Determining expected behavior of fraudsters for a continuous audit system

Abstract

This research attempts to determine the behavior of fraudsters in a continuous audit system where the fraudsters have multiple options for committing fraud. The system is modeled as a Continuous Time Markov Chain where the state changes are caused by the fraudster’s actions. The model uses a dynamic game with probabilistic transitions to determine the expected behavior of the fraudster. Managements and external auditors are expected to assess and evaluate controls that prevent and detect fraud. A key part of this assessment is determination of fraud-prone areas and the determination of those systems or sub-systems that are susceptible to fraud. There are several approaches that are used to find such areas. In this research, we attempt to model the behavior of fraudsters in order to help determine systems that require close monitoring. In this research we attempt to determine the behavior of fraudsters in a continuous audit system and where the fraudsters have multiple options for committing fraud. The system is modeled as a Continuous Time Markov Chain where the state changes are caused by the fraudster’s actions. In each state the fraudster can either choose to continue the fraud or to resign. The possibility of success in each case is assumed to be probabilistic. This state of affairs is represented as a dynamic game with probabilistic transitions. Since the fraudster assumes that audit system is a rational player who seeks to minimize the fraudster’s expected payoff, the Nash Equilibrium of the complete stochastic game is calculated. The Nash Equilibrium is a set of minimax vectors. These solution vectors represent a complete fraud strategy, which maximizes the expected payoff of the fraudster and ensures that she has no ex-post regrets.