During a recent red team engagement, we found that the target organization used a well-known identity access management (IAM) product for their multi-factor authentication (MFA) solution. Most of their Internet-facing login portals, such as Office365 and Citrix portals, were behind this IAM platform; however, their users were allowed to enable multiple MFA options, including Email OTP (one-time passcode). Given these situations, we thought a phishing campaign abusing Microsoft 365 OAuth device code flow would be a great attack vector for potentially bypassing their MFA and gaining access to the applications from the Internet.

Image

Figure 1: Email OTP Enabled for MFA

Microsoft OAuth Authentication Flow

According to Microsoft, it uses OAuth device authorization grant to allow users to sign in to input-constrained devices such as smart TVs, IoT devices or printers. To enable this authorization flow, the device will have the user visit a webpage from another device’s browser to sign in. Once the user signs in, the device can request access tokens as needed.

Image

Figure 2: Microsoft OAuth Device Code Flow

To explain what this is doing:

How Can We Abuse This for Phishing?

Practical Example of Microsoft OAuth Device Code Phishing

One perk about this phishing technique is that we do not need to build custom phishing infrastructure (e.g., creating a phishing server and landing page). We can use pretty much everything provided by Microsoft since we are technically leveraging their legitimate OAuth authentication flow.

Creating Azure Application

First, we need to create an Azure application. This application will serve as an endpoint where we will generate client_id and device_code, which are later used for obtaining the victim’s access_token, refresh_token and id_token.

Steps: Azure Portal → Azure Active Directory → App registrations → New registration

Image

Figure 3: Azure App Registrations

Next, enter a name for the application. This can be anything such as “Optiv Remote User Management”, etc. For Supported account types, select “Accounts in any organizational directory (Any Azure AD directory – Multitenant) and personal Microsoft accounts (e.g., Skype, Xbox)” so that target users could use their business Microsoft accounts to allow the application consents.

Image

Figure 4: Azure Register an Application

This will create our Azure application and make a note for the Application (client) ID value. This will serve as client_id value.

Image

Figure 5: Azure App client_id

Finally, change Allow public client flows to “Yes” under the Authentication section and select “Save.”

Image

Figure 6: Azure App Authentication

Launching Phishing Campaign

To conduct the device code phishing, we first need to make a POST request to /devicecode endpoint to obtain user_code and device_code. Postman, an API testing and development platform, can be very useful for this.

*Note: A full list of the scope can be found at Microsoft.com. Make sure to carefully check permissions required for each scope since some require admin consent.

Image

Figure 7: POST Request to /devicecode Endpoint

Next, upon obtaining the user_code, we need to send a phishing email to the victim. This is the only tricky part of this phishing attack since you need to lure the victim to visit the verification_uri and enter the user_code. Another caveat is that obtained user_code and device_code will expire 15 minutes after they have been generated. Thus, if they expire, you will be required to regenerate them and resend the phishing email.

For our engagement, we came up with the following phishing email template:

Image

Figure 8: Example Phishing Email

We registered our phishing domain to send emails from Microsoft Outlook. Also, the reason why we did not embed the hyperlink for the verification_uri (https://microsoft.com/devicelogin) was that our phishing email kept landing in the spam folder. Though we are not 100% sure, we think Microsoft may have done some subtle filtering for hyperlinked verification_uri and user_code combinations not originating from a Microsoft email domain. We observed this behavior from our dynamic testing at that time, so we decided to make our phishing email like the above.

If the phishing attempt was successful and the victim visited the verification_uri, entered the user_code, and completed authentication, the victim would have seen the following devicelogin and permission consent pages:

Image

Figure 9: Microsoft Devicelogin Page for Entering user_code

Image

Figure 10: Example Sign in Page

Image

Figure 11: Example Device Permissions Requested Page

Image

Figure 12: Example Final Page After Accepting Permissions

As for the attacker, they will need to keep checking whether the victim completed the authorization process by sending a POST request to /token API endpoint after the phishing email is sent. Once the authorization process is completed, it will present a bearer token formatted access_token and refresh_token, which you can use to mimic the victim’s access.

Image

Figure 13: Authorization Pending

Image

Figure 14: Authorization Granted & Gain Bearer access_token

For example, with the access_token, the attacker can now access the victim’s email inbox using Microsoft Graph API.

Image

Figure 15: Accessing Victim's Outlook Mail Inbox

Automation of Device Code Phishing

Going back to our red team scenario, our goals were to:

We originally found a great open-source framework, TokenTactics created by revrsh3ll to conduct our phishing attack. The tool was designed to contain all the necessary functions and tactics to abuse Azure JSON Web Token (JWT). We highly recommend trying it out. Leveraging his tool, however, we decided to create a simple Python script to automate some of the device code phishing flow to solely meet our goals. The first function getDeviceCode() was created to request for user_code and device_code.

Image

Figure 16: getDeviceCode

Image

Figure 17: Example of Running devicePhish.py – Getting user_code & device_code

Next, using the newly obtained device_code, the getAccessToken() function continuously loops through the authorization process. If the phishing attack is successful, it will yield an access_token and refresh_token; otherwise, the loop will end after 15 minutes due to the token expiration.

Image

Figure 18: getAccessToken

Image

Figure 19: Example of Victim Entering user_code

Image

Figure 20: Example of Running devicePhish.py – Obtaining access_token & refresh_token

With access_token, the getMail() function keeps looking for an email containing specific text. For our case, it was looking for a string “MFA – One Time Code” within the email text, so if we were to trigger the Email OTP, it would have sent the temporary code to the victim’s inbox. If it finds the right email, it prints out the email context with the OTP code and saves the email_id value.

Image

Figure 21: getMail

Image

Figure 22: Example OTP Email

Image

Figure 23: Example of Running devicePhish.py – Reading OTP Email

Finally, with the stored email_id value, the deleteMail() function will delete the OTP email from the victim’s inbox.

Image

Figure 24: deleteMail

Image

Figure 25: Example of Running devicePhish.py – Deleting OTP Email

The script used in this demo can be found here: Microsoft365-devicePhish

Mitigation of Microsoft OAuth Device Code Phishing

This device code phishing attack can be mitigated by making configuration changes for Azure AD. Under the Enterprise applications settings, prevent users from giving consent to all applications.

Image

Figure 26: Mitigation

If the above solution is too restrictive, then please check out the following best practices provided by Microsoft:

References