Synthetic Identity Detection via ML Collision Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current synthetic identity detection methods are limited in effectiveness and accuracy due to the incomplete utilization of stored information, leading to potential failures in identifying synthetic identities.
Innovation Solution
A computing platform with a synthetic identity detection model is trained to identify collisions between received identity information and stored information, generating a synthetic identity score that indicates the likelihood of a synthetic identity generation attempt. The platform compares this score to predefined thresholds to trigger security actions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional synthetic identity detection methods are used, then the system is simple to implement, but the detection accuracy and effectiveness are limited
Solution Approach 1:
The patent replaces traditional rule-based mechanical detection systems with a machine learning model that automatically learns patterns from data. The synthetic identity detection model uses machine learning algorithms to analyze identity information and detect synthetic identities, substituting manual rule creation with automated intelligent analysis that improves accuracy while managing complexity through algorithmic processing.
Solution Approach 2:
The patent changes the detection parameters from fixed rules to dynamic learned parameters. The machine learning model continuously learns from new data, adjusting its detection parameters and thresholds based on patterns in the training data. This allows the system to adapt to evolving synthetic identity creation methods while maintaining high detection accuracy.
2Reliability
If more stored information is utilized for detection, then the detection effectiveness improves, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by pre-processing and storing identity information in structured formats before detection is needed. The system pre-computes features from identity data and stores them in optimized data structures, so that during actual detection, the machine learning model can quickly access and analyze pre-prepared information rather than processing raw data from scratch.
Solution Approach 2:
The patent creates simplified copies of complex identity information in the form of feature vectors and extracted attributes. Instead of processing entire identity records, the system creates condensed representations that retain essential detection-relevant information, allowing faster processing while maintaining detection effectiveness through the use of these copied feature sets.
3Measurement precision
If collision detection is performed on all identity information, then comprehensive detection is achieved, but the computational complexity increases
Solution Approach 1:
The patent extracts only the most relevant features and attributes from complete identity information for collision detection. The machine learning model identifies and extracts key discriminative features that are most indicative of synthetic identities, performing collision detection on these extracted features rather than on all raw identity data, thereby reducing computational complexity while maintaining detection comprehensiveness.
Solution Approach 2:
The patent segments identity information into distinct feature categories and processes them through different detection pathways. The system divides identity data into segments such as demographic features, account information, and behavioral patterns, applying specialized detection methods to each segment before integrating results, which reduces overall computational complexity through divide-and-conquer processing.
Data Source
AI summary
A computing platform may train a synthetic identity detection model to detect synthetic identity information. The computing platform may receive identity information corresponding to an identity generation request. The computing platform may input, into the synthetic identity detection model, the identity information, which may cause the synthetic identity detection model to: identify at least one collision between the received identity information and stored identity information, and generate, based on the at least one collision, a synthetic identity score indicating a likelihood that the received identity information corresponds to a request to generate a synthetic identity. The computing platform may compare the synthetic identity score to at least one synthetic identity detection threshold. Based on identifying that the synthetic identity score meets or exceeds the at least one synthetic identity detection threshold, the computing platform may execute one or more security actions.