Team Name: MINI-DORAS
Problem ID: 1672
Team Leader: Naman Bansal @Nb4159
Team Members:
-
MEMBER_1 - 2022UIN3312 Naman Bansal - @Nb4159
-
MEMBER_2 - 2022UIN3341 Arunima Banerjee- @elisedare28
-
MEMBER_3 - 2022UIN3371 Hemant Kumar- @HemantKr071
-
MEMBER_4 - 2022UIN3356 Pranav Varshney- @pranav2310
-
MEMBER_5 - 2022UIN3339 Prasann- @Prasann2004
-
MEMBER_6 - 2022UIN3330 Ritobroto Mukherjee- @RMZeroFour
- Final SIH Presentation: Final SIH Presentation
- Video Demonstration: Watch Video
- Source Code: GitHub Repository
The Unique Identification Authority of India (UIDAI) manages crucial Aadhaar-related operations and uses CAPTCHA to protect its portals, but it has drawbacks. CAPTCHA can be time-consuming and frustrating for users and poses accessibility challenges for people with disabilities.
This solution must follow strict privacy policies, collecting minimal, anonymized data and ensuring compliance with legal and ethical standards.
The Solution follows a 3 Layer approach:
- Layer 1(ML Judge): Uses a ML model to classify the user as a bot or Human. If the model has low Confidence it will move to layer 2.
- Layer 2(Simple Task + Sus Meter): Displays 9 problems and an image of number displaying which problem to solve. This image would be generated using SOTA addversal attack methods making it very difficult for the bot to identify the image. If the bot is still able to solve the problem there will be a model known as Sus Meter that will help us identifying if the bot has solved the problem. The problems will be chosen which can be solved by majority indians deduced by a Survey.
- Layer 3(Normal Captcha): If the model is still unsure we will just use Normal Captcha.
- React - react.dev
- Vite JS - vitejs.dev
- PyTorch - pytorch.org
- FastAPI - fastapi.tiangolo.com
- TensorFlow - tensorflow.org
- CatBoost - catboost.ai
- XGBoost - xgboost.readthedocs.io
This project requires the following npm packages:
- React.js: Frontend library for building user interfaces.
- react-router-dom: For dynamic routing.
- Vite: Build tool for faster development.
git clone https://github.com/RMZeroFour/DCAPTCHA/tree/main
- Navigate to the back directory
cd back
- Install requirements
pip install -r requirements.txt
- Run Backend
python main.py
- Navigate to the vite-project folder in the front directory
cd vite-project
- Run the website
npm run dev
- Navigate to the bot directory
cd bot
- Install required dependencies
pip install -r requirements
- Run bot script
python bot.py
Distributed under the MIT License. See LICENSE.txt
for more information.