Privacy Safe Identity Matching Using Salted Hashing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data analysis systems face challenges in performing attribution processes for external content while maintaining user privacy, as they often involve handling personally identifiable information (PII) that can be exposed during impression and conversion tracking.
Innovation Solution
A privacy-safe attribution system is implemented using salt values to anonymize PII from impression records and conversion requests, employing a daily salt generation and discarding process, combined with machine learning techniques for matching user identifiers, to ensure privacy and prevent inference of individual user data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional attribution tracking methods are used to measure conversion effectiveness, then accurate attribution metrics can be obtained, but personally identifiable information (PII) is exposed during impression and conversion tracking
Solution Approach 1:
The patent introduces an intermediary anonymization layer that transforms PII into hashed identifiers before processing. Conversion requests and impression records both pass through this layer, which uses hashing functions to create anonymized versions of user identifiers. This intermediary process enables accurate matching and attribution while preventing direct exposure of original PII, as the hashed values cannot be reversed to reveal original identities.
Solution Approach 2:
The patent creates anonymized copies of user identifiers through hashing. Instead of using original PII directly in attribution tracking, the system generates hashed copies that preserve the ability to match impressions to conversions while eliminating the ability to identify individual users. These copied hashed identifiers are used throughout the attribution process, allowing metrics accuracy to be maintained without exposing original sensitive data.
2Reliability
If PII is anonymized using hashing functions, then user privacy is protected, but the ability to accurately match impressions to conversions may be reduced
Solution Approach 1:
The patent applies preliminary hashing to both impression records and conversion requests before they enter the attribution matching process. By pre-anonymizing both data streams using the same hashing function, the system ensures that when matching occurs, both sides are already in the correct anonymized format. This preliminary action eliminates the need to hash data during matching, preserving accuracy while maintaining privacy throughout the entire process.
Solution Approach 2:
The patent transforms the parameter format of user identifiers from plaintext PII to hashed values. This parameter change is applied consistently across all impression and conversion data, allowing the matching algorithm to operate on the transformed parameters. The hashing function preserves the uniqueness and matching capability of identifiers while changing their representation form to protect privacy, thus maintaining measurement precision in the anonymized domain.
3Reliability
If salt values are generated and discarded daily, then privacy security is enhanced by preventing inference of individual user data, but system complexity increases
Solution Approach 1:
The patent implements periodic salt generation and discarding on a daily basis. Each day, new salt values are generated for hashing operations, and after use, they are discarded. This periodic rotation ensures that even if one day's hashed data is compromised, attackers cannot infer information about users from other days' data, as different salts were used. The periodic action enhances privacy security by preventing longitudinal inference while maintaining a manageable complexity through automated daily processes.
Data Source
AI summary
An example computer-implemented system maintains user profiles and displays external content. Method and system are provided for performing attribution of conversions with respect to the external content in a privacy safe manner by anonymizing personally identifiable information utilizing cryptographic salt.