Fraud is an inevitable cost of doing business. Organisations are responding to the pervasiveness of fraud by employing increased fraud risk management strategies. Internal audit is the most effective corporate control available to management to address the threat of fraud. Internal audit as an effective corporate control is studied in the context of the mandatory obligation imposed by The Institute of Internal Auditors’ 2009 International Professional Practices Framework (hereafter, IPPF). This mandatory obligation requires the internal audit function to ‘evaluate the potential for the occurrence of fraud and [to evaluate] how the organisation manages fraud risk’ (IPPF performance standard 2120.A2 Risk Management). At the individual auditor level, the internal auditor is required to ‘possess the knowledge, skills, and other competencies to perform their individual responsibilities’ (IPPF attribute standard 1210 Proficiency). These standards have the potential to increase expectations on the internal audit profession to prevent and detect the threat of fraud. This expectations gap raises two research questions: (1) What are the determinants of auditors’ fraud detection capabilities? and (2) What constitutes expert performance in the fraud detection task? This thesis aims to address these research questions through the performance of two studies. Study 1 used interviews to investigate the determinants of fraud detection capabilities of both the external auditor and internal auditor. Study 2 drew on the results from Study 1 and was an online survey of Australian internal audit practitioners. Study 1 confirmed that the determinants of auditor expert performance in other audit tasks established in literature, namely, certification, continuous learning, practical experience, analytical reasoning, data analysis skills, communication skills, are also applicable to the fraud detection task. Further, the fraud detection task requires key, unique capabilities because fraud has an inherent element of deception and concealment by fraud perpetrators. The determinants of auditor expert performance unique to the fraud detection task include mentoring, technical skills, and the ability to work in a team. An additional finding is the identification of an effective control environment as a determinant related to the environment where the auditor performs audit work. The resulting model of auditor expert performance in fraud detection depicts the relationship of these determinants (independent variables) with expert performance in fraud detection (the dependent variable). The inclusion of new determinants and the revision of the definitions of determinants established in literature provided the solution to Research Question 1. Subsequently, auditor expert performance in fraud detection was defined based on the resulting combination of determinants. This was the solution to Research Question 2. Study 2 was an online survey with new scales of measurement that were developed from Study 1 interview data. Expert studies and pilot studies were conducted to validate these new scales of measurement. The online survey captured the perceptions of Australian internal audit practitioners about the determinants identified in Study 1. The survey data was applied to the model of auditor expert performance in fraud detection. The main findings of this study are: (1) the assessment of the effectiveness of the strategies to develop auditors’ knowledge of fraud and fraud detection - mentoring, practical experience, continuous learning, and certification; (2) the assessment of the effectiveness of each determinant in contributing towards auditor expert performance in fraud detection. The findings of this thesis supported the expectation that the uniqueness of the fraud detection task impacted on the composition of auditors’ fraud detection capabilities and subsequently, the composition and definition of auditor expert performance in fraud detection. The participation of practitioners – internal auditors, external auditors, and fraud investigators - and academics in the data collection and validation processes provided valuable insight into the research design and provided helpful data for the two studies. The main contribution of this thesis is the extension of Bonner and Lewis’ (1990) model of auditor expert performance to the fraud detection task. Next, the resulting model of auditor expert performance in fraud detection provides the internal audit profession, organisations, and the individual internal auditor with an understanding of the factors that impact on the individual internal auditor’s fraud detection capabilities. Therefore, this practical understanding of internal auditors’ fraud detection capabilities has the potential to: (1) contribute to the development and improvement of an organisation’s fraud risk management strategy; (2) inform the policy debate regarding the promulgation of professional and mandatory standards; and (3) contribute to auditing practice and the audit profession through the identification of strategies to educate the audit profession about fraud detection. The final contribution is the research design where the qualitative study (Study 1) contributed to the development of the survey instrument and provided insights into the results of the structural mode (Study 2).

