Homomorphic Data Analysis for Secure Cloud Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Businesses using cloud computing face security vulnerabilities as they lack control over their data, making them susceptible to attacks and legal issues when using cloud services for data analysis.
Innovation Solution
A novel encryption scheme that allows arithmetic operations on encrypted data using a public computation key, while keeping the data secure with a private encryption and decryption key, ensuring that the data remains encrypted throughout the analysis process, even on a potentially compromised server.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is encrypted to maintain security, then data security is improved, but the ability to perform data analysis is worsened
Solution Approach 1:
The patent transforms data from plaintext to ciphertext form, changing its state while preserving mathematical properties that allow homomorphic operations. This parameter change enables both security (encrypted form) and analyzability (mathematical operations on ciphertext)
Solution Approach 2:
Homomorphic encryption acts as an intermediary that allows cloud servers to process encrypted data without decrypting it. The encryption scheme mediates between the need for security and the need for computation by enabling operations on encrypted data
2Power
If data is stored in cloud computing resources, then computational power and availability are improved, but security control is worsened
Solution Approach 1:
Homomorphic encryption serves as an intermediary layer between the client's data and the cloud server's processing capabilities. The server operates on encrypted data through the encryption scheme, which mediates the interaction and ensures security while enabling computation
Solution Approach 2:
The data is transformed into an encrypted state that preserves computational properties. This parameter change from plaintext to ciphertext allows the data to maintain its utility for computation while being protected during storage and processing in the cloud
3Productivity
If data is decrypted to enable analysis operations, then data analyzability is improved, but security vulnerability is worsened
Solution Approach 1:
Instead of changing data from encrypted to decrypted form, the patent enables operations directly on the encrypted data through homomorphic properties. This parameter approach allows analyzability without the security vulnerability of decryption
Data Source
AI summary
Systems, methods, and computer-executable instructions for homomorphic data analysis. Encrypted data is received, from a remote system, that has been encrypted with an encryption key. A number of iterations to iterate over the encrypted data is determined. A model is iterated over by the number of iterations to create an intermediate model. Each iteration updates the model, and the model and the intermediate model encrypted with the encryption key. The intermediate model is provided to the remote system. An updated model based upon the intermediate model is received from the remote system. The updated model is iterated over until a predetermined precision is reached to create a final model. The final model is provided to the remote system. The final model is encrypted with the encryption key.