Stochastic Secure Computing for Bootstrapping-Free Homomorphic Encryption

The patent changes the fundamental parameter of how homomorphic operations are performed by switching from traditional bootstrapping-based fully homomorphic encryption to a stochastic computing approach. This involves representing encrypted data in a stochastic format where arithmetic operations can be performed directly on the stochastic representations without triggering noise accumulation that requires bootstrapping, thereby maintaining unlimited operation capability while dramatically reducing computation time

The patent replaces the mechanical bootstrapping process (which involves complex decryption and re-encryption operations) with a stochastic computing mechanism. Instead of periodically resetting noise through bootstrapping, the system uses probabilistic bit string operations that inherently avoid noise accumulation, substituting a fundamentally different computational mechanism that achieves the same goal of enabling unlimited operations without the time penalty

The patent fundamentally changes the operational parameters by adopting stochastic computing representations where encrypted values are expressed as probabilistic bit strings. This parameter change allows addition and multiplication to be performed through simple bitwise operations rather than complex homomorphic transformations, reducing the calculation amount while preserving the ability to perform unlimited operations

The patent substitutes the traditional homomorphic encryption mechanical system (which requires bootstrapping to manage noise) with a stochastic computing system. This substitution eliminates the need for bootstrapping operations entirely, as the stochastic representation naturally handles unlimited operations without noise accumulation, thereby significantly reducing the overall calculation amount

The patent extracts and eliminates the bootstrapping step from the homomorphic encryption process by adopting stochastic computing. Instead of performing bootstrapping to reduce noise after a certain number of operations, the system uses stochastic representations that allow operations to continue indefinitely without noise accumulation, effectively taking out the time-consuming bootstrapping operation from the computational workflow

The patent changes the fundamental parameter of noise management by switching from deterministic noise reduction through bootstrapping to probabilistic noise avoidance through stochastic computing. This parameter change allows the system to perform an unlimited number of operations without the time penalty of periodic bootstrapping, as the stochastic representation inherently prevents noise from exceeding threshold values