Fraud Account Classification Using Shared-Attribute Graph Links
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to effectively detect fraudulent user accounts before they engage in fraudulent activities, allowing malicious users to create new accounts and continue their illicit actions, leading to ongoing losses for service providers.
Innovation Solution
An account classification system analyzes shared attributes between a new user account and multiple known fraudulent accounts to determine a risk level, using a graph-based approach to identify links and assign weights to different attributes and accounts, employing machine learning to classify the new account as fraudulent.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the service provider deactivates fraudulent user accounts after detection, then further losses from detected accounts are prevented, but malicious users can create new accounts and continue fraudulent activities
Solution Approach 1:
The system performs preliminary actions by analyzing account attributes and detecting fraudulent patterns before fraudulent activities are conducted. The account classification module evaluates new accounts against known fraudulent account characteristics, enabling the service provider to prevent fraudulent activities proactively rather than reactively after detection
Solution Approach 2:
The system implements feedback mechanisms by continuously monitoring account activities and updating the classification of accounts based on shared attributes with known fraudulent accounts. The loss value derived from shared attributes provides feedback that adjusts the risk assessment, allowing the system to adapt and improve detection accuracy over time
2Measurement precision
If the service provider monitors account activities to detect fraudulent accounts, then fraudulent activities can be identified, but detection occurs only after fraudulent activities have already been conducted
Solution Approach 1:
The system performs preliminary classification of accounts by evaluating account attributes against known fraudulent account characteristics before fraudulent activities are conducted. This allows the service provider to identify and prevent fraudulent accounts proactively, eliminating the detection delay inherent in traditional post-activity monitoring approaches
3Extent of automation
If the service provider analyzes individual fraudulent accounts independently, then each account can be classified, but links between multiple fraudulent accounts created by the same malicious user are missed
Solution Approach 1:
The system merges the analysis of multiple fraudulent accounts by collecting and evaluating shared attributes across accounts. The account classification module derives loss values based on the collective shared attributes between a particular user account and two or more known fraudulent user accounts, enabling the detection of patterns linking accounts to the same malicious user
Solution Approach 2:
The system implements a universal classification approach that evaluates accounts based on shared attributes with multiple known fraudulent accounts simultaneously. The graph-based loss value calculation methodology can be applied across different account pairs, making the system versatile in detecting various fraudulent patterns regardless of the specific accounts involved
Data Source
AI summary
Methods and systems are presented for classifying a particular user account as a fraudulent user account by analyzing links between the user account and two or more known fraudulent user accounts collectively. Attributes of the particular user account are compared against attributes of a plurality of known fraudulent accounts to determine that the particular user account has shared attributes with a first known fraudulent account and a second known fraudulent account. The shared attributes with the first known fraudulent account and the second known fraudulent account are analyzed collectively to determine a risk level for the particular user account. The risk level may indicate a likelihood that the particular user account corresponds to a fraudulent account.