Τα φαινόμενα απάτης τείνουν να κυριαρχήσουν τις τελευταίες δεκαετίες σε κάθε τομέα. Ένας τομέας που πλήττεται ιδιαίτερα στις μέρες μας είναι αυτός της υγειονομικής περίθαλψης γενικά και ειδικά της συνταγογράφησης των φαρμάκων. Οι υγειονομικές υπηρεσίες είναι ιδιαίτερα τρωτές στην απάτη και την κατάχρηση. Τόσο οι φορείς κοινωνικής ασφάλισης, όσο και οι ιδιωτικές ασφαλιστικές εταιρείες χάνουν όλο και περισσότερα χρήματα κάθε χρόνο, λόγω ψευδών αιτιών αποζημιώσεων. Το αντικείμενο της παρούσας διατριβής είναι ο σχεδιασμός και η ανάπτυξη μιας μεθοδολογίας ανίχνευσης και πρόληψης της απάτης, που θα μπορεί να εφαρμοστεί στις επιχειρησιακές διεργασίες των υπηρεσιών υγειονομικής περίθαλψης και θα εξασφαλίζει την ελαχιστοποίηση της απώλειας των σχετικών κεφαλαίων. Η ίδια θα είναι σε θέση να ανιχνεύει τα ύποπτα προς απάτη περιστατικά, εξασφαλίζοντας έτσι την ποιότητα και την συνέπεια των παρεχόμενων υπηρεσιών. / Fraud phenomena tend to dominate the last decades. A sector that is particularly affected in our days is that of healthcare domain in general and specifically prescriptions reimbursement. Healthcare services are particularly vulnerable in fraud and abuse. Not only institutions of social insurance, but also private companies lose more money each year, because of false causes of compensations. This thesis intends to illustrate the planning and development of a fraud detection methodology, which is accompanied and supported by a generic fraud ontological framework. This methodology will be able to detect erroneous or suspicious records, ensuring thus the quality and the consequence of provided services.

The use of analytical techniques for fraud detection and the design of fraud detection systems have been topics of several research projects in the past and have seen varying degrees of success in their practical implementation. In particular, several authors regard the use of credit risk scorecards for fraud detection as a useful analytical detection tool. However, research on analytical fraud detection for the South African insurance industry is limited. Furthermore, real world restrictions like the availability and quality of data elements, highly unbalanced datasets, interpretability challenges with complex analytical techniques and the evolving nature of insurance fraud contribute to the on-going challenge of detecting fraud successfully. Insurance organisations face financial instability from a global recession, tighter regulatory requirements and consolidation of the industry, which implore the need for a practical and effective fraud strategy. Given the volumes of structured and unstructured data available in data warehouses of insurance organisations, it would be sensible for an effective fraud strategy to take into account data-driven methods and incorporate analytical techniques into an overall fraud risk assessment system. Having said that, the complexity of the analytical techniques, coupled with the effort required to prepare the data to support it, should be carefully considered as some studies found that less complex algorithms produce equal or better results. Furthermore, an over reliance on analytical models can underestimate the underlying risk, as observed with credit risk at financial institutions during the financial crisis. An attractive property of the structure of the probabilistic weights-of-evidence (WOE) formulation for risk scorecard construction is its ability to handle data issues like missing values, outliers and rare cases. It is also transparent and flexible in allowing the re-adjustment of the bins based on expert knowledge or other business considerations. The approach proposed in the study is to construct fraud risk scorecards at entity level that incorporate sets of intrinsic and relational risk factors to support a robust fraud risk assessment. The study investigates the application of an integrated Suspicious Activity Assessment System (SAAS) empirically using real-world South African insurance data. The first case study uses a data sample of short-term insurance claims data and the second a data sample of life insurance claims data. Both case studies show promising results. The contributions of the study are summarised as follows: The study identified several challenges with the use of an analytical approach to fraud detection within the context of the South African insurance industry. The study proposes the development of fraud risk scorecards based on WOE measures for diagnostic fraud detection, within the context of the South African insurance industry, and the consideration of alternative algorithms to determine split points. To improve the discriminatory performance of the fraud risk scorecards, the study evaluated the use of analytical techniques, such as text mining, to identify risk factors. In order to identify risk factors from large sets of data, the study suggests the careful consideration of both the types of information as well as the types of statistical techniques in a fraud detection system. The types of information refer to the categories of input data available for analysis, translated into risk factors, and the types of statistical techniques refer to the constraints and assumptions of the underlying statistical techniques. In addition, the study advocates the use of an entity-focused approach to fraud detection, given that fraudulent activity typically occurs at an entity or group of entities level. / PhD, Operational Research, North-West University, Vaal Triangle Campus, 2011

