Overview

Motivation

Since we have FedFT (see platform) algorithm whose correctness is verified by some experiences, devices, and student resources on our campus, we can make this theory a practical application. Therefore, we decided to develop a FedCampus platform based on FedFT algorithm and may cater to more FA algorithms in the future.

Preview

This project implements an Android App to collect health and exercise data for FL training. It has two main function: data collection and local differential privacy (DP). Besides, this project also needs support iot for FL training, e.g: using smart watch.

Functions (HealthKit Application)

• Data Collect: data from wearable devices.

  • Store data in a secure and encrypted format, and apply privacy-preserving technologies to the data, such as differential privacy.

• Data Analytics: Process data to extract insights and patterns.

  • Data cleaning: This involves identifying and correcting errors in the data.

  • Data transformation: This involves converting the data into a format that is easier to analyze.

  • Data analysis: This involves using statistical and mathematical techniques to identify patterns and trends in the data.

  • Data visualization: This involves creating graphs, charts, and other visuals to help communicate the findings of the data analysis.

• Data Collaboration: Share data with authorized users, recommendation, FL...

Additionally functions:

• Data Collaboration: Share data with authorized users, recommendation, FL...

Person	Work	Time Due
Jingheng, Beilong	Data Collection from wearable devices.	By the end of May
Beilong	Backend & privacy techniques (Store data in a secure or encrypted format and apply privacy-preserving technologies)	Mid of May
Beilong	Data Collaboration & Privacy techniques for FL	By the End of Summer
Jingheng, Shuhao	App Implementation - Frontend	End of June
Jingheng	Data Analytics (Data cleaning, Data transformation, Data analysis, Data visualization)	By the End of Summer

Resources

• Apollo server: https://blog.csdn.net/weixin_39808143/article/details/112576683

• Ali Cloud: https://help.aliyun.com/document_detail/163032.html

• Health Kit: https://developer.huawei.com/consumer/cn/doc/development/HMSCore-Guides/apply-kitservice-0000001050071707

• Health Kit Data: https://developer.huawei.com/consumer/cn/doc/development/HMSCore-Guides/overview-0000001147638696

• Health Kit Overview：https://developer.huawei.com/consumer/cn/codelab/HMSHealthKit-StepCount/index.html#0

• Semi-finished code

GitHub - Thea-Feng/FedCGitHub

Semi-code: for easiness, just look at what is used by MainActivity.

Time	Task	Progress
During Semester (A)	Communication
During Semester (A)	Data Storage
During Semester (B)	Algorithm
During Semester (B)	UI
1st month (Summer) (A,B)	Health Kit
2nd month (Summer) (B)	Smart Watch
2nd month (Summer) (A,B)	Test App

Task

• Communication

• Data Storage

• Algorithm

• UI

• HealthKit & FL Platform

Communication

• Build a Apollo broker

  • Build a server to monitor message communication

  • To verify correctness, use Mqtt.fx / MqttX / EMqtt

    
• Fix the bug of MQTT subscribe

  • Publish function is implemented well. But there's a little bug in subscribe function: messageArrived is not called when subscribing message (in MainActivity.java)

  • You can also implement TCP version and forget about all the structure of MQTT if you are good at TCP

  

Data Storage

• Local storage: LitePal (finished)

• Cloud storage: Ali Cloud (supporting MQTT), setup a dashboard by the instruction in resources (to do)

UI (Easy task)

• Beautify

• English version for international student

• 

Algorithm

• Implement local differential privacy algorithm

Health Kit

• After obtaining the authentication, plug in health kit to let the app automatically collect data

• Smart watch plug-in: collect data from the watch