Differential Privacy

  1. Models of Differential Privacy Deployment
    1. The Central Model
      1. Architecture with Trusted Data Curator
        1. Centralized Noise Addition
          1. Lower Noise Requirements
            1. Use Cases and Assumptions
              1. Security and Trust Considerations
                1. Implementation Challenges
                2. The Local Model
                  1. Architecture with Untrusted Aggregators
                    1. Noise Addition at Data Source
                      1. Higher Noise Requirements
                        1. Privacy Amplification Techniques
                          1. Protocols and Systems
                            1. Google's RAPPOR
                              1. Protocol Overview
                                1. Bloom Filter Encoding
                                  1. Applications and Limitations
                                  2. Apple's Private Data Collection
                                    1. Protocol Overview
                                      1. Count Mean Sketch
                                        1. Applications and Limitations
                                        2. Microsoft's Telemetry Collection
                                        3. Trade-offs Between Local and Central Models
                                        4. The Shuffle Model
                                          1. Hybrid Approach Architecture
                                            1. Using Trusted Shuffler
                                              1. Privacy Amplification by Shuffling
                                                1. Implementation Considerations
                                                  1. Theoretical Guarantees
                                                    1. Practical Protocols
                                                    2. The Secure Aggregation Model
                                                      1. Cryptographic Protocols
                                                        1. Multi-Party Computation Integration
                                                          1. Privacy Amplification Properties
                                                          2. Federated Learning Models
                                                            1. Distributed Training with DP
                                                              1. Client-Server Architectures
                                                                1. Privacy Accounting Across Rounds

                                                              Previous

                                                              4. Fundamental Properties of Differential Privacy

                                                              Go to top

                                                              Next

                                                              6. Differential Privacy in Practice