User Account Characteristic Group Detection for Fraud Prevention
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for preventing malicious account registrations on websites are inadequate, as they fail to detect a majority of fraudulent accounts, especially those generated through manual registrations and do not effectively utilize user identifier connections to identify abnormal activity.
Innovation Solution
A system and method for determining characteristic groups of user accounts by analyzing mapping relationship data between user account numbers and user identifiers, identifying candidate groups through direct and indirect associations, and determining characteristic values to identify potentially malicious accounts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If verification code input step is added during user registration process, then automated program registration is prevented, but malicious registrations through manual registrations are not prevented
Solution Approach 1:
The system performs preliminary analysis of user identifier mapping relationships during and after the registration process. By pre-establishing the mapping framework between user account numbers and user identifiers (cookies, device IDs, IP addresses), the system can proactively identify characteristic groups of malicious accounts before they cause significant harm, rather than merely reacting to automated registration attempts.
Solution Approach 2:
The patent introduces user identifier mapping relationships as an intermediary mechanism to connect user account numbers with various identifying information (cookies, device IDs, IP addresses). This intermediary layer enables the system to indirectly detect malicious registrations by analyzing the relationships between accounts and identifiers, even when traditional verification methods fail to prevent manual fraud.
2Measurement precision
If post-registration processing checks number of account logins associated with same cookie, then some malicious accounts are discovered, but majority of malicious accounts are not detected
Solution Approach 1:
The patent transitions from checking a single dimension (number of logins per cookie) to analyzing multi-dimensional mapping relationships between user account numbers and various user identifiers including cookies, device IDs, IP addresses, and registration information. This dimensional expansion allows the system to detect malicious accounts that would be invisible to single-dimension checking methods.
Solution Approach 2:
The mapping relationship data structure serves multiple functions: it tracks login behavior, identifies characteristic groups, detects malicious patterns, and supports various detection strategies simultaneously. This universal data structure replaces multiple separate checking mechanisms, providing both precise detection and broad coverage through its multi-functional design.
3Reliability
If characteristic group recognition mechanism is implemented, then malicious account detection is improved, but system complexity increases
Solution Approach 1:
The patent segments the malicious account detection problem into distinct modules: (1) collecting mapping relationship data between accounts and identifiers, (2) determining characteristic groups based on predefined rules, and (3) processing and analyzing these groups to identify malicious accounts. This segmentation reduces system complexity by breaking down the complex recognition task into manageable, independent components that can be implemented and maintained separately.
Data Source
AI summary
Determining a characteristic group is disclosed, including: retrieving stored mapping relationship data for first objects and second objects, wherein the mapping relationship data describes connections between the first objects and the second objects; determining a candidate group including at least some of the first objects and at least some of the second objects based at least in part on the mapping relationship data, wherein any first object included in the candidate group is associated with another first object included in the candidate group via one or more other first object(s) or second object(s) included in the candidate group; determining a first characteristic value associated with the candidate group based at least in part on a number of first objects that are included in the candidate group; and determining whether the candidate group comprises a characteristic group based at least in part on the first characteristic value.