vFHE: Verifiable Fully Homomorphic Encryption with Blind Hash (2303.08886v1)

Abstract: Fully homomorphic encryption (FHE) is a powerful encryption technique that allows for computation to be performed on ciphertext without the need for decryption. FHE will thus enable privacy-preserving computation and a wide range of applications, such as secure cloud computing on sensitive medical and financial data, secure machine learning, etc. Prior research in FHE has largely concentrated on improving its speed, and great stride has been made. However, there has been a scarcity of research on addressing a major challenge of FHE computation: client-side data owners cannot verify the integrity of the calculations performed by the service and computation providers, hence cannot be assured of the correctness of computation results. This is particularly concerning when the service or computation provider may act in an untrustworthy, unreliable, or malicious manner and tampers the computational results. Prior work on ensuring FHE computational integrity has been non-universal or incurring too much overhead. We propose vFHE to add computational integrity to FHE without losing universality and without incurring high performance overheads.