openmined/pyariesfl

Federated Learning on HyperLedger Aries

Distributed Deep-learning with HL-Aries

This repo has been forked from Hyperledger Aries Cloud agent link provided by British Columbia Government.

We have created a set of agents that can be run using docker to experiment with a federated deep learning scenario using didCOMM to pass secure messages between the participants - e.g. the ML model parameters. Furthermore the scenario explores using credentials so that participants can verify that who they are communicating with. In the example this is verifying a hospital is indeed a hospital and a researcher has been accredited by a regulator. However the credentials abnd issuers within this use case could be adapted to any ecosystem.

The agents connect, by default, to the BC government development instance of Hyperledger Indy ledger link.

Guide

To run the scenario:

git clone https://github.com/blockpass-identity-lab/aries-fl-demo.git

cd demo

Hospital gets verified by the NHS Trust.

1. Start the NHS Trusted agent.

   ./run_demo nhsheadoffice
   

2. Start each of the hospitals in separate terminals.

   ./run_demo hospital1(2,3)
   

3. Copy NHS's invitation URL into the hospitals terminal to initiate DIDExchange and establish pairwise DID connection.

4. Issue verified hospital credential to the Hospital. This first requests a self attested proof of hospital name from the hospital then issues this name in a credential.

5. Create new invitation from the NHSTrust and repeat steps 3 and 4 for each hospital

The Researcher gets accredited by the regulatory authority.

6. Start the Coordinator.

./run_demo coordinator

7. Start the regulator.

./run_demo regulator

8. Enter command to allow coordinator to accept new invitation

9. Copy regulator's invitation URL into coordinator terminal.

10. Issue a credential from the Regulator to the Coordinator.

The researcher validates the hospital's credential. The hospital validates the researcher's accreditation. The training procedure starts.

11. Create New Invitation in the coordinator for the hospital.

12. Input new connection into the hospital terminal

13. Coordinator request proof connection is from a valid hospital. If proof is valid, the coordinator adds this hospital to the list of the trusted connections.

14. Hospital to request proof that the coordinator connection is from a certified research. If so add to list of trusted connections.

15. Either add another trusted connection between a new hospital and current researcher by repeating steps 11 - 14

or..

16. In the coordinator terminal initiate training process which loops through all trusted hospital connections in turn, sending a message containing the model. The hospital checks if the message is from a trusted research, if so they initiate the training process using their private date. Once complete the trained model is sent back to the coordinator who validates the model and forwards it onto the next trusted connection.

Created by

Will Abramson (wip-abramson)
Adam James Hall (H4LL)
Pavlos Papadopoulos (pavlos-p)

pyariesfl's Issues

Adapt FL example to include weighted average

• Make NHS Head Office issue rating to hospitals
• Coordinator request proofs of rating from hospital
• Coordinator tracks rating for each hospital
• Coordinator aggregates training result based on weights for each hospital

Create new Basic Message Test

Rename project

Description

We need to rename this project slightly to fit our new project naming standards.

The new name for this repo should be: PyAriesFL
The new release name for this repo should be: openmined/ariesfl

I have gone ahead and renamed the repo, but I will need someone else to be in charge of changing the release name.

If you have any questions about this issue, please comment here or message @cereallarceny on Slack.

Create new message type in aries agent code.

Initially, simply look to replicate the code in basic message and give it a new name.

See - https://github.com/hyperledger/aries-rfcs/tree/master/features/0095-basic-message

Create a better Readme

Based on this readme template, we should improve our readmes across all OpenMined projects.

Please consult the following document for our desired best practices.