d/acc homomorphic

We built infrastructure for training biometric models in a federated way while maintaining data privacy by encrypting the model and user data.

User data is stored locally, owners can register and administrate access for data scientists in a database, and can receive and approve code requests to be run on their data in order to earn money.

We combine PySyft with TenSEAL, a library for federated training and one for homomorphic encryption, respectively.

One of the purposes of the project is to increase accessibility of disabled individuals to earn from watching or interacting with content while wearing their brain computer interfaces. This projects allows them to keep ownership of their data and to earn an income

For login, we used Dynamic for abstracting accounts. PySyft enabled federated training, allowing users to create Datasite structures to securely store their data, register data scientists, and manage encrypted code requests. TenSEAL was used to implement homomorphic encryption, enabling computations directly on encrypted data without compromising privacy.