Guardora VFL

Key features

• Supports tabular data.
• Ready-to-use training algorithms:
  • Linear regression.
  • Logistic regression.
  • Softmax regression (multiclass classification).
  • Decision-tree-based gradient boosting.
• User-friendly API for integration into your existing ML pipelines.

For whom:

• Financial sector.
• Telecom providers.
• Retail chains, marketplaces.
• Other scoring service providers.

Security

All raw data computations are performed within the data owner’s perimeter without transferring data to other parties.

1. Create a cloud network and a subnet to host the virtual machine (VM).

2. Create a security group in the cloud network and configure the rules as follows:

   Traffic direction	Description	Port range	Protocol	Source / Destination	CIDR blocks
   Ingress	Swagger API	7171	TCP	CIDR	0.0.0.0/0
   Ingress	MLFLow	5555*	TCP	CIDR	0.0.0.0/0
   Ingress	TensorBoard	6006*	TCP	CIDR	0.0.0.0/0
   Ingress	SSH	22	TCP	CIDR	0.0.0.0/0
   Egress	Any Egress	0-65535	Any	CIDR	0.0.0.0/0

   ^* Port used to monitor model quality metrics.

3. Get an SSH key pair for connection to the VM.

4. In the Marketplace, find Guardora VFL and click Create VM:

   1. Under Network settings, specify the cloud network, subnet, and security group you created earlier.

   2. Under Access, specify the username and public SSH key for connection to the VM.

   3. Complete setting up your VM and click Create VM.

      Wait for the VM to be created and the application to be installed: this may take up to 10 minutes.

      After the VM is created, the API becomes available automatically.

5. Learn the VM’s public IP address.

6. To connect to Swagger’s web UI, open this URL in your browser:

   http://<VM_public_IP_address>:7171/docs
   

7. To initialize a node (create a party):

   1. In Swagger’s web UI, under Node, expand Post Node.

   2. Click Try it out.

   3. In the Name field, specify the organization name.

      The name may contain Latin letters, digits, and spaces.

   4. Click Execute.

      The request’s properties and result will be displayed below.

Training scoring and anti-fraud models on third-party data

• What it can help with:

  • Improves scoring accuracy by using partner data without direct access to it.
  • Detects complex fraud schemes using cross-corporate data.

• For whom:

  • Credit bureaus.
  • Scoring service providers.
  • Banks and microlenders.
  • Insurance companies.
  • Payment systems.
  • Telecom providers.

Creating collaborative ML products

• What it can help with:

  • Development of new products that combine data from several companies.

• For whom:

  • Scoring service providers.
  • Fintech companies.

Using data within corporate groups

• What it can help with:

  • Overcoming internal data exchange barriers between the companies of the group.
  • Creating unified models for the whole group.

• For whom:

  • Fintech holdings.
  • Corporate groups, including cross-border ones.

Pilot projects and retro-tests

• What it can help with:

  • Rapid testing of hypotheses based on real data of potential customers.
  • Product value demonstration without risking a data leak.

• For whom:

  • Scoring service providers.
  • Providers of personal data processing solutions.

Guardora
The Guardora team provides 24/7 technical support to Guardora VFL users. If you have questions or issues, email us at help@guardora.ru.

Yandex Cloud
Yandex Cloud does not provide technical support for this product. If you have any issues, please refer to the vendor’s information resources.