Machine-Learning Data Encryption for Selective, Coherent De-Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional encryption and de-identification techniques fail to comprehensively remove sensitive information, often rendering digital documents unusable, and struggle with context-dependent information, leading to retrieval errors and imperfect document coherence.
Innovation Solution
A parallel model pipeline combining machine learning components and encryption techniques to detect and encrypt protected data elements, replacing them with anonymized tags while preserving document coherence, enabling secure storage and selective repopulation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional de-identification techniques remove sensitive information, then security is improved, but document usability deteriorates due to loss of non-sensitive information
Solution Approach 1:
The patent segments the document into sensitive and non-sensitive portions by detecting protected data elements and applying encryption only to those segments, leaving the rest of the document plaintext and usable. This selective encryption approach maintains document usability while protecting sensitive information.
Solution Approach 2:
The patent applies different quality treatments to different parts of the document: encrypted protection is applied locally to sensitive data elements, while non-sensitive portions remain in their original plaintext form, preserving their usability and readability.
2Productivity
If pattern matching or keyword lists are used to detect sensitive information, then detection speed is improved, but detection accuracy deteriorates due to inability to capture context-dependent information
Solution Approach 1:
The patent introduces an intermediary layer of encrypted element representations that bridge the gap between fast pattern matching and accurate context understanding. These representations allow the system to use efficient keyword-based detection while maintaining the ability to handle context-dependent information through the encryption/decryption mechanism.
3Reliability
If sensitive information is redacted, then security is improved, but retrieval accuracy deteriorates since sensitive information may have been used for file retrieval
Solution Approach 1:
The patent creates encrypted copies of sensitive information that can be used for retrieval purposes without exposing the actual sensitive data. The encrypted element representations serve as functional copies that maintain retrieval capability while ensuring security.
4Loss of information
If advanced natural language processing is used to maintain document coherence, then document usability is improved, but detection complexity increases
Solution Approach 1:
The patent segments the complex NLP task into simpler sub-tasks: detecting protected data elements through pattern matching, determining their sensitivity, and applying encryption. This segmentation reduces detection complexity while maintaining document coherence through the structured encryption approach.
Data Source
AI summary
Various embodiments of the present disclosure provide automated encryption and data de-identification techniques for improving computer security. The techniques apply machine learning and encryption techniques to transform input data objects to tagged data objects that may be locally decrypted using encrypted element representation stored within the tagged data objects. The techniques may include determining a protected data element from an input data object based on privacy criteria and generating the tagged data object from the input data object by replacing the protected data element with an anonymized privacy tag that identifies a privacy type of the protected data element. The techniques may further include generating an encrypted element representation of the protected data element and inserting the encrypted element representation to a portion of the tagged data object to enable decryption of the tagged data object by authorized entities.