Process-Based Fraud (PBF) is fraud enabled by process deviations that occur in business processes. Several studies have proposed PBF detection methods; however, false decisions are still often made because of cases with low deviation. Low deviation is caused by ambiguity in determining fraud attribute values and low frequency of occurrence. This paper proposes a method of detecting PBF with low deviation in order to correctly detect fraudulent cases. Firstly, the fraudulence attributes are established, then a fuzzy approach is utilized to weigh the importance of the fraud attributes. Further, multi-attribute decision making (MADM) is employed to obtain a PBF rating according to attribute values and attribute importance weights. Finally, a decision is made whether the deviation is fraudulent or not, based on the PBF rating. Experimental validation showed that the accuracy and false discovery rate of the method were 0.98 and 0.17, respectively.

Process-based fraud (PBF) is fraud caused by deviation from a business process model. Some studies have proposed methods for PBF detection; however, these are still not able to fully detect the occurrence of fraud. In this context, we propose a new method of PBF detection which carries out the behavior of the originators (users who perform events) to adjust the levels of fraud occured in the events. In this research, we propose a method of PBF detection with behavior model in order to increase accuracy. This is done firstly by analyzing the business processes that correspond to those in the standard operating system (SOP). Secondly, by calculating the event execution performed by the originator and his/her relations within the organization, whose behavior is then analyzed. Thirdly, by using the number of deviations and the originator behavior to calculate the attribute value. By using attribute importance weights, an attribute rating of each originator is kept. Finally, Multi Attribute Decision Making is used to decide the PBF rating of a case, on the basis of which it is decided whether fraud occurred or not. The experimental results show that this behavior model is able to reduce false positive and false negative, therefore, the method can increase the accuracy level by 0.01.

No abstract

Permainan batu, gunting, dan kertas sangat populer di seluruh dunia. Permainan ini biasanya dimainkan saat sedang berkumpul untuk mengundi ataupun hanya bermain untuk mengetahui yang menang dan yang kalah. Namun, perkembangan zaman dan teknologi mengakibatkan orang dapat berkumpul secara virtual. Untuk bisa melakukan permainan ini secara virtual, penelitian ini membuat model klasifikasi citra untuk membedakan objek tangan yang menunjuk batu, kertas, dan gunting. Performa metode klasifikasi merupakan hal yang harus diperhatikan dalam kasus ini. Salah satu metode klasifikasi citra yang populer adalah Convolutional Neural Network (CNN). CNN adalah salah satu jenis neural network yang biasa digunakan pada data klasifikasi citra. CNN terinspirasi dari jaringan syaraf manusia. Algoritma ini memiliki 3 tahapan yang dipakai, yaitu convolutional layer, pooling layer, dan fully connected layer. Uji coba 5-Fold cross validation klasifikasi objek tangan yang menunjuk citra batu, kertas, dan gunting menggunakan CNN pada penelitian ini menghasilkan rata-rata akurasi sebesar 97.66%.

Fraud detection has become an important research topic in recent years. In online sales transaction, fraud can occur on a business process. Fraud which occurs on business process is popularly known as process-based fraud (PBF). Previous studies have proposed PBF detection on process business model, however, false decisions are still often made because of new fraud pattern in online sales transactions. False decision mostly occurs since the method cannot identify the attributes of fraud in online sales transaction. This research proposes new fraud attributes and fraud patterns in online transactions. The attributes can be identified by exploring the event logs and Standard Operating Procedure (SOP) of online sales transactions. First, this is conducted by collecting event logs and creating SOP of online sales transaction; then, performing conformance between event logs and SOP; further, discussing with fraud experts about the result of SOP deviations which have been identified; moreover, determining convention value of the SOP deviation to fuzzy value, and classifying the SOP deviation; and at last, establishing fraud attributes and fraud patterns based on classification result. The new fraud attribute and fraud patterns are expected to increase accuracy of fraud detection in online sales transaction. Based on the evaluation, this method resulted a better accuracy 0.03 than the previous one.