A subfield of cryptography, Secure Multi-Party Computation (MPC) enables multiple parties to compute a function while keeping their input data private. The goal is to create methods for parties to jointly compute a function over their inputs while keeping those inputs private. Unlike traditional cryptography, which secures communication or storage against external adversaries, MPC protects participants' privacy from each other.
Foundation and Evolution
The foundation for secure multi-party computation began in the late 1970s with the work on mental poker. This cryptographic work simulated game playing and computational tasks over distances without requiring a trusted third party. Traditionally, cryptography focused on concealing content, while MPC introduced a new type of cryptography aimed at protecting privacy among participants.
Key Characteristics
- Data Privacy: The data remains encrypted and distributed across different parties during the computation.
- No Single Decryption: No single party can decrypt or view the raw data inputs from other parties.
- Computed Output: The final computed output is the same as if the parties pooled their unencrypted data centrally.
Applications
- Secure Business Analytics: Provides meaningful results from data analytics without exposing underlying sensitive data.
- Privacy-Preserving Machine Learning: Enables machine learning models to be trained on sensitive data without revealing the raw data.
- Fraud Detection: Allows institutions to collaborate on fraud detection without sharing raw data.
- SaaS Providers: Offers a strong technical control for handling sensitive personally identifiable information (PII).
Technical Techniques
MPC uses sophisticated cryptographic techniques such as:
- Secret Sharing: Divides data into parts and distributes them among multiple parties.
- Oblivious Transfer: Allows a party to transfer information to another party without revealing anything beyond the transferred data.
- Garbled Circuits: Enables secure computation of a function by transforming it into an encrypted circuit.
MPC enables collaborative data sharing and processing in an encrypted and privacy-preserving manner, unlocking insights without compromising data confidentiality.
