Prompt Engineering for Secure LLM Query Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current prompt engineering in generative AI systems lacks sufficient security measures, allowing unauthorized users to access sensitive information and posing a risk of inappropriate or unanswerable responses.
Innovation Solution
A prompt engineering system that includes an acquisition unit, detection unit, extraction unit, and prompt generation unit to analyze question data, detect user levels, and generate prompts that modify or refuse to answer based on the question's meaning and user authorization, ensuring secure interactions with large-scale language models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If prompt engineering allows high flexibility in asking questions, then the ease of operation is improved, but security is worsened due to unauthorized access to sensitive information
Solution Approach 1:
The patent introduces an intermediary system between the user's flexible question input and the generative AI model. This intermediary includes a sentence meaning determination unit that acts as a mediator to evaluate whether the question's meaning is answerable based on extracted keywords and questioner level, thereby maintaining security while preserving user flexibility.
Solution Approach 2:
The system performs preliminary actions by extracting keywords from the question before the generative AI processes it. The sentence meaning determination unit then evaluates the question's answerability in advance, preventing unauthorized access to sensitive information before it can be compromised.
2Adaptability or versatility
If the system accepts questions with canceling-out meanings, then the adaptability is improved, but the reliability is worsened due to unanswerable responses
Solution Approach 1:
The patent implements a feedback mechanism where the sentence meaning determination unit evaluates the question's meaning and provides feedback by determining whether it is answerable. This feedback loop ensures that questions with canceling-out meanings or unanswerable content are identified and handled appropriately, maintaining response accuracy while accepting diverse question types.
3Productivity
If the system processes all question data without filtering, then the productivity is improved, but security is worsened due to exposure of sensitive information
Solution Approach 1:
The patent extracts keywords from question data as a separate step before full processing. This extraction allows the system to identify and flag potentially sensitive or unanswerable questions efficiently, enabling security filtering without significantly impacting overall processing productivity.
Data Source
AI summary
A prompt engineering computer that generates a prompt for input into a large-scale language model is provided. The prompt engineering computer acquires question data, detects a questioner level, extracts a first keyword from the question data; determines whether or not a sentence meaning of the question data is answerable based on the first keyword and the questioner level, and generates a prompt that modifies at least a part of the first keyword or a prompt that refuses to answer if the sentence meaning is unanswerable.