Database Entity Sensitivity Classification via Schema Tokenization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing database entity sensitivity classification methods face challenges in large and complex systems, including security, reliability, and efficiency issues, particularly in identifying sensitive information without exposing confidential data and requiring human intervention.
Innovation Solution
The method generates database entity sensitivity classifications based on entity schema, using tokenization and corpus-based frequency values to create feature data, and employs a sensitivity classification model selected through expert feedback, enabling secure and accurate classification without analyzing underlying data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional database entity sensitivity classification methods are used, then classification can be performed, but security is compromised due to potential exposure of confidential data and requirement for human intervention
Solution Approach 1:
The patent extracts only the necessary metadata (table names, column names, data types) from the database schema without accessing or exposing the actual confidential data values. The classification model is trained on this extracted metadata to determine sensitivity, thereby achieving accurate classification while preventing data exposure.
Solution Approach 2:
The patent introduces a sensitivity classification model as an intermediary between the database schema and the classification output. This model processes metadata and generates sensitivity classifications without requiring human intervention or direct access to confidential data, thereby maintaining security while improving reliability.
2Measurement precision
If comprehensive data analysis is performed for sensitivity classification, then classification accuracy improves, but processing time and computational resources increase significantly
Solution Approach 1:
The patent extracts only essential metadata features (table names, column names, data types) from the database schema, avoiding the need to analyze entire datasets. This selective extraction maintains classification accuracy while dramatically reducing processing time and computational resource requirements.
Solution Approach 2:
The patent performs preliminary extraction and preparation of metadata features before classification. By pre-processing the schema information to create relevant features, the system enables faster classification processing without sacrificing accuracy, as the essential information is already organized and ready for analysis.
3Reliability
If manual review and human intervention are used for sensitivity classification, then classification reliability improves, but productivity and efficiency decrease
Solution Approach 1:
The patent implements a self-service classification system where the sensitivity classification model automatically processes metadata and generates classifications without human intervention. The model learns from training data and independently determines sensitivity levels, maintaining reliability through consistent application of learned patterns while dramatically improving productivity.
Solution Approach 2:
The patent incorporates feedback mechanisms where the classification model is trained on labeled data and continuously improves its accuracy. The model learns from correct and incorrect classifications, adjusting its parameters to enhance reliability over time while maintaining automated high-throughput processing capability.
Data Source
AI summary
There is a need for solutions that perform entity sensitivity classification for a database entity associated with plurality of database elements. This need can be addressed by, for example, for each database element, generating element tokens for the database element based on a textual identifier for the database element, generating token combinations for the database element based on the element tokens for the database element and a token order associated with the textual identifier, generating a token-combination document for the database element based on the token combinations for the database element, and generating element feature data for the database element based on the token-combination document for the database element, and determining the entity sensitivity classification on each element feature data associated with a database element of the plurality of database elements and using a sensitivity classification model.