MNIST Classifier

• 

  Demo App

Inspiration 🌟

Our inspiration comes from the growing need for privacy in the digital age. With machine learning becoming increasingly pervasive, we realized the importance of protecting sensitive data. This led us to create a solution that not only leverages the power of ML but also ensures the utmost privacy.

What it does 🤖

The MNIST Classifier app utilizes the Leo language on the Aleo blockchain to classify handwritten digits. It uniquely applies zero-knowledge proofs, ensuring the privacy of both input data and classification results. This approach offers a new standard in privacy-preserving machine learning.

How we built it 🔨

We built the app using the Leo programming language, specifically designed for the Aleo blockchain. We focused on integrating zero-knowledge proofs to secure the data processing, ensuring that the input and output remain private, without compromising the accuracy and efficiency of the classifier.

Challenges we ran into 🚧

One of the biggest challenges was implementing zero-knowledge proofs in a way that is both efficient and secure. We also faced hurdles in optimizing the app for the Aleo blockchain, ensuring that it operates seamlessly while maintaining user privacy. Also the neural network in Leo code has a lot of constraints. Therefore deploying to the Aleo testnet has some issues. I have encounter 503 errors while deploying using snarkos. Also Aleo.tools can not deploy the code since it has a lot of constraints.

Accomplishments that we're proud of 🏆

We're proud of successfully creating a fully private MNIST classifier that stands as a testament to innovation in privacy-preserving machine learning. Our use of the Leo language and Aleo blockchain to implement zero-knowledge proofs is a significant accomplishment in this field.

What we learned 📚

Throughout this project, we've gained profound insights into the complexities of blockchain technology, the intricacies of the Leo programming language, and the powerful potential of zero-knowledge proofs in privacy preservation. I have in the Aleo workshops, I have learn a lot in about Leo programming as well

What's next for MNIST Classifier 🔮

Moving forward, we plan to enhance the scalability and efficiency of our app. We aim to expand its capabilities to classify more complex neural network while continuing to uphold the highest standards of privacy and security. To do that we need a ** some math library** in Leo. This library will simplify the process to code neural network in Leo file.

Built With

• aleo
• keras
• leo
• node.js
• python
• react
• tensorflow