Skip to main content

Part I: Data Owner - FL Client

The Data Owner provides the secure environment and raw data required for training.

1. Setup & SyftBox Execution​

  • Install: Run the SyftBox installation script. This establishes your local machine as a "datasite."
  • Run Client: Start the SyftBox client. It will run as a background service, managing secure file synchronization with the network.
  • Login: Your client automatically registers you. Verify your identity by checking your local config at ~/.syftbox/config.json.

2. Datasite Administration​

  • Admin Access: Launch the RDS-Dashboard (via Docker). This is your command center.
  • Login: Open localhost:8000. You are automatically logged in as the Admin of your local datasite.

3. Creating Syft Datasets​

  • Private Data: Upload your sensitive diabetes_train.csv. This data never leaves your machine.
  • Mock Data: Upload a synthetic diabetes_mock.csv with identical columns (e.g., Glucose, BMI, Age). This allows Data Scientists to write code without seeing real patient info.
  • Metadata Sync: Once created in the Dashboard, the metadata (name and schema) is synced to the network, making it "discoverable."

4. Review & Approval​

  • Audit: When a job arrives, open the Jobs tab. Use the "View Code" feature to audit the Python scripts.
  • Execute: Click "Approve" and "Run". SyftBox will now execute the Data Scientist's model training locally on your private data.