The use of analytical techniques for fraud detection and the design of fraud detection systems have been topics of several research projects in the past and have seen varying degrees of success in their practical implementation. In particular, several authors regard the use of credit risk scorecards for fraud detection as a useful analytical detection tool. However, research on analytical fraud detection for the South African insurance industry is limited. Furthermore, real world restrictions like the availability and quality of data elements, highly unbalanced datasets, interpretability challenges with complex analytical techniques and the evolving nature of insurance fraud contribute to the on-going challenge of detecting fraud successfully. Insurance organisations face financial instability from a global recession, tighter regulatory requirements and consolidation of the industry, which implore the need for a practical and effective fraud strategy. Given the volumes of structured and unstructured data available in data warehouses of insurance organisations, it would be sensible for an effective fraud strategy to take into account data-driven methods and incorporate analytical techniques into an overall fraud risk assessment system. Having said that, the complexity of the analytical techniques, coupled with the effort required to prepare the data to support it, should be carefully considered as some studies found that less complex algorithms produce equal or better results. Furthermore, an over reliance on analytical models can underestimate the underlying risk, as observed with credit risk at financial institutions during the financial crisis. An attractive property of the structure of the probabilistic weights-of-evidence (WOE) formulation for risk scorecard construction is its ability to handle data issues like missing values, outliers and rare cases. It is also transparent and flexible in allowing the re-adjustment of the bins based on expert knowledge or other business considerations. The approach proposed in the study is to construct fraud risk scorecards at entity level that incorporate sets of intrinsic and relational risk factors to support a robust fraud risk assessment. The study investigates the application of an integrated Suspicious Activity Assessment System (SAAS) empirically using real-world South African insurance data. The first case study uses a data sample of short-term insurance claims data and the second a data sample of life insurance claims data. Both case studies show promising results. The contributions of the study are summarised as follows: The study identified several challenges with the use of an analytical approach to fraud detection within the context of the South African insurance industry. The study proposes the development of fraud risk scorecards based on WOE measures for diagnostic fraud detection, within the context of the South African insurance industry, and the consideration of alternative algorithms to determine split points. To improve the discriminatory performance of the fraud risk scorecards, the study evaluated the use of analytical techniques, such as text mining, to identify risk factors. In order to identify risk factors from large sets of data, the study suggests the careful consideration of both the types of information as well as the types of statistical techniques in a fraud detection system. The types of information refer to the categories of input data available for analysis, translated into risk factors, and the types of statistical techniques refer to the constraints and assumptions of the underlying statistical techniques. In addition, the study advocates the use of an entity-focused approach to fraud detection, given that fraudulent activity typically occurs at an entity or group of entities level. / PhD, Operational Research, North-West University, Vaal Triangle Campus, 2011

abstract: Using historical data from the third-party payment acquiring industry, I develop a statistical model to predict the probability of fraudulent transactions by the merchants. The model consists of two levels of analysis – the first focuses on fraud detection at the store level, and the second focuses on fraud detection at the merchant level by aggregating store level data to the merchant level for merchants with multiple stores. My purpose is to put the model into business operations, helping to identify fraudulent merchants at the time of transactions and thus mitigate the risk exposure of the payment acquiring businesses. The model developed in this study is distinct from existing fraud detection models in three important aspects. First, it predicts the probability of fraud at the merchant level, as opposed to at the transaction level or by the cardholders. Second, it is developed by applying machine learning algorithms and logistical regressions to all the transaction level and merchant level variables collected from real business operations, rather than relying on the experiences and analytical abilities of business experts as in the development of traditional expert systems. Third, instead of using a small sample, I develop and test the model using a huge sample that consists of over 600,000 merchants and 10 million transactions per month. I conclude this study with a discussion of the model’s possible applications in practice as well as its implications for future research. / Dissertation/Thesis / Doctoral Dissertation Business Administration 2015

