Node Setup

To participate in federated learning on Unity Network, each organization must provide a dedicated node for performing model training. This node, fully controlled and owned by the organization, ensures that all data processing remains within the organization’s secure environment. Unity’s SDK handles node management, facilitating seamless and secure integration with Unity Network.

1. Provisioning the Node

• Node Ownership: The organization is responsible for setting up and maintaining a node, which will be used exclusively for model training.

• Security and Privacy: Since the node remains under the organization’s control, it provides full data privacy, ensuring that no raw data is exposed outside the organization’s environment. Only model updates (not data) will be shared with Unity Network after training.

2. Node Registration on Unity Portal

• Submitting Node Details: During the registration process, the organization must enter the necessary node specifications in the Unity portal. This includes information such as the IP address or connection endpoint, node configuration, and performance details.

• Performance Specifications: Unity Network requires nodes to meet specific performance standards to ensure efficient training and secure operations. Minimum CPU, memory, and storage requirements are outlined. These specifications ensure that the node can handle the demands of federated learning.

3. Temporary IAM Role Creation

• Security Role Setup: The organization creates a temporary IAM (Identity and Access Management) role to provide Unity Network with controlled access to the node for verification purposes.

• Permissions: This IAM role grants Unity Network limited, temporary permissions to perform initial setup and verification checks on the node. The permissions are designed to be as restrictive as possible, ensuring that Unity Network can only conduct necessary tests without accessing sensitive data or configurations.

• Submission: Once created, the IAM role credentials are securely submitted through the Unity portal, allowing Unity to initiate verification.

4. Node Capability Verification by Unity Network

• Basic Testing: Unity Network performs a series of automated tests on the node to verify its compliance with required specifications. These tests check:

  • Processing Power: Ensuring the node meets CPU and GPU requirements for effective model training.

  • Memory and Storage Capacity: Verifying that there’s sufficient memory and storage to handle large datasets and training processes.

  • Network Speed and Reliability: Confirming the node’s network connection speed and stability, as federated learning relies on timely and secure transmission of model updates.

• Security Compliance: Unity also verifies that the node is configured with recommended security protocols to protect data during training and transmission.

5. Finalizing Node Approval

• Temporary Role Removal: After the node’s capabilities are verified, the organization can safely remove the temporary IAM role. Unity Network’s testing process is complete, and the SDK will handle all further node interactions securely.

• Node Activation: Once verification is complete, the node is activated on the Unity Network, and it’s ready to begin processing model training requests as part of the federated learning workflow.

6. Ongoing Node Management through Unity SDK

• Automatic Node Management: The Unity SDK continuously monitors and manages the node to ensure it remains compliant with Unity Network’s standards. This includes checking for performance consistency, managing secure model updates, and maintaining connection stability.

• Node Monitoring and Logging: The SDK also logs key performance metrics and provides insights into the node’s activity for the organization. This monitoring is designed to be lightweight and minimally intrusive, keeping the node running efficiently while ensuring data security.