Configure an MPC project

Prerequisites

Before you configure an MPC project, ensure that you have completed the following operations:

1. You have installed and deployed nodes, granted authorization between the nodes, and ensured that the nodes use both training and prediction engines. For more information about deploying nodes and granting authorization, see Privacy-preserving computation nodes.

2. You have added the data tables required for the MPC project. For more information, see Data tables.

3. Confirm that the online features for the MPC project have been published. For more information, see Publish Features.

4. The members required for the MPC project have been added to the console. For more information, see Add a member.

Step 1: Configure nodes

Grant authorization for the MPC project to the nodes created by your organization. These nodes store the authorized data for the project.

1. Log on to the Ant Privacy-Preserving Computation Platform.

2. In the navigation pane on the left, click Privacy Computing Projects.

3. On the Privacy Computing Projects page, find your project on the Initiated Projects or Invited Projects tab.

4. On the Privacy Computing Projects page, click Resource Configuration to the right of the target MPC project.

5. On the Node Configuration tab of the Resource Configuration panel, configure the following parameters.

   Parameter	Description
   Training node	Select a node. The node must be a node that uses a training engine from the Node Management module.
   Model transfer method	The following two model transfer methods are supported: Platform-encrypted transfer The model file of the training node is encrypted and transferred to the platform. The platform then downloads the model file to the prediction node. The file is decrypted on the prediction node. Therefore, configure the same key pair when you deploy the training and prediction nodes. For more information about how to configure a key pair, see Node configuration and installation. Local node transfer The model file of the training node is transferred directly from the local machine to the prediction node. Therefore, ensure network connectivity between the training and prediction nodes in the project.
   Staging environment	Select a node to serve as the privacy-preserving computation node for the staging environment. The node must be a node that uses a prediction engine in a staging environment from the Node Management module.
   Production environment	Select a node to serve as the privacy-preserving computation node for the production environment. The node must be a node that uses a prediction engine in a production environment from the Node Management module.

6. After you complete the configuration, click Save. In the dialog box that opens, click OK.

   If the Operation Successful message appears, the node configuration is successful.

Step 2: Configure data

Grant authorization to the data that the project requires. The data is granted to the nodes that you authorized in Step 1: Configure nodes. This lets you use the authorized data in the project. Data configuration involves configuring offline samples and online features.

Configure offline samples

Offline samples are the raw data in an MPC project. They are stored on nodes and can be used only locally. To use offline samples in a project, you must grant them authorization for the project.

1. On the Privacy Computing Projects page, click Resource Configuration to the right of the target MPC project.

2. In the Resource Configuration panel, click the Data Configuration tab, and then click Add Data Table Authorization.

3. In the Add Data Table Authorization panel, add the target data table in the Select Data Table section.

   Note

   Ensure that the target data table is in the Available state on the target node. If the target data table is not found, see Add a data table, and ensure that the new data table is in the Available state.

4. In the Data Table Alias section, enter an alias for the data table. By default, the data table name is used as the alias.

   In an MPC project, data tables cannot have the same name. If multiple participants in a project have data tables with the same name, you can use aliases to distinguish them.

5. Select the Association Key or Group Column checkbox for the target field, and then click Save.

   An association key is a field used in a JOIN ON SQL statement for private set intersection. A group column is a field used for conditional splitting during model training. To set fields in the table schema as association keys or group columns, you can select the Association Key or Group Column checkbox to the right of the relevant field.

   Note

   • In the Offline Samples section of the Resource Configuration panel, you can click Revoke Authorization to the right of a data table to revoke its authorization.

   • After the authorization is revoked, the data table is removed from the MPC project. However, the virtual wide tables and models generated from this data table are not affected.

Configure online features

Online features are required for online prediction with a multi-party secure model. To use online features in a project, you must grant authorization for the MPC project to the published online features.

1. In the Resource Configuration panel, click the Data Configuration tab.

2. In the Online Features section, under the staging environment, click Add Feature API Authorization.

3. In the Add Feature API Authorization panel, add the target feature group, grant authorization to the required online features, and then click OK.

   These are the online features that you prepared as described in the Prerequisites section.

   Note

   • In the Online Features section of the Resource Configuration panel, you can click Revoke Authorization to revoke the authorization of a feature group.

   • After the authorization is revoked, the feature group is removed from the MPC project. However, deployed model services are not affected.

4. Repeat the previous two steps to add feature API authorization for the production environment.

5. After all configurations are complete, click Save in the lower-right corner of the Resource Configuration panel.

   If the Operation Successful message appears, the data configuration is successful.

Step 3: Configure members

Configure the members who will participate in the project. After a member is granted project authorization, they can access the project console to begin development.

1. On the Privacy Computing Projects page, click Member Configuration to the right of the target MPC project.

2. In the Member Configuration panel, click Add Member.

3. Select a member that is associated with your organization and click Save.

   If the Operation Successful message appears, the member configuration is successful.