LLM Tagging for Consistent Data Authorization Policies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data authorization systems in Zero Trust Architecture (ZTA) face challenges in deriving scalable and efficient arbitrary policies due to reliance on user-defined tags, subjectivity noise, and limited computational methods, which result in inconsistent access policies and security risks.
Innovation Solution
A data authorization decision point utilizing a large language model (LLM) with pre-trained rich semantics and prompt engineering to automatically generate arbitrary tags and access policies, reducing the need for extensive data and computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If user-defined tags are used for data authorization, then the system is easy to operate, but the policy consistency and security reliability deteriorate due to user subjectivity and potential errors
Solution Approach 1:
The system performs preliminary automated tagging of data files using an LLM before authorization decisions are made. The LLM analyzes the semantic content of each data file and generates appropriate tags in advance, eliminating the need for manual user tagging and ensuring consistent, accurate tagging that reflects the actual content semantics.
Solution Approach 2:
The patent introduces an LLM-based intermediary component that mediates between the raw data content and the authorization policy enforcement. This intermediary automatically extracts semantic meaning from data files and generates tags that serve as the basis for policy matching, removing human subjectivity from the tagging process while maintaining policy enforceability.
2Adaptability or versatility
If expert-driven approaches are used to derive access policies, then the policies can be customized, but the scalability and generalization capability deteriorate due to the complexity of deriving general rules
Solution Approach 1:
The system enables self-service automated policy derivation where the LLM autonomously analyzes data files, generates appropriate tags, and applies policies without requiring expert intervention for each case. The LLM learns from training data and automatically generalizes to new scenarios, maintaining both customization capability and scalability.
Solution Approach 2:
The patent transforms the complex problem of policy derivation into a parameter-based tagging system. Instead of requiring experts to define complex policies for each scenario, the system uses the LLM to extract semantic parameters (tags) from data content, which then map to predefined policies. This parameter transformation enables scalable policy application across diverse data types.
3Use of energy by moving object
If current data-driven methods are used for tag generation, then computational resources are limited, but the semantic richness and tag diversity deteriorate due to limited training corpus and modeling approaches
Solution Approach 1:
The patent replaces traditional mechanical NLP processing methods with an LLM-based semantic understanding system. The LLM leverages pre-trained knowledge and advanced transformer architecture to extract rich semantic meaning from data files, generating diverse and accurate tags without requiring exponentially more computational resources. This substitution enables high semantic richness while maintaining computational efficiency.
Data Source
AI summary
One example method includes receiving a data file at a large language model (LLM). Arbitrary tags that include labels that are attachable to the data file and prompts are also received. The prompts are paired with the arbitrary tags to form arbitrary tag-prompt pairs and include information that is used by the LLM to find the paired arbitrary tag. The LLM determines a selected subset of the arbitrary tags that apply to the data file. A trust module receives the selected subset of the arbitrary tags that apply to the data file and data access policies that specify access rules for the data file. A conditional access decision is determined that specifies whether access should be given to the data file.