in the GD like credit card numbers, social security numbers, email addresses, etc.…
For more Google API abuse, checkout my blog post and tool for Google Calendar Phishing:
A special thanks to my good friend Cedric Owens for helping me with the threading on this app. The mass download FLIES now!
As always, thanks for reading!