Humans have always used sketches to explain the visual world. It is a simple and straight- forward mean to communicate new ideas and designs. Consequently, as in almost every aspect of our modern life, the relatively recent major developments in computer science have highly contributed to enhancing individual sketching experience. The literature of sketch related research has witnessed seminal advancements and a large body of interest- ing work. Following up with this rich literature, this dissertation provides a holistic study on sketches through three proposed novel models including sketch analysis, transfer, and geometric representation. The first part of the dissertation targets sketch authorship recognition and analysis of sketches. It provides answers to the following questions: Are simple strokes unique to the artist or designer who renders them? If so, can this idea be used to identify authorship or to classify artistic drawings? The proposed stroke authorship recognition approach is a novel method that distinguishes the authorship of 2D digitized drawings. This method converts a drawing into a histogram of stroke attributes that is discriminative of authorship. Extensive classification experiments on a large variety of datasets are conducted to validate the ability of the proposed techniques to distinguish unique authorship of artists and designers. The second part of the dissertation is concerned with sketch style transfer from one free- hand drawing to another. The proposed method exploits techniques from multi-disciplinary areas including geometrical modeling and image processing. It consists of two methods of transfer: stroke-style and brush-style transfer. (1) Stroke-style transfer aims to transfer the style of the input sketch at the stroke level to the style encountered in other sketches by other artists. This is done by modifying all the parametric stroke segments in the input, so as to minimize a global stroke-level distance between the input and target styles. (2) Brush-style transfer, on the other hand, focuses on transferring a unique brush look of a line drawing to the input sketch. In this transfer stage, we use an automatically constructed input brush dictionary to infer which sparse set of input brush elements are used at each location of the input sketch. Then, a one-to-one mapping between input and target brush elements is learned by sparsely encoding the target sketch with the input brush dictionary. The last part of the dissertation targets a geometric representation of sketches, which is vital in enabling automatic sketch analysis, synthesis and manipulation. It is based on utilizing the well known convolutional sparse coding (CSC) model. We observe that CSC is closely related to how line sketches are drawn. This process can be approximated as the sparse spatial localization of a number of typical basic strokes, which in turn can be cast as a non-standard CSC model that forms a line drawing from parametric curves. These curves are learned to optimize the fit between the model and a specific set of line drawings. Each part of the dissertation shows the utility of the proposed methods through a variety of experiments, user studies, and proposed applications.

E-commerce has seen a rapid growth the last two decades, making it easy for customers to shop wherever they are. The growth has also led to new kinds of fraudulent activities affecting the customers. To make customers feel safe while shopping online, companies like Resurs Bank are implementing different kinds of fraud filters to freeze transactions that are thought to be fraudulent. The latest type of fraud filter is based on machine learning. While this seems to be a promising technology, data and algorithms need to be tuned properly to the task at hand. This thesis project gives a proof of concept of realizing a machine learning based fraud filter for Resurs Bank. Based on a literature study, available data and explainability requirements, this work opts for a supervised learning approach based on Random Forests with a sliding window to overcome concept drift. The inherent class imbalance of the setting makes the area-under-the-receiver operating-curve a suitable metric. This approach provided promising results that a machine learning based fraud filter can add value to companies like Resurs Bank. An alternative approach on how to incorporate non-numerical features by using recurrent neural networks (RNN) was implemented and compared. The non-numerical feature was transformed by a pre-trained RNN-model to a numerical representation that reflects the features suspiciousness. This new numerical feature was then included in the Random Forest model and the result demonstrated that this approach can add valuable insight to the fraud detection field.

