Nondeterministic Model Output Permission Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Nondeterministic models, such as large language models, are difficult to secure due to their opaque nature of accessing and processing information, making it challenging to determine and apply appropriate permissions to their outputs, which inhibits their use in secure applications.
Innovation Solution
A system that determines and propagates permissions to nondeterministic model outputs during runtime operations, allowing for the secure management of access to data inputs by identifying the data used by the model and applying permissions accordingly, enabling secure chaining with other functions and models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If permissions are applied to nondeterministic model outputs, then security and data access control are improved, but the complexity of determining and applying permissions increases due to the opaque nature of model operations
Solution Approach 1:
The patent introduces a permission determination system as an intermediary component that sits between the nondeterministic model and the data inputs/outputs. This mediator tracks which data inputs are accessed by the model during runtime and automatically determines which permissions should be applied to the outputs, thereby managing the complexity centrally rather than distributed throughout the system.
Solution Approach 2:
The system performs preliminary tracking of data input access during model execution before permissions need to be applied. By monitoring and recording which data inputs are accessed during runtime, the system prepares permission information in advance, making the subsequent permission application process more straightforward and reducing overall system complexity.
2Reliability
If runtime permission propagation is implemented, then data access control is improved, but the operational complexity and processing overhead increase
Solution Approach 1:
The permission determination system operates autonomously by automatically tracking data input access and determining applicable permissions without requiring manual intervention. The system self-manages the complexity of runtime permission propagation, reducing operational burden on users while maintaining strong data access control.
3Reliability
If permission tracking is performed during runtime, then security is improved, but processing time and computational overhead increase
Solution Approach 1:
The patent extracts the permission tracking functionality from the main model execution path by implementing it as a separate, dedicated tracking mechanism. This allows the model to execute efficiently while the tracking system operates in parallel, minimizing the impact on processing time while maintaining security.
Data Source
AI summary
Computer-implemented systems and methods are disclosed, including for determining permissions for nondeterministic model output. A computer-implemented method may include, for example, receiving one or more user inputs including a first user input providing at least a portion of a first prompt for a query for a first nondeterministic model. A computer-implemented method may in response to receiving the one or more user inputs include: executing the query, by the first nondeterministic model, to generate an output, determining a first one or more data inputs used by the first nondeterministic model during execution of the query, determining a first set of permissions associated with the first one or more data inputs; and applying a second set of permissions to at least a first portion of the output based on the first set of permissions.