What is Privacy-Preserving Machine Learning?

Privacy-Preserving Machine Learning (PPML) refers to techniques that enable machine learning models to be trained without data leakage and exposure of sensitive data. It incorporates privacy-enhancing strategies that enable multiple input sources to collaboratively train machine learning models without exposing their private data in its original form. Common PPML methods include:

• Differential privacy: Adds noise to the data or outputs to prevent identification of individual data points.
• Federated learning: Allows models to be trained on decentralized data (on devices or different locations) without the data leaving its source.
• Homomorphic encryption: Enables computations on encrypted data without decrypting it, ensuring privacy during processing.