Detecting Financial Collusion Through Graph Analytics: A Procure-to-Pay and Payroll Fraud Detection Framework

Abstract

Collusion remains one of the most difficult fraud schemes to detect because it is designed to defeat traditional rule-based controls and segregation of duties. Conventional audit analytics focus on individual transactions and attributes, while collusion operates through relationships between people, vendors, bank accounts, and approval chains. This article proposes a graph-analytics approach for identifying collusive behavior in procure-to-pay and payroll processes. The article introduces a network data model, risk indicators, and detection algorithms that combine structural graph metrics with financial attributes. The proposed framework demonstrates how internal audit functions can move from exception testing to relationship-driven continuous monitoring, improving detection of kickbacks, ghost employees, vendor favoritism, and approval rings.