This thesis introduces a financial simulation model covering two related financial domains: Mobile Payments and Retail Stores systems.

The problem we address in these domains is different types of fraud. We limit ourselves to isolated cases of relatively straightforward fraud. However, in this thesis the ultimate aim is to introduce our approach towards the use of computer simulation for fraud detection and its applications in financial domains. Fraud is an important problem that impact the whole economy. Currently, there is a lack of public research into the detection of fraud. One important reason is the lack of transaction data which is often sensitive. To address this problem we present a mobile money Payment Simulator (PaySim) and Retail Store Simulator (RetSim), which allow us to generate synthetic transactional data that contains both: normal customer behaviour and fraudulent behaviour. 

These simulations are Multi Agent-Based Simulations (MABS) and were calibrated using real data from financial transactions. We developed agents that represent the clients and merchants in PaySim and customers and salesmen in RetSim. The normal behaviour was based on behaviour observed in data from the field, and is codified in the agents as rules of transactions and interaction between clients and merchants, or customers and salesmen. Some of these agents were intentionally designed to act fraudulently, based on observed patterns of real fraud. We introduced known signatures of fraud in our model and simulations to test and evaluate our fraud detection methods. The resulting behaviour of the agents generate a synthetic log of all transactions as a result of the simulation. This synthetic data can be used to further advance fraud detection research, without leaking sensitive information about the underlying data or breaking any non-disclose agreements.

Using statistics and social network analysis (SNA) on real data we calibrated the relations between our agents and generate realistic synthetic data sets that were verified against the domain and validated statistically against the original source.

We then used the simulation tools to model common fraud scenarios to ascertain exactly how effective are fraud techniques such as the simplest form of statistical threshold detection, which is perhaps the most common in use. The preliminary results show that threshold detection is effective enough at keeping fraud losses at a set level. This means that there seems to be little economic room for improved fraud detection techniques.

We also implemented other applications for the simulator tools such as the set up of a triage model and the measure of cost of fraud. This showed to be an important help for managers that aim to prioritise the fraud detection and want to know how much they should invest in fraud to keep the loses below a desired limit according to different experimented and expected scenarios of fraud.

RetSim is a multi-agent based simulator (MABS) calibrated with real transaction data from one of the largest shoe retailers in Scandinavia. RetSim allows us to generate synthetic transactional data that can be publicly shared and studied without leaking business sensitive information, and still preserve the important characteristics of the data.

In this paper we extended the fraud model of RetSim to cover more cases of internal fraud perpetrated by the staff and allow inventory control to flag even more suspicious activity. We also generated sufficient number of runs using a range of fraud parameters to cover a vast number of fraud scenarios that can be studied. We then use RetSim to simulate some of the more common retail fraud scenarios to ascertain exactly the cost of fraud using different fraud parameters for each case.

This thesis introduces a financial simulation model covering two related financial domains: Mobile Payments and Retail Stores systems. The problem we address in these domains is different types of fraud. We limit ourselves to isolated cases of relatively straightforward fraud. However, in this thesis the ultimate aim is to cover more complex types of fraud, such as money laundering, that comprises multiple organisations and domains. Fraud is an important problem that impact the whole economy. Currently, there is a general lack of public research into the detection of fraud. One important reason is the lack of transaction data which is often sensitive. To address this problem we present a Mobile Money Simulator (PaySim) and Retail Store Simulator (RetSim), which allow us to generate synthetic transactional data. These simulations are based on real transaction data. These simulations are multi agent based simulations. Hence, we developed agents that represent the clients in PaySim and customers and salesmen in RetSim. The normal behaviour was based on behaviour observed in data from the field, and is codified in the agents as rules of transactions and interaction between clients, or customers and salesmen. Some of these agents were intentionally designed to act fraudulently, based on observed patterns of real fraud. We introduced known signatures of fraud in our model and simulations to test and evaluate our fraud detection results. The resulting behaviour of the agents generate a synthetic log of all transactions as a result of the simulation. This synthetic data can be used to further advance fraud detection research, without leaking sensitive information about the underlying data. Using statistics and social network analysis (SNA) on real data we could calibrate the relations between staff and customers and generate realistic synthetic data sets that were validated statistically against the original. We then used RetSim to model two common retail fraud scenarios to ascertain exactly how effective the simplest form of statistical threshold detection commonly in use could be. The preliminary results show that threshold detection is effective enough at keeping fraud losses at a set level, that there seems to be little economic room for improved fraud detection techniques.

Managing fraud is important for business, retail and financialalike. One method to manage fraud is by \emph{detection}, wheretransactions etc. are monitored and suspicious behaviour is flaggedfor further investigation. There is currently a lack of publicresearch in this area. The main reason is the sensitive nature of thedata. Publishing real financial transaction data would seriouslycompromise the privacy of both customers, and companies alike. Wepropose to address this problem by building RetSim, a multi-agentbased simulator (MABS) calibrated with real transaction data from oneof the largest shoe retailers in Scandinavia. RetSim allows us togenerate synthetic transactional data that can be publicly shared andstudied without leaking business sensitive information, and stillpreserve the important characteristics of the data.

We then use RetSim to model two common retail fraud scenarios toascertain exactly how effective the simplest form of statisticalthreshold detection could be. The preliminary results of our testedfraud detection method show that the threshold detection is effectiveenough at keeping fraud losses at a set level, that there is littleeconomic room for improved techniques.

