Graph Neural Network Data Governance for Sensitive Information
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying and governing sensitive data at scale are inefficient, as they rely on manual processes that lose value in readability and are complex to manage, especially when data is accessed and used downstream from its initial point, leading to difficulties in controlling access and compliance with regulatory protections.
Innovation Solution
The use of machine learning models and graph neural networks to classify and govern sensitive data by generating nodes and edges in a graph based on governance policies and user interactions, predicting data flow, and refining the graph with new connections and nodes, allowing for automated data classification and access control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual classification methods are used for sensitive data, then classification accuracy can be maintained, but scalability and efficiency deteriorate at large data volumes
Solution Approach 1:
The patent replaces manual mechanical classification processes with automated machine learning models and graph neural networks. The system uses AI algorithms to automatically classify sensitive data, predict data flows, and generate governance actions, eliminating the need for manual review while maintaining high accuracy through sophisticated pattern recognition capabilities.
Solution Approach 2:
The system enables self-service classification where the machine learning models autonomously process data, generate classifications, and update governance policies without human intervention. The graph neural networks automatically learn from data patterns and improve classification accuracy over time through self-training on observed data flows and user interactions.
2Reliability
If mechanical masking is applied to classified sensitive data, then data protection is improved, but data readability and value are lost
Solution Approach 1:
Instead of applying uniform masking to all sensitive data, the system applies different governance actions based on local context. The graph neural networks analyze specific data elements, their relationships, and access patterns to determine appropriate actions for each data point, preserving readability where protection is less critical while applying strong protection where needed.
Solution Approach 2:
The system implements dynamic governance actions that adapt based on real-time analysis. Rather than static masking, the system generates context-aware actions such as access restrictions, approval workflows, or selective masking that change based on user role, data sensitivity, and usage context, maintaining data utility while ensuring protection.
3Adaptability or versatility
If data is accessed and used downstream from initial access points, then data utility is improved, but governance complexity and access control difficulty increase
Solution Approach 1:
The graph neural network framework provides a universal governance platform that handles multiple functions: tracking data flows, predicting unauthorized access, generating governance actions, and updating policies. This single system manages governance across the entire data lifecycle from source to downstream usage, reducing overall system complexity despite increased data mobility.
Solution Approach 2:
The system implements continuous feedback loops where graph neural networks monitor actual data flows, compare them against predicted patterns, and automatically update governance policies. This closed-loop approach adapts to downstream data usage patterns, maintaining effective control while accommodating legitimate data utility needs through learned behavioral patterns.
4Productivity
If automated machine learning models are deployed for data classification, then processing efficiency is improved, but system complexity increases
Solution Approach 1:
The system segments the complex governance task into distinct components handled by specialized machine learning models: one model for data classification, another for flow prediction, and a third for action generation. The graph neural network framework organizes these segmented functions into a coherent system, managing complexity through modular architecture while maintaining high processing efficiency.
Data Source
AI summary
Systems and methods for data classification and governance are disclosed. In accordance with aspects, a method may include retrieving data related to a first-level entity, wherein the first-level entity is associated with a governance policy; using one or more machine learning models, a method may include: adding labels to the data; adding classifications to the data based on the labels; generating nodes and edges in a graph based on the governance policy and known user interactions with the data; resolving ambiguous nodes into specific nodes; predicting edge probabilities in the graph; and predict a flow of data in a network based on the edge probabilities.