TeleTect

Getting Started

Before getting started, you must have the following on your machine.

• Frontend:
  • Node.js (https://nodejs.org/en)
  • Ionic CLI (https://ionicframework.com/docs/intro/cli)
  • Capacitor CLI (https://capacitorjs.com/docs/getting-started)
  • Android Studio (including Android SDK installation)
• Backend:
  • Python 3.8
  • Python modules declared on requirements.txt (pip install -r requirements.txt)
  • Java (some Python modules require JDK)
  • gcloud CLI (requires setting up application credentials)

Frontend

The frontend project includes two principal parts:

• Web-level project (Ionic UI-toolkit-based web application) [Root directory: /2t_frontend]
• Native-level project (Android Studio Project) [Root directory: /2t_frontend/android]

To deploy an Android application, you must first build the web-level project through Capacitor, then use Android Studio to build the native-level application (APK).

1. Please confirm the following before setting up the project.

   • Node.js
   • Ionic CLI
   • Capacitor CLI
   • Android Studio
   • Android SDK

2. Run the following commands to set up the web-level project.

   npm install

3. Set up the web-level project’s environment variable (/2t_frontend/.env) based on the template file (/2t_frontend/.env.template).

4. Run the following commands to build the web-level project and sync with the native-level project.

   npm run build
   npx cap sync

5. Open the Android Studio project using the following command. (Alternatively, you can open /2t_frontend/android on Android Studio)

   npx cap open android

6. Use Android Studio tools to run the frontend project as an Android application.

   • Documentation: https://developer.android.com/studio/run

Backend

1. Please confirm the following before setting up the project.

   • Python 3.8 (Note: As KoBERT uses the legacy versions of some modules, running the backend application on the environment beyond Python 3.8 may not work.)
   • Java (some Python modules require JDK) sudo apt install openjdk-11-jdk
   • gcloud CLI (requires setting up application credentials)

2. Run the following commands to set up the project.

   cd 2t_backend
   pip install -r requirements.txt

3. Set up application credentials using gcloud CLI.

   • Documentation: https://cloud.google.com/docs/authentication/gcloud

4. Proceed with the following to set up the voice phishing detection features.

   • KoBERT: Locate the trained model file (teletect_kobert_train.pt) to the following directory: /2t_backend/teletect/detection/kobert_detection/KoBERTModel/model. You can acquire the model file through the following methods:

     • Method 1: Run train_kobert_detection.py on the local machine.
     • Method 2: Download our pre-trained model file from the following URL and locate it in the model directory. https://link.geonhui.com/gdsc/sc2024/detection-model-file

   • Vertex AI/Gemini: No further actions are necessary if the application credentials setup is complete.

5. Set up the environment variable (.env) based on the template file (.env.template).

   • The environment variable file should be located in /2t_backend/.env

6. Run the backend server using the following command.

   uvicorn main:app --reload

Notes:

[Backend] Voice Phishing Detection (KoBERT)

• Voice phishing detection works faster on a GPU server that supports CUDA.
• To force the backend to run the detection on CPU mode, set the environment variable DETECTION_FORCE_CPU to true.