On-Device PII Neutralization for Personalized AI Responses

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The increasing use of generative conversational AI, such as ChatGPT, has highlighted the need for an on-device system that can suppress the leakage of Personal Identifiable Information (PII) while providing personalized responses without transmitting PII to management servers.

Innovation Solution

An on-device-based system that detects PII from user queries, converts it into neutral information, and transmits user-neutral queries to a management server. The management server trains a language model to generate common response patterns for each neutral query pattern, ensuring that personalized responses are generated locally on the user device without exposing PII.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If PII is transmitted to management server for training language model, then personalized response can be provided, but PII leakage risk increases

Engineering Contradiction:

Improvepersonalized response capabilityVSAvoidPII leakage risk

Core Design Contradiction:

Adaptability or versatilityVSObject-affected harmful factors

Solution Approach 1:

The patent extracts PII from user queries using detection models before transmission, removing the harmful element (PII) while retaining the useful information (query intent) for training the language model on the management server

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces an intermediary processing step where PII detection models and neutralization techniques are applied between the user device and management server, acting as a mediator that protects PII while enabling personalized response training

Inventive Principle:

Principle #24Intermediary (Mediator)

2Object-affected harmful factors

If PII is stored only on user device, then PII leakage is suppressed, but personalized response generation capability is reduced

Engineering Contradiction:

ImprovePII leakage suppressionVSAvoidpersonalized response generation capability

Core Design Contradiction:

Object-affected harmful factorsVSAdaptability or versatility

Solution Approach 1:

The patent segments the system into two parts: PII detection and neutralization on user device, and language model training on management server with neutralized data, allowing each component to perform its function optimally while maintaining overall system security and capability

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent changes the state of PII data by converting it into neutralized information that retains structural and semantic properties for training purposes while removing personally identifiable characteristics, enabling model training without actual PII exposure

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If more PII is processed on device for personalization, then user-customized response improves, but device storage and processing requirements increase

Engineering Contradiction:

Improveuser-customized responseVSAvoiddevice storage and processing requirements

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

Instead of storing and processing large amounts of PII data on the user device, the patent inverts the approach by sending neutralized data to the management server for model training, reducing device complexity while maintaining personalization capability

Inventive Principle:

Principle #13The other way round (Inversion)

Data Source

PatentUS20250086318A1On device-based system and method for suppressing leakage of personal information and for providing personalized response

Publication Date: 2025.03.13 SEOUL NATIONAL UNIVERSITY R&DB FOUNDATION

AI summary

According to an embodiment, an on device-based system for suppressing the leakage of personal information and for providing personalized response includes: a plurality of user devices configured to detect Personal Identifiable Information (PII) from a user query and transmit a user-neutral query which is obtained by converting the PII into neutral information; and a management server configured to receive the user-neutral query and train a management language model that generates a common response pattern for each neutral query pattern.