FHE Computation Fingerprints for Verifiable Delegated Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for verifying delegated computation tasks in cloud computing environments, particularly with Fully Homomorphic Encryption (FHE), fail to effectively detect deviations from requested computations, especially when dealing with malicious servers that can produce consistent wrong results without access to the public key, and lack efficient mechanisms to ensure correct execution with minimal overhead.
Innovation Solution
A system and method that incorporates computation fingerprints, calculated a-priori using FHE, are embedded within input data to detect deviations by comparing the a-priori fingerprint results with the computed fingerprint results, ensuring correct execution by restricting server operations to black box additions and multiplications, and using discrete logarithmic representations to verify computations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If computation delegation is used in cloud computing to mask tasks, then privacy protection is improved, but the ability to verify computation correctness is worsened
Solution Approach 1:
The patent introduces computation fingerprints as an intermediary mechanism that bridges the gap between encrypted computation and verification. The fingerprint acts as a mediator that can be computed on encrypted data while still providing verifiable proof of correct execution, thus maintaining both privacy protection and verification capability
Solution Approach 2:
The patent segments the computation verification process into two independent parts: the main computation task and the fingerprint verification. This segmentation allows the fingerprint to be verified separately without decrypting or revealing the actual computation inputs, thus maintaining privacy while enabling verification
2Productivity
If zero-knowledge proofs like SNARK are used for computation verification, then verification efficiency is improved, but the complexity of creating verification is worsened
Solution Approach 1:
Instead of requiring full zero-knowledge proofs for complete computation verification, the patent uses computation fingerprints that provide partial verification through polynomial evaluation. This partial action approach achieves sufficient verification efficiency without the excessive complexity of complete SNARK implementation
Solution Approach 2:
The patent changes the verification parameter from complex cryptographic proofs to simpler polynomial evaluation problems. By transforming the verification task into evaluating whether a polynomial equals zero at a specific point, the system achieves efficient verification with reduced computational complexity
3Loss of information
If FHE schemes are used to cope with honest but curious servers, then privacy protection is improved, but the ability to detect malicious servers is worsened
Solution Approach 1:
The patent implements a feedback mechanism where computation fingerprints provide verifiable proof of correct execution. This feedback loop allows the system to detect when a server deviates from the requested computation, thus enabling detection of malicious behavior while maintaining privacy protection through encrypted computation
Solution Approach 2:
The patent performs preliminary computation of expected fingerprints before delegating the actual computation task. This preliminary action creates a reference value that can be used to verify the server's execution, enabling detection of malicious servers while the actual computation remains encrypted and private
4Reliability
If computation verification methods are implemented, then computation correctness is improved, but the overhead of verification is worsened
Solution Approach 1:
The patent performs preliminary computation of fingerprints and expected results before the actual delegated computation. This preliminary action allows the verification to be done efficiently by simply comparing the returned fingerprint against the pre-computed expected value, minimizing verification overhead while ensuring correctness
Data Source
AI summary
A system for verifying the execution of requested computation tasks, delegated by a computerized delegator device, to one or more computerized devices executing the tasks, comprising one or more computerized devices that contain one or more processors being adapted to define an input data for each delegated computation task; add to the input data, one or more computation fingerprints being executable encrypted input control bits, capable of detecting deviation from each requested computation task; allow the delegator to calculate the computation fingerprints once, by executing the requested computation tasks on predetermined random base values, to obtain an a-priori fingerprint result; couple the base values to each requested computation task; allow the one or more computerized devices to execute each requested computation task along with the base values, for returning an output consisting of a combination of a computed task result section and a calculated fingerprint result section; verify the returned output by comparing the a-priori fingerprint result to the calculated fingerprint and accepting the returned output if and only if the returned fingerprint result was identical to the a-priori fingerprint result; and use the a-priori fingerprint result or new verified result to verify further requested computation tasks delegated to the same one or more computerized devices.