Sensity Deepfake Detection for Microsoft Teams
Complete setup and usage guide
Overview
Sensity Deepfake Detection is a Microsoft Teams app that analyzes participants' live audio and video streams for signs of deepfake manipulation. Once installed and configured, it runs automatically during meetings, flagging any suspicious patterns in real-time.
Prerequisites
- Microsoft Teams: Admin rights to add apps at the organization level.
- Sensity Account: A valid Sensity subscription (trial or paid).
- Permissions: Your Teams tenant must allow third-party app installations and Bot Framework integrations.
Installation
Open Teams Admin Center
Go to Teams Apps → Manage apps (or directly within the Teams client, click Apps in the sidebar).
Search for "Sensity Deepfake Detection"
In the marketplace search bar, type Sensity Deepfake Detection.
Add to Organization
Select the app and click Add. Under Setup policies, ensure it's published to all users or specific groups as needed.
Make Available in Meetings
Under your org's App permission policies, verify the app is allowed in meeting contexts.
Initial Configuration
The first time you add Sensity Deepfake Detection, Teams will redirect you to a configuration page.
Authorize Your Microsoft Account
Click the Connect Sensity button on the configuration screen.
Log in to Sensity
You'll be taken to your Sensity dashboard. If needed, sign in or create an account.
Enable the Teams Integration
In the Sensity app, navigate to Integrations → Teams and click Connect.
Return to Teams & Save
Close the Sensity window, return to Teams, then click Save on the configuration page.
Using the App in a Meeting
Open Your Meeting
Add the App (One-Time)
Click Apps (or the "+" tab) at the top of the meeting. Search "Sensity Deepfake Detection" and select it.
Access the Side Panel
Once added, the Sensity icon appears among your meeting tabs—click it to open.
Deepfake Detection Workflow
View Participant List
The side panel displays every attendee's name and current status (Valid / Suspicious).
Start Analysis
Click Analyze at the bottom. A small processing indicator will appear.
Processing
Behind the scenes, our detection bot joins the call and captures 30 seconds of audio/video.
Results
After processing, each participant's status updates to Valid or Suspicious.
Audio vs. Video Detection
📹 Video Detection
Always runs when the bot sees a face.
🎤 Audio Detection
Requires the participant to speak at least once during the 30 second window.
If no audio is captured, only video analysis results are shown.
Re-Running an Analysis
If you need fresh data or someone missed speaking:
- Click Analyze again.
- The bot re-joins (or stays) and captures another 30 seconds.
- New statuses overwrite previous results.
Best Practices
🗣️ Inform Participants
Let attendees know they'll be analyzed—this ensures they speak and face the camera.
🔇 Quiet Backgrounds
Minimize noise to improve audio detection accuracy.
💡 Stable Lighting
Good lighting helps the video detector distinguish facial features.
Troubleshooting
Support & Feedback
For additional help or to report bugs: