Asymmetric Encryption for Secure Data Pipeline Model Development
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional digital rights management technologies for securing sensitive data are complex, expensive, and difficult to maintain, leading to increased infrastructure processing and storage resource usage, especially in scenarios with multiple users accessing sensitive data for analytics model development and use.
Innovation Solution
A system and method utilizing asymmetric encryption to secure data pipelines, integrating security and model development controllers, which obfuscate and decrypt data using public and private key pairs, reducing the need for strict access control and minimizing exposure of sensitive data in plain text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional digital rights management technologies are used to secure sensitive data, then data security is improved, but infrastructure processing and storage resource usage increases
Solution Approach 1:
The patent extracts the sensitive data from the plaintext domain by applying encryption transformations, separating the security-critical information from the processing pipeline. Only encrypted representations are stored and processed, with original data never exposed in plaintext during model development, thereby reducing security risks without proportionally increasing resource usage
Solution Approach 2:
The patent introduces encrypted data as an intermediary between the original sensitive data and the model processing system. This intermediary layer allows model training and analysis to proceed on transformed data that preserves statistical properties while eliminating direct exposure to sensitive information, reducing both security threats and the need for complex access control infrastructure
2Reliability
If traditional digital rights management technologies are used to secure sensitive data, then data security is improved, but system complexity increases
Solution Approach 1:
The patent transforms the security approach by changing the state of data from plaintext to encrypted format through mathematical transformations. This parameter change (from readable to encrypted state) maintains security while simplifying the system architecture, as the encryption/decryption process is more straightforward than traditional digital rights management frameworks requiring complex access control lists, permission hierarchies, and audit trails
3Reliability
If strict access control is implemented for sensitive data, then data security is improved, but model development progress is impeded
Solution Approach 1:
The patent creates encrypted copies of the sensitive data that preserve the statistical and relational properties needed for model training while eliminating the security risks of plaintext exposure. Data scientists can work with these encrypted copies using standard tools and techniques, maintaining productivity while security is ensured through the encryption layer. The encrypted copies function as surrogates that enable development work without requiring access to original sensitive data
Data Source
AI summary
A method of securing a data set using encryption and scaling. The data set comprises a modeling and validation section with each section having text and numeric data. For each section, the method encrypts text data using a public key and scales numeric data using a scaling factor. The method builds a model by applying the encrypted modeling section to the algorithm to generate modeling text data, numeric data, and patterns derived therefrom. The method generates validation text data, numeric data, and patterns derived therefrom by applying the encrypted validation section to the model. The method compares the patterns from each section and validates the model based on the comparison. The method decrypts the generated modeling text data and validation text data using a private key and descales the modeling numeric data and validation numeric data using the scaling factor. The method verifies the model by comparing the decrypted text data and descaled numeric data to the same in the data set.