Inferred Identifier Generation for Hashed Email Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online systems face challenges in matching users with hashed email addresses from third-party systems, as they cannot directly recover useful data from hashed identifiers, limiting the ability to identify and connect users with different but equivalent email addresses.
Innovation Solution
The online system generates local hashes for its users and infers additional identifiers based on user profile characteristics, such as preferences and demographics, to match with hashed identifiers from third-party systems, while maintaining a low error rate by adjusting generation parameters and discarding common elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the online system uses only exact hash matching for user identification, then the matching process is simple and fast, but the ability to match users with different email addresses is lost
Solution Approach 1:
The patent segments the user identification process into two distinct phases: exact hash matching for direct matches, and inferred identifier matching for alternative email addresses. This segmentation allows the system to maintain fast exact matching while adding versatility through multiple matching strategies without compromising overall performance
Solution Approach 2:
The patent introduces inferred identifiers as an intermediary mechanism between exact hash matching and user identification. These inferred identifiers act as mediators that bridge the gap when exact matches fail, enabling the system to match users with different email addresses while maintaining the simplicity of hash-based comparison
2Adaptability or versatility
If the online system generates multiple inferred identifiers for each user, then the ability to match additional users increases, but the complexity of the identification process increases
Solution Approach 1:
The patent changes parameters such as email domain variations, username formats, and identifier structures to generate multiple inferred identifiers from a single user profile. By systematically varying these parameters, the system expands matching capability while maintaining a structured, manageable process rather than arbitrary complexity
Solution Approach 2:
The patent creates a universal identifier generation system that produces multiple inferred identifiers serving the same matching function. Each inferred identifier is designed to be interchangeable in the matching process, allowing the system to handle diverse email formats and structures through a unified multi-functional approach
3Reliability
If the online system uses hashed identifiers, then user privacy is protected, but the ability to recover and use the original identifier data is lost
Solution Approach 1:
The patent performs preliminary actions by generating inferred identifiers from available user profile information before the matching process begins. This allows the system to prepare multiple potential match candidates in advance, maximizing the utility of hashed identifiers while maintaining privacy, as all inference is done server-side without exposing sensitive data
Data Source
AI summary
An online system receives third party hashes for a plurality of targeted users and generates local hashes for one or more local users of the online system. The online system identifies as matched users those local users with local hashes that match those of the third party hashes. The online system generates one or more inferred identifiers for each of the one or more local users, the inferred identifiers being of the same type as the local unique identifiers, and the inferred identifiers generated based on characteristics of each corresponding local user. The online system identifies as inferred matched users at least one of the local users that have local hashes of corresponding inferred identifiers that match a third party hash of a third party unique identifier. The online system provides, to a third party system, a selection including the matched users and a selection including the inferred matched users.