Local LLM Authentication From Multidimensional Data Without Hardware
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user authentication methods, particularly those relying on passwords and two-factor authentication, are vulnerable to breaches and require additional hardware, making them cumbersome and insecure.
Innovation Solution
A multidimensional local large language model (LLM) on a user's device collects data from multiple dimensions such as geographic location, purchase history, and social media engagement to generate non-intersecting authentication questions, ensuring security by preventing a malicious user from answering all questions accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional passwords and two-factor authentication are used, then authentication security is improved, but device complexity and ease of operation deteriorate due to additional hardware requirements
Solution Approach 1:
The patent extracts the authentication process from traditional server-based systems and relocates it to the user's local device. The local LLM stores and processes user data independently, eliminating the need for additional authentication hardware while maintaining security through local computation and data processing.
Solution Approach 2:
The system enables self-service authentication by using the user's own device and data to generate and answer authentication questions. The local LLM automatically creates questions based on user profile information and verifies answers without requiring external authentication services or additional hardware devices.
2Reliability
If multiple dimensions of user data are collected for authentication, then authentication security is improved, but loss of information increases due to data handling requirements
Solution Approach 1:
The patent applies local quality by processing and storing user data locally on the user's device rather than centrally. The local LLM maintains user profiles and generates authentication questions from this local data, ensuring that sensitive information remains under user control and is not exposed to external systems.
Solution Approach 2:
The local LLM acts as an intermediary between the authentication system and user data. It processes data locally to generate authentication questions and verifies answers without transmitting or exposing the underlying user information, thus maintaining security while preventing information loss.
3Reliability
If local LLM is stored on user's device for privacy, then information security is improved, but device complexity increases due to LLM storage requirements
Solution Approach 1:
The patent applies parameter changes by optimizing the LLM for specific authentication tasks rather than using a full-scale general model. The local LLM is tailored to process user profile data and generate authentication questions, reducing computational and storage requirements while maintaining security effectiveness.
Data Source
AI summary
A computer user is authenticated by collecting user data from at least two dimensions of user activity into a local large language model (LLM) residing on a user device, and generating user authentication questions from at least those two dimensions of user activity using the large language model. Questions having intersecting time or location are discarded, and questions from two or more dimensions are presented to the user for response. The user is authenticated by comparing the user response to known correct answers to the questions.