Secure Multi-Party Computation (SMPC)

The fundamental guarantee that no party learns anything about another party's private input beyond what can be logically inferred from the final output. This is achieved through secret sharing schemes where data is split into mathematically random shares that reveal no information in isolation. For example, in a three-party salary comparison, each participant's actual salary remains hidden; only the result of the comparison is revealed. This property holds even if a subset of parties is actively malicious.

The protocol ensures that the computed output is mathematically identical to what would have been obtained if a trusted third party had performed the computation on the plaintext inputs. This is enforced through verifiable secret sharing and message authentication codes embedded in the computation circuit. Even if some parties deviate from the protocol or submit malformed data, the honest majority can detect the fault and abort, preventing a false result from being accepted.

Fairness ensures that if any party receives the output, all designated parties receive it. This prevents a malicious participant from aborting the protocol after learning the result but before others do. In practice, this is achieved through a commitment and reveal phase where the final output shares are exchanged simultaneously. If a party aborts early, the remaining honest parties can reconstruct the output from their shares without the malicious actor.

A stronger property than fairness, Guaranteed Output Delivery (GOD) ensures that the protocol always terminates with a correct output regardless of adversarial behavior. This requires an honest majority assumption and typically uses Byzantine agreement protocols as a sub-routine. In the context of private synthetic data factories, GOD means that a scheduled joint computation across isolated data silos will always complete successfully, preventing denial-of-service attacks on the synthesis pipeline.

Protocols are classified by their adversarial model. Semi-honest security assumes parties follow the protocol but try to learn extra information from the transcript. Malicious security protects against parties that arbitrarily deviate from the protocol—injecting false data, aborting early, or sending corrupted messages. Maliciously secure SMPC uses zero-knowledge proofs and cut-and-choose techniques to force correct behavior, essential for high-stakes financial or healthcare computations.

Universal composability guarantees that an SMPC protocol remains secure even when executed concurrently with other protocols or as a sub-routine in a larger system. This is critical for complex private synthetic data factories where multiple SMPC instances may run in parallel—one for feature selection, another for aggregation. Without composability, an attacker could exploit interactions between protocol instances to leak information across sessions.

An encryption scheme permitting computation directly on ciphertexts. The generated encrypted result, when decrypted, matches the outcome of operations performed on the plaintext. Unlike SMPC, which requires interactive communication between parties, fully homomorphic encryption (FHE) allows a single party to compute on encrypted data without interaction.

• Key Distinction: Non-interactive computation on encrypted data vs. interactive joint computation in SMPC.
• Use Case: Outsourcing computation to an untrusted cloud while maintaining data confidentiality.
• Limitation: High computational overhead compared to SMPC for complex functions.

A mathematical framework that injects calibrated statistical noise into datasets or queries to guarantee that the presence or absence of any single individual's record is indistinguishable. While SMPC protects inputs during computation, differential privacy protects the output from revealing individual contributions.

• Epsilon Parameter: Quantifies the privacy loss budget; lower values enforce stronger guarantees.
• Composition: Often combined with SMPC to ensure that the final aggregated result does not leak individual data through the output itself.
• Mechanism: Gaussian or Laplacian noise addition calibrated to the sensitivity of the query function.

A fundamental cryptographic primitive where a sender transfers one of potentially many pieces of information to a receiver, but remains oblivious to which piece was transferred. Oblivious Transfer is a core building block for many SMPC protocols.

• 1-out-of-2 OT: The sender inputs two messages; the receiver chooses one without learning the other, and the sender learns nothing about the choice.
• OT Extension: Techniques to efficiently extend a small number of base OTs into many OTs using only symmetric-key operations.
• Role in SMPC: Enables secure evaluation of garbled circuits for boolean functions.

A cryptographic protocol where a function is represented as a boolean circuit, and one party garbles the circuit by encrypting the truth table of each gate. The other party evaluates the garbled circuit obliviously using keys obtained via Oblivious Transfer.

• Constant-Round Protocol: Requires a fixed number of communication rounds regardless of circuit depth.
• Yao's Protocol: The foundational garbled circuit construction for secure two-party computation.
• Optimizations: Free-XOR and half-gates techniques dramatically reduce computational and communication costs.

A method for distributing a secret among a group of participants, each receiving a share of the secret. The secret can only be reconstructed when a sufficient number of shares are combined. SMPC protocols leverage secret sharing to perform linear operations on distributed data without reconstruction.

• Shamir's Secret Sharing: Based on polynomial interpolation; requires t+1 shares to reconstruct from an n-party distribution.
• Additive Secret Sharing: Splits a value into random shares that sum to the original; enables linear operations with zero communication.
• Threshold Schemes: Define the minimum number of parties required for reconstruction, balancing availability and confidentiality.