Privacy-on-Federated-Learning

Comparative Analysis of Data Anonymization Techniques on Federated Learning

Techniques used:

1. Differential Privacy (Dependancies: IBM's diffprivlib library)
2. Fully Homomorphic Encryption (Dependancies: TFHE (Fast Fully Homomorphic Encryption over the Torus) open source library)
3. Secure Multi-Party Computation (Dependancies: Python's PySyft Library)

Fedearated Learning implemented using TensorFlow's toolkit.

Dataset used for analysis: https://archive.ics.uci.edu/ml/datasets/adult