IMPROVED CONFIDENTIAL COMPUTING

Smart overview of the Invention

An innovative scheme for confidential computing integrates multi-party computation (MPC) and fully homomorphic encryption (FHE). This approach ensures that private information remains secure, as no data leaks occur unless the foundational assumptions of both MPC and FHE are compromised. The system aims to perform computations over private data without exposing any sensitive information to unauthorized parties.

Background of Confidential Computing

Confidential computing enables multiple parties to compute functions over private inputs while ensuring that sensitive data remains confidential. The challenge lies in transferring the computed result without revealing any information about the individual private inputs. Current methods, such as MPC and FHE, rely on specific assumptions regarding the honesty of participants or the security of mathematical parameters, respectively.

Multi-Party Computation (MPC)

MPC operates under the assumption that a limited number of computation parties can be corrupt, while the majority remain honest. This guarantees that any colluding subset cannot access private inputs, allowing the output party only to learn the final result. The process involves each input party sending messages to computation parties, which then collaboratively compute the desired function while maintaining input privacy.

Fully Homomorphic Encryption (FHE)

FHE allows computations to be performed on encrypted data without needing to decrypt it first. This method relies on secure algorithms that generate keys for encryption and decryption, ensuring that the output remains confidential. The process involves generating keys, encrypting input data, performing computations on ciphertexts, and finally decrypting the results to reveal the output without exposing any private inputs during the computation.

Hybrid Approach: Combining MPC and FHE

The hybrid MPC/FHE scheme enhances security by addressing vulnerabilities present in using either method alone. By combining these approaches, it leverages the strengths of both frameworks while mitigating risks associated with incorrect assumptions or configurations. In this hybrid model, the function input for MPC is substituted with FHE's evaluation algorithm, creating a robust solution for confidential computing that maintains strict privacy standards.