Textual Embedding Space Authentication Answer Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing knowledge-based authentication (KBA) schemes are vulnerable to fraud as answers to authentication questions can be easily obtained by fraudsters, leading to unauthorized access to user accounts.
Innovation Solution
The use of textual embedding space software engines to generate embedding vectors for authentication answers, allowing for the comparison of semantic and discourse similarity between verified and unverified user responses, thereby enhancing authentication security.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional knowledge-based authentication schemes are used, then the authentication process is simple and easy to implement, but the system becomes vulnerable to fraud as answers can be easily obtained by fraudsters
Solution Approach 1:
The patent transforms authentication answers from discrete text strings into continuous embedding vectors in a high-dimensional space. This parameter transformation allows for semantic similarity comparison, enabling the system to verify identity even when answers are phrased differently, thus improving security against fraud while maintaining system feasibility
Solution Approach 2:
The patent introduces textual embedding space as an intermediary representation layer between the authentication questions/answers and the verification process. This embedding space acts as a mediator that captures semantic meaning, allowing the system to compare answers based on their meaning rather than exact text matching, thereby enhancing security without excessive complexity
2Adaptability or versatility
If exact text matching is used for authentication answers, then the verification process is straightforward, but the system cannot accommodate varied phrasing of the same meaning
Solution Approach 1:
The patent changes the parameter representation from exact text matching to semantic similarity measurement in embedding space. This allows the system to tolerate varied phrasing while maintaining verification accuracy by comparing the semantic meaning of answers rather than their literal text
Solution Approach 2:
The patent moves authentication verification from one-dimensional exact text matching to multi-dimensional semantic space comparison. By projecting answers into high-dimensional embedding vectors, the system gains the ability to measure similarity across multiple semantic dimensions, accommodating answer variation while preserving verification precision
Data Source
AI summary
A method and system performed by a processor includes receiving from a verified user, an authentication answer for identity-authentication questions. An authentication answer embedding vector in a textual embedding space is generated by inputting each authentication answer into an embedding engine and stored. An unverified-user authentication answer is received, in response to posing to the unverified user, a specific identity-authentication question of the verified user. An unverified-user authentication answer embedding vector is generated using the embedding engine. An embedding space distance is computed between the unverified-user authentication answer embedding vector and the authentication answer embedding vector for the specific identity-authentication question of the verified user posed to the unverified user. A similarity score based on the embedding space distance is computed. The unverified user is identified as the verified user when the similarity score is higher than a predefined threshold score.