Spatial dimensionality reduction-based secure two-party vector element-wise multiplication method and related apparatus

The patent transforms the problem from vector element-wise multiplication directly in one dimension to matrix multiplication in two dimensions, then back to vector form. By converting vectors to matrices, performing secure multiplication, and then extracting diagonal elements, the patent achieves secure computation with reduced complexity compared to direct garbled circuit approaches.

The patent replaces the mechanical construction of garbled circuits with a mathematical transformation approach. Instead of building Boolean circuits for each multiplication operation, the patent uses matrix multiplication properties and dimensionality reduction to achieve the same security goal with significantly lower computational overhead.

The patent combines multiple secret sharing operations into a single matrix multiplication operation. By merging the vector operations into matrix form, the patent reduces the number of separate message exchanges required, as matrix multiplication can be performed more efficiently with fewer communication rounds.

The patent extracts the secure computation functionality from the third-party cloud platform and implements it locally using matrix multiplication. By taking out the reliance on external platforms and using local cryptographic operations with dimensionality reduction, the patent significantly reduces both computing storage overhead and communication overheads while maintaining security.