DataTerminology.comNEW
Also known as: MPC, SMPC
Cryptographic protocols enabling parties to compute a function over their inputs while keeping each input private from other parties.
DESCRIPTION
Secure multi-party computation (MPC) allows collaborative computation without revealing individual inputs, reducing trust requirements. Techniques include secret sharing, garbled circuits, and oblivious transfer, each offering different performance/security trade-offs. MPC underpins privacy-preserving analytics, ad measurement, fraud detection, and joint model training. Compared to using a trusted third party, MPC distributes trust and reduces single points of failure. Performance has improved with hardware acceleration and protocol optimization, though costs remain higher than plaintext computation.
ALTERNATIVE DEFINITIONS
Wikipedia
Secure multi-party computation (MPC) is a subfield of cryptography with the goal of creating methods for parties to jointly compute a function over their inputs while keeping those inputs private.
Canonical definition emphasizing privacy of inputs during joint computation.
NIST
Multiparty computation enables parties to jointly compute a function over their inputs while keeping those inputs private, supporting privacy-preserving data analysis.
Positions MPC within privacy-enhancing cryptography for data analysis.
MPC Alliance
MPC is a cryptographic technique that allows multiple parties to compute on data without revealing the data to each other.
Plain-language explanation oriented to industry adopters.
RELATED TERMS
Secret Sharing Garbled Circuits Oblivious Transfer Zero-Knowledge Proof Private Set Intersection Federated Learning Threshold Cryptography Differential Privacy Privacy-Preserving Machine Learning
COMPETING TERMS
Trusted Third Party Homomorphic Encryption Trusted Execution Environment
KNOWLEDGE AREAS
Data Security Data Privacy & Compliance AI Data Science
TARGET AUDIENCE
Software Developer AI Engineer Data Scientist Data Engineer MLOps Engineer Cybersecurity Engineer Chief Information Security Officer Data Privacy Officer
EXTERNAL SOURCES
- Wikipedia - Secure multi-party computation
- NIST - NIST overview of privacy-enhancing cryptography
- MPC Alliance - Industry explainer and use cases for MPC
- MP-SPDZ - Open-source MPC framework implementation
TOP QUESTIONS
What problems does Secure Multi-Party Computation solve?
MPC enables collaborative analytics and model training without sharing raw data, addressing privacy, regulatory, and competitive constraints.
Category: Relevancy
How does MPC compare to using a trusted third party?
MPC spreads trust among participants and avoids single points of compromise, albeit with higher computational overhead.
Category: Comparison
What are primary MPC paradigms?
Secret sharing-based protocols and garbled circuits dominate, often combined with oblivious transfer and precomputation.
Category: Concepts
What are common performance bottlenecks in MPC?
Communication latency, cryptographic operations, and precomputation dominate costs. Protocol choices and batching mitigate overhead.
Category: Performance
Where is MPC used in practice today?
Use cases include ad conversion measurement, fraud detection, joint KYC, and privacy-preserving model training and inference.
Category: Application
How does MPC relate to federated learning?
Federated learning distributes training; MPC protects computations on sensitive aggregates. They can be combined for strong privacy.
Category: Concepts
What security assumptions underpin MPC?
Models include semi-honest and malicious adversaries, with different guarantees and protocol costs.
Category: Requirements
How do I start building with MPC?
Prototype with open-source frameworks, define threat models, benchmark operations, and start with small, well-scoped functions.
Category: Learning
What’s a hot topic in MPC research?
Hardware acceleration, efficient malicious-secure protocols, and combining MPC with zero-knowledge proofs are active areas.
Category: Topic