Social Media - Private Set Intersection

Introduction

• We use a Privacy Enhancing Technology(PET), Private Set Intersection(PSI) to run a intersection check on two datasets and return the common values.
• Uses the Top 1000 youtuber list from Kaggle, which have been randomly split into two databases.
• In this project, we develop a POC that allows this phenomenon to run using Openmined's PSI library.

---

Project Structure

• This project contains four subdirectories:
  • Server: The server API that performs majority of the PSI computation
  • Client: The client API that retrieves client preferences and queries from the UI, creates a serializable/computable object for the server and returns intersection.
  • Web: The web directory contains source code for the client UI.
  • Data: The data subdirectory contains the intial dataset, that has been split into two subsets uding a 50% row selection probability. It also contains a MongoDB Population script.

---

Running the APIs

• Project contains a client and server setup [Both currently Node.js APIs]
• Prerequisites: Node.js, Python(For running scripts), React(Client Web) and MongoDB
• Once the prerequisites are installed, we run Taskfile to start our servers

Available Commands

$ ./Taskfile.sh                     # Starts Client, Server and UI concurrently
$ ./Taskfile.sh installAPI          # Installs the dependencies
$ ./Taskfile.sh startUI             # Starts Client UI

---

Step by Step Run

• Step 1: To run the client and server, we need to setup the environment variables. Make a copy of .env.example and name the copied file .env and fill in the respective details.
• Step 2: Install dependencies for both the client, server and the web UI.

$ ./Taskfile.sh installAPI
$ cd web && yarn add

• Step 3: Once the dependencies are installed, we can run the project.

$ ./Taskfile.sh

• Step 4: Navigate to localhost:3000 to access the web UI.

---

Contributors

• Swadhin Routray
• Kathryn Earles
• Wenhan Zhou