Multi-Party AI Query Segmentation for Confidential Data Security
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-party data science collaboration systems fail to prevent data leakage when interacting with large-scale public AI and LLMs, and they lack the ability to separate and manage confidential from non-confidential sub-queries effectively, leading to potential data breaches.
Innovation Solution
The system disassembles human user queries into sub-queries, tokens, and tensors, identifies and tags them as confidential or non-confidential using an AI-trained learning model, and processes confidential data within a secure Multi-Party Collaborator while routing non-confidential data to public AI/LLM systems for processing, ensuring confidential data remains quarantined and preventing breaches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If all queries are processed by large public AI/LLM systems, then processing capability and intelligence are improved, but confidential data security deteriorates due to data leakage risk
Solution Approach 1:
The system segments queries into confidential and non-confidential portions using an AI classifier that analyzes each query and separates sensitive data from non-sensitive data. This allows different processing paths: confidential queries are handled by local AI models while non-confidential queries can be processed by public AI/LLM systems, thus maintaining both security and processing capability.
Solution Approach 2:
The patent introduces an intermediary system comprising an AI classifier and query separation mechanism that acts as a mediator between user queries and processing systems. This intermediary identifies and extracts confidential information, routing it appropriately while allowing non-confidential portions to utilize public AI/LLM resources, thereby resolving the contradiction between security and capability.
2Reliability
If confidential data is quarantined and processed separately, then data security is improved, but system complexity increases due to multiple processing paths
Solution Approach 1:
The system employs a universal AI classifier that can handle multiple functions: identifying confidential information, separating queries, and routing to appropriate processing systems. This multi-functional approach reduces overall system complexity compared to having separate specialized components for each function, while still maintaining secure confidential data processing.
Solution Approach 2:
The AI classifier automatically identifies and separates confidential information without requiring manual intervention or complex configuration. The system self-manages the classification and routing processes, reducing operational complexity while maintaining high security standards for confidential data.
3Measurement precision
If AI classifiers are trained on organization-specific confidential data, then classification accuracy is improved, but data breach risk increases during the training process
Solution Approach 1:
The system extracts only the necessary characteristics and patterns from confidential data for training purposes, rather than using the actual confidential data itself. The AI classifier is trained on synthesized or anonymized representations that capture the essential features needed for accurate classification without exposing the actual confidential information, thus maintaining both accuracy and security.
Solution Approach 2:
Instead of training directly on confidential data, the system uses copies or synthesized representations that preserve the structural and contextual features needed for accurate classification while eliminating the security risks associated with using real confidential data during training processes.
Data Source
AI summary
This present disclosure includes Multi-Party Collaborators and methods wherein: A given business or public organization establishes a secure digital workspace, a multi-party collaborator. This organization trains the multi-party collaborator using a multi-party AI secrets learner with a training data set of confidential/non-confidential data samples. Once the multiparty collaborator is trained, human users submit AI/LLM queries within the Multi-Party Collaborator. The human submitted query is disaggregated into machine readable tagged, sub-queries, and the AI classifier sorts the dis-aggregated sub-queries into confidential and non-confidential. The Query manager, within the multi-party collaborator, routes the non-confidential sub-query to the large public AI/LLMs, for public processing, routes the confidential sub-query within the Multi-party collaborator for quarantined processing. Taking advantage of large public AI/LLMs where possible and quarantining confidential data where appropriate, and combining the results into a coherent, human-readable result, for additional human interaction and AI Machine iteration.