Embedding-Space Query Obfuscation for Secure Search Intent
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional query obfuscation methods expose sensitive information and search intent to content providers by sending decoy queries that still reveal the actual query, compromising security.
Innovation Solution
Utilize large language models to distort query regions in an embedding space, generating modified and decoy queries that conceal the search intent, allowing secure search results to be obtained without exposing the original query to the content provider.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If decoy queries are added to the actual query to conceal search intent, then search security is improved, but the actual query is still exposed to the content provider
Solution Approach 1:
The patent segments the query processing into multiple independent components: embedding generation, region distortion, and query modification. Each component operates separately to transform the original query into obfuscated versions, ensuring that no single point exposes the full query intent to the content provider
Solution Approach 2:
The patent introduces a geometric dimension to query obfuscation by working with embedding spaces and distorting regions within them. By transforming queries through spatial operations (rotation, scaling, translation) in the embedding space, the system creates obfuscated queries that are semantically different from the original, preventing the content provider from inferring the actual search intent
2Reliability
If decoy queries are generated to confuse the content provider, then search intent concealment is improved, but the query processing complexity increases
Solution Approach 1:
The patent creates multiple copies of the query in the form of obfuscated versions by applying different distortion transformations to the embedding region. These copied queries are then processed independently and their results aggregated, allowing the system to maintain simplicity at each processing step while achieving complex obfuscation through repetition
Solution Approach 2:
The patent changes parameters of the query representation by modifying embedding space coordinates through geometric transformations. By altering positional parameters (rotation angles, scaling factors, translation vectors) rather than the query content itself, the system achieves obfuscation through parameter manipulation while keeping the overall processing framework simple
Data Source
AI summary
According to a present invention embodiment, a system for processing queries for secure searches comprises one or more memories and at least one processor. The system determines for a query, via a first machine learning model, a region of an embedding space corresponding to search results. The region of the embedding space is distorted along one or more dimensions to produce distorted regions in the embedding space corresponding to different search results. A second machine learning model determines modified queries corresponding to the different search results of the distorted regions to produce obfuscated queries. Results are obtained from processing the obfuscated queries. A response to the query is produced based on the results for the obfuscated queries. Embodiments of the present invention further include a method and computer program product for processing queries for secure searches in substantially the same manner described above.