Fraud detection in process aware systems

Abstract

In the last years, some large companies have been involved in scandals related to accounting and financial mismanagement, which represented a large financial damage to their stockholders. To recover the stock market confidence, certifications and manuals for best practices of governance were developed, and in some cases, harder federal laws were implemented (eg.: Sarbox, in USA). Companies adhered to these changes as a response to the market, deploying process aware information systems (PAS) and adopting the best practices of governance. On the other hand, companies demand a rapid response to strategic changes, so the adoption of normative PAS may compromise their competitiveness. That is, the companies need flexible PAS for competitiveness reasons; whereas, flexibility may compromise the security of the system because the users can execute tasks that characterize a fraudulent execution. In order to re-balance the trade-off between security and flexibility, we present in this work an anomaly detection algorithm for logs of PAS. We believe that the identification of anomalous events can help the adoption of flexible PAS without the loss of security properties.