Criminals use money laundering to make the proceeds from their illegal activities look legitimate in the eyes of the rest of society. Current countermeasures taken by financial organizations are based on legal requirements and very basic statistical analysis. Machine Learning offers a number of ways to detect anomalous transactions. These methods can be based on supervised and unsupervised learning algorithms that improve the performance of detection of such criminal activity. In this study we present an analysis of the difficulties and considerations of applying machine learning techniques to this problem. We discuss the pros and cons of using synthetic data and problems and advantages inherent in the generation of such a data set. We do this using a case study and suggest an approach based on Multi-Agent Based Simulations (MABS).

Mobile money is a service for performing financial transactions using a mobile phone. By law it has to have protection against money laundering and other types of fraud. Research into fraud detection methods is not as advanced as in other similar fields. However, getting access to real world data is difficult, due to the sensitive nature of financial transactions, and this makes research into detection methods difficult. Thus, we propose an approach based on a Multi-Agent Based Simulation (MABS) for the generation of synthetic transaction data. We present the generation of synthetic data logs of transactions and the use of such a data set for the study of different detection scenarios using machine learning.

We present a social simulation model that covers three main financialservices: Banks, Retail Stores, and Payments systems. Our aim is toaddress the problem of a lack of public data sets for fraud detectionresearch in each of these domains, and provide a variety of fraudscenarios such as money laundering, sales fraud (based on refunds anddiscounts), and credit card fraud. Currently, there is a general lackof public research concerning fraud detection in the financial domainsin general and these three in particular. One reason for this is thesecrecy and sensitivity of the customers data that is needed toperform research. We present PaySim, RetSim, and BankSim asthree case studies of social simulations for financial transactionsusing agent-based modelling. These simulators enable us to generatesynthetic transaction data of normal behaviour of customers, and alsoknown fraudulent behaviour. This synthetic data can be used to furtheradvance fraud detection research, without leaking sensitiveinformation about the underlying data. Using statistics and socialnetwork analysis (SNA) on real data we can calibrate the relationsbetween staff and customers, and generate realistic synthetic datasets. The generated data represents real world scenarios that arefound in the original data with the added benefit that this data canbe shared with other researchers for testing similar detection methodswithout concerns for privacy and other restrictions present when usingthe original data.

The investigation of fraud in business has been a staple for the digital forensics practitioner since the introduction of computers in business. Much of this fraud takes place in the retail industry. When trying to stop losses from insider retail fraud, triage, i.e. the quick identification of sufficiently suspicious behaviour to warrant further investigation, is crucial, given the amount of normal, or insignificant behaviour. It has previously been demonstrated that simple statistical threshold classification is a very successful way to detect fraud~\cite{Lopez-Rojas2015}. However, in order to do triage successfully the thresholds have to be set correctly. Therefore, we present a method based on simulation to aid the user in accomplishing this, by simulating relevant fraud scenarios that are foreseeing as possible and expected, to calculate optimal threshold limits. This method gives the advantage over arbitrary thresholds that it reduces the amount of labour needed on false positives and gives additional information, such as the total cost of a specific modelled fraud behaviour, to set up a proper triage process. With our method we argue that we contribute to the allocation of resources for further investigations by optimizing the thresholds for triage and estimating the possible total cost of fraud. Using this method we manage to keep the losses below a desired percentage of sales, which the manager consider acceptable for keeping the business properly running.

RetSim is an agent-based simulator of a shoe store basedon the transactional data of one of the largest retail shoesellers in Sweden. The aim of RetSim is the generationof synthetic data that can be used for fraud detection re-search. Statistical and a Social Network Analysis (SNA)of relations between staff and customers was used to de-velop and calibrate the model. Our ultimate goal is forRetSim to be usable to model relevant scenarios to gen-erate realistic data sets that can be used by academia, andothers, to develop and reason about fraud detection meth-ods without leaking any sensitive information about theunderlying data. Synthetic data has the added benefit ofbeing easier to acquire, faster and at less cost, for exper-imentation even for those that have access to their owndata. We argue that RetSim generates data that usefullyapproximates the relevant aspects of the real data.

The investigation of fraud in the financial domain has been restricted to those who have access to relevant data. However, customer financial records are protected by law and internal policies, therefore they are not available for most of the researchers in the area of fraud detection. This paper aims to present the work of those researchers who have had access to data and present an interesting approach to fraud detection research; which is the generation of a synthetic data set to work on fraud detection research. Some of the domains covered in this review include mobile money payments, e-payments, retail stores, online bank services and credit card payments. We also cover some of the most relevant surveys in the field and point out the impossibility to compare this work due to the lack of common public data set to test different results.

The lack of legitimate datasets on mobile money transactions to perform research on in the domain of fraud detection is a big problem today in the scientific community. Part of the problem is the intrinsic private nature of financial transactions, that leads to no public available data sets. This will leave the researchers with the burden of first harnessing the dataset before performing the actual research on it. This paper propose an approach to such a problem that we named the PaySim simulator. PaySim is a financial simulator that simulates mobile money transactions based on an original dataset. In this paper, we present a solution to ultimately yield the possibility to simulate mobile money transactions in such a way that they become similar to the original dataset. With technology frameworks such as Agent-Based simulation techniques, and the application of mathematical statistics, we show in this paper that the simulated data can be as prudent as the original dataset for research.