Segundo relatório da auditoria KPMG, 69% das empresas admitiram ser vítimas de algum tipo de fraude. Em 2010, no setor bancário foram perdidos aproximadamente R$ 1,5 bilhões devido às fraudes financeiras cometidas em clientes considerando apenas as fraudes documentais e as perdas com fraudes bancárias eletrônicas superaram os 900 milhões neste mesmo ano. Os tipos de fraudes cometidas foram diversos, dentre eles a fraude durante a abertura de contas, cheques clonados, falsificação de documentos, alterações de códigos de barras e clonagem de cartões. A fraude é um problema frequente nas organizações e bastante discutido no mercado, porém verificou-se a existência de uma lacuna teórica quando se trata de gestão da fraude externa. O objetivo do trabalho foi a estruturação de um quadro conceitual para a Gestão da Fraude Financeira e a sua comparação com a prática.Este é um estudo qualitativo exploratório e foi realizado por meio da análise baseada na Teoria Fundamentada definindo categorias a partir da literatura disponível e a sua comparação com entrevistas feitas em um banco de varejo brasileiro e uma associação de instituições financeiras, além dos artigos jornalísticos. Com relação à utilização dos resultados esta é uma pesquisa aplicada já que seu resultado pode, além de contribuir para a discussão teórica, ser aplicada em qualquer organização interessada em gerir a fraude financeira. Os resultados da elaboração do quadro conceitual mostram que a gestão da fraude financeira externa tem quatro fases: a Contínua, a Prevenção, Detecção e a Reação e as categorias definidas estão inseridas nelas. Quanto à comparação da teoria com a prática, nem todos os aspectos verificados na literatura puderam ser encontrados nos relatos das entrevistas e nos artigos jornalísticos analisados. / According to KPMG audit report, 69% of companies admitted being victims of some kind of fraud. In 2010, the banking sector have lost approximately R$ 1.5 billion due to financial fraud perpetrated on customers considering only documentary fraud and the electronic banking fraud losses exceeded R$ 900 million in the same year. The types of fraud were many, including fraud during account opening, cloned checks, forgery, alteration barcode and card cloning. Fraud is a common problem in organizations and widely discussed in the market, however it was found that there is a theoretical gap when it comes to managing external fraud. The objective of this research was to structure a conceptual framework for the Management of Fraud and its comparison with the practice. This is an exploratory qualitative study and was conducted through analysis based on Grounded Theory defining categories from the available literature and interviews with comparison to a bank and an association of financial institutions, in addition to news articles. Regarding the use of results is an applied research its result can also contribute to the theoretical discussion, and be applied to any organization interested in managing financial fraud. The results of the development of the conceptual framework shows that the management of external financial fraud has four phases: Continuous, Prevention, Detection and Reaction and the defined categories are located in them. Regarding the comparison of theory with practice, not all aspects verified in the literature could be found in the reports of interviews and newspaper articles analyzed.

The thesis focus on fraud management - the detection of fraudulent behavior. In the first part current market situation in finance, telecommunications and healthcare is mapped. Following section contains the review of methods used for data mining which is closely related to detection of fradulent or non-standard behaviour. Part of the thesis is focused on the fraudulent behaviour detection in the procurement area. The whole cycle of deployment and use of the tool for detection is introduced. It covers data collection, identification of suspected cases, outcomes analysis and recommendations how to prevent from fraud.

<p>This thesis investigates fraud and the situation on Internet with e-commerce today, to point on some potential threats and needed countermeasures. The work reviews several state of the art e-fraud schemes, techniques used in the schemes and statistics on the extent of e-fraud. This part shows that e-frauds are today both sophisticated and widespread. </p><p>Since real world frauds are deemed impossible to fully cover in order to predict potential new e-frauds, the thesis adopts a different approach. It suggests two abstraction models for fraud cases, a protocol model and a functional model. These are used to perform analysis on case studies on both telecom frauds and e-frauds. The analysis presents characteristics for both types of frauds. Using one of the abstraction models, the functional model, conceptually similar cases among telecom frauds as well as e-fraud cases are identified. The similar cases in each category are then compared, using the other abstraction model, the protocol model. The study shows that concepts from telecom frauds already exist in e-frauds. </p><p>Several challenges and some possibilities in e-fraud prevention and detection are also extracted in the comparative study of the different categories. The major consequence of the challenges is e-frauds’ higher scalability compared to telecom frauds. </p><p>Finally, this thesis covers several existing countermeasures in e-commerce along with specific countermeasures against auction fraud, phishing and spam. However, it is shown that these countermeasures do not address the challenges in e-fraud prevention and detection to a satisfactory extent. Therefore, this thesis proposes several high-level countermeasures in order to address the challenges.</p>