Fraud Detection via Cluster Analysis of Accountant-Entity Connections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing business management applications (BMAs) lack effective methods to detect fraudulent loan applications submitted by accountants with access to multiple business entities, as current multi-factor authentication techniques do not adequately address the risk of account takeover leading to extensive fraudulent activities.
Innovation Solution
A method and system utilizing cluster analysis to determine connection strength between business entities and a fraud score for accounting firms, combined with a behavioral model to assess the probability of fraudulent loan applications, which involves receiving loan applications, calculating connection strength and fraud scores, and applying these to determine the likelihood of fraud.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multi-factor authentication (MFA) is implemented to safeguard accountant access to business entity accounts, then account security is improved, but it does not adequately detect fraudulent loan applications submitted by compromised accountant accounts
Solution Approach 1:
The patent segments the fraud detection process into multiple independent analysis components: cluster analysis for identifying groups of related business entities, connection strength analysis for measuring relationships between entities, fraud score calculation for assessing risk levels, and behavioral model analysis for detecting suspicious patterns. This segmentation allows each component to specialize in detecting specific fraud indicators without being constrained by a single authentication mechanism.
Solution Approach 2:
The patent introduces an intermediary fraud detection system that operates between the accountant's authenticated access and the loan application processing. This intermediary layer analyzes the accountant's actions, business entity relationships, and application patterns to detect fraud without blocking legitimate accountant activities. The system acts as a mediator that allows authorized access while filtering out fraudulent behavior.
2Productivity
If accountants are granted access to multiple business entity accounts via BMA, then operational efficiency is improved, but the impact of a single accountant account takeover becomes more extensive
Solution Approach 1:
The patent implements preliminary fraud detection measures before loan applications are processed. The system pre-calculates connection strengths between business entities, pre-assesses fraud scores for accountant accounts, and pre-identifies suspicious patterns through behavioral modeling. This preliminary analysis allows the system to detect fraud risks before they materialize into financial losses, maintaining operational efficiency while preventing extensive fraud impact.
Solution Approach 2:
The patent establishes a feedback mechanism where the fraud detection system continuously monitors accountant activities across multiple business entities and adjusts fraud scores based on observed patterns. When suspicious behavior is detected in one entity, the system provides feedback that triggers enhanced scrutiny of related entities through cluster analysis. This feedback loop allows the system to dynamically respond to fraud risks while maintaining efficient operations for legitimate activities.
3Measurement precision
If cluster analysis is used to determine connection strength and fraud scores, then fraud detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies partial cluster analysis by focusing computational resources on identifying and analyzing only the most relevant clusters of business entities connected to accounts showing fraud indicators. Rather than performing exhaustive cluster analysis on all possible entity relationships in the system, the method selectively analyzes connections for accounts that trigger fraud detection thresholds, thereby maintaining high detection accuracy while reducing overall computational complexity.
Data Source
AI summary
A method for fraud detection may include receiving, via a first user account of a business management application (BMA), a first loan application for a first business entity. The first user account may be accessible to an accountant of an accounting firm. The method may further include receiving, via a second user account of the BMA, a second loan application for a second business entity. The second user account may be accessible to the accountant. The method may further include determining, using a cluster analysis, (i) a connection strength between the first business entity and the second business entity relative to the accounting firm, and (ii) a fraud score for the accounting firm, and determining, based on the connection strength and the fraud score, a probability that the first loan application is fraudulent.