Probabilistic Frequency Filters for Cross-Domain Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting online fraud across multiple domains are inadequate, particularly in browsers that do not support third-party cookies, and they often fail to detect fraudulent activity that spans multiple domains.
Innovation Solution
The use of probabilistic data structures, such as Bloom filters, to represent frequency filters, which are used to track event counts across multiple domains, allowing for the detection of fraudulent activity while preserving user privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If third-party cookies are used to track event counts across multiple domains, then fraud detection capability is improved, but user privacy protection deteriorates
Solution Approach 1:
The patent introduces a probabilistic data structure (Bloom filter) as an intermediary between the tracking system and user data. This structure allows the system to verify whether a user has exceeded frequency thresholds without revealing which specific domains or events triggered the thresholds, thus mediating between fraud detection needs and privacy protection
Solution Approach 2:
The patent creates a simplified copy of the frequency filtering logic that runs locally in the browser. Instead of relying on server-side third-party cookies to track and report all user activity, the system copies the essential filtering capability to the client side, where it can operate with minimal privacy exposure
2Measurement precision
If detailed cross-domain user activity information is collected, then fraud detection precision is improved, but user privacy loss increases
Solution Approach 1:
The patent extracts only the essential information needed for fraud detection (whether frequency thresholds are exceeded) while leaving out detailed information about which specific domains or events caused the thresholds to be exceeded. This extraction approach maintains detection precision by capturing the critical signal while filtering out unnecessary privacy-sensitive details
Solution Approach 2:
The patent applies different levels of information detail to different parts of the system. The browser maintains detailed local state for accurate frequency counting, but only reports aggregated threshold-exceeded information to content providers, creating a local quality distinction between detailed local tracking and minimal remote reporting
3Reliability
If frequency filters are applied across multiple domains, then fraud detection coverage is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple frequency filter configurations from different content providers into a single probabilistic data structure (Bloom filter). Instead of implementing separate tracking mechanisms for each domain's frequency filters, the system combines them all into one unified structure that can be efficiently queried by any content provider
Solution Approach 2:
The patent changes the parameter representation from detailed domain-specific frequency counts to a probabilistic threshold-exceeded indicator. This parameter transformation simplifies the data structure by converting multiple detailed parameters (domain, event type, count) into a single probabilistic state (threshold exceeded or not)
Data Source
AI summary
This disclosure relates to using probabilistic data structures to enable systems to detect fraud while preserving user privacy. In one aspect, a method includes obtaining a set of frequency filters. Each frequency filter defines a maximum event count for a specified event type over a specified time duration and corresponds to a respective content provider. A subset of the frequency filters are identified as triggered frequency filters for which an actual event count for the specified event type corresponding to the frequency filter exceeds the maximum event count defined by the frequency filter during a time period corresponding to a specified time duration for the frequency filter. A probabilistic data structure that represents at least a portion of the frequency filters in the subset of frequency filters is generated. A request for content is sent to multiple content providers. The request for content includes the probabilistic data structure.