U.S.-U.K. PETs Prize Challenge: Anomaly Detection via Privacy-Enhanced Federated Learning.

Privacy Enhancing Technologies (PETs) have the potential to enable collaborative analytics without compromising privacy. This is extremely important for collaborative analytics can allow us to really extract value from the large amounts of data that are collected in domains such as healthcare, finance, and national security, among others. In order to foster innovation and move PETs from the research labs to actual deployment, the U.S. and U.K. governments partnered together in 2021 to propose the PETs prize challenge asking for privacy-enhancing solutions for two of the biggest problems facing us today: financial crime prevention and pandemic response. This article presents the Rutgers ScarletPets privacy-preserving federated learning approach to identify anomalous financial transactions in a payment network system (PNS). This approach utilizes a two-step anomaly detection methodology to solve the problem. In the first step, features are mined based on account-level data and labels, and then a privacy-preserving encoding scheme is used to augment these features to the data held by the PNS. In the second step, the PNS learns a highly accurate classifier from the augmented data. Our proposed approach has two major advantages: 1) there is no noteworthy drop in accuracy between the federated and the centralized setting, and 2) our approach is flexible since the PNS can keep improving its model and features to build a better classifier without imposing any additional computational or privacy burden on the banks. Notably, our solution won the first prize in the US for its privacy, utility, efficiency, and flexibility.

Gas chromatography ion mobility spectrometry (GC-IMS) dataset of honey samples with different botanical origins.

Gas chromatography ion mobility spectrometry (GC-IMS) is a robust and sensitive benchtop technique commonly used for non-target screening of volatile organic compounds. It has been applied to authenticity analysis by generating characteristic "fingerprints" of food samples, well suited for chemometric data analysis. This dataset contains headspace GC-IMS spectra from 50 monofloral honey samples from three different botanical origins, 18 acacia honeys (Robinia pseudoacacia), 19 canola honeys (Brassica napus) and 18 honeydew honeys (forest flowers). Honeys were sourced from the beekeepers directly or obtained from governmental food inspectors from Baden-Wuerttemberg, Germany. Authenticity was confirmed by pollen analysis in the framework of the official control of foodstuffs. The data was acquired using a setup based on an Agilent 6890N gas chromatograph (Agilent Technologies, Palo Alto, CA) and an OEM Standalone IMS cell from G.A.S Sensorsysteme m. b. H. (Dortmund, Germany). All samples were recorded in duplicates and spectra are presented as raw data in the .mea file format. The dataset is available on Mendeley Data: https://data.mendeley.com/datasets/jxj2r45t2x.

Health insurance fraud detection based on multi-channel heterogeneous graph structure learning.

Health insurance fraud is becoming more common and impacting the fairness and sustainability of the health insurance system. Traditional health insurance fraud detection primarily relies on recognizing established data patterns. However, with the ever-expanding and complex nature of health insurance data, it is difficult for these traditional methods to effectively capture evolving fraudulent activity and tactics and keep pace with the constant improvements and innovations of fraudsters. As a result, there is an urgent need for more accurate and flexible analytics to detect potential fraud. To address this, the Multi-channel Heterogeneous Graph Structured Learning-based health insurance fraud detection method (MHGSL) was proposed. MHGSL constructs a graph of health insurance data from various entities, such as patients, departments, and medicines, and employs graph structure learning to extract topological structure, features, and semantic information to construct multiple graphs that reflect the diversity and complexity of the data. We utilize deep learning methods such as heterogeneous graph neural networks and graph convolutional neural networks to combine multi-channel information transfer and feature fusion to detect anomalies in health insurance data. The results of extensive experiments on real health insurance data demonstrate that MHGSL achieves a high level of accuracy in detecting potential fraud, which is better than existing methods, and is able to quickly and accurately identify patients with fraudulent behaviors to avoid loss of health insurance funds. Experiments have shown that multi-channel heterogeneous graph structure learning in MHGSL can be very helpful for health insurance fraud detection. It provides a promising solution for detecting health insurance fraud and improving the fairness and sustainability of the health insurance system. Subsequent research on fraud detection methods can consider semantic information between patients and different types of entities.

Enhancing fraud detection in banking by integration of graph databases with machine learning.

The banking sector's shift from traditional physical locations to digital channels has offered customers unprecedented convenience and increased the risk of fraud for customers and institutions alike. In this study, we discuss the pressing need for robust fraud detection & prevention systems in the context of evolving technological environments. We introduce a graph-based machine learning model that is specifically designed to detect fraudulent activity in various types of banking operations, such as credit card transactions, debit card transactions, and online banking transactions. This model uses advanced methods for anomalies, behaviors, and patterns to analyze past transactions and user behavior almost immediately. We provide an in-depth methodology for evaluating fraud detection systems based on parameters such as Accuracy Recall rate and False positive rate ROC curves. The findings can be used by financial institutions to develop and enhance fraud detection strategies as they demonstrate the effectiveness and reliability of the proposed approach. This study emphasizes the critical role that innovative technologies play in safeguarding the financial sector from the ever-changing strategies of fraudsters while also enhancing banking security.•This paper aims to implement the detection of fraudulent transactions using a state-of-the-art Graph Database approach.•The relational graph of features in the dataset used is modelled using Neo4J as a graph database.•Applying JSON features from the exported graph to various Machine Learning models, giving effective outcomes.

Hyperparameter Optimization with Genetic Algorithms and XGBoost: A Step Forward in Smart Grid Fraud Detection.

This study provides a comprehensive analysis of the combination of Genetic Algorithms (GA) and XGBoost, a well-known machine-learning model. The primary emphasis lies in hyperparameter optimization for fraud detection in smart grid applications. The empirical findings demonstrate a noteworthy enhancement in the model's performance metrics following optimization, particularly emphasizing a substantial increase in accuracy from 0.82 to 0.978. The precision, recall, and AUROC metrics demonstrate a clear improvement, indicating the effectiveness of optimizing the XGBoost model for fraud detection. The findings from our study significantly contribute to the expanding field of smart grid fraud detection. These results emphasize the potential uses of advanced metaheuristic algorithms to optimize complex machine-learning models. This work showcases significant progress in enhancing the accuracy and efficiency of fraud detection systems in smart grids.

Towards generalizable food source identification: An explainable deep learning approach to rice authentication employing stable isotope and elemental marker analysis.

In addressing the generalization issue faced by data-driven methods in food origin traceability, especially when encountering diverse input variable sets, such as elemental contents (C, N, S), stable isotopes (C, N, S, H and O) and 43 elements measured under varying laboratory conditions. We introduce an innovative, versatile deep learning-based framework incorporating explainable analysis, adept at determining feature importance through learned neuron weights. Our proposed framework, validated using three rice sample batches from four Asian countries, totaling 354 instances, exhibited exceptional identification accuracy of up to 97%, surpassing traditional reference methods like decision tree and support vector machine. The adaptable methodological system accommodates various combinations of traceability indicators, facilitating seamless replication and extensive applicability. This groundbreaking solution effectively tackles generalization challenges arising from disparate variable sets across distinct data batches, paving the way for enhanced food origin traceability in real-world applications.

Assessing and Improving Data Integrity in Web-Based Surveys: Comparison of Fraud Detection Systems in a COVID-19 Study.

BACKGROUND: Web-based surveys increase access to study participation and improve opportunities to reach diverse populations. However, web-based surveys are vulnerable to data quality threats, including fraudulent entries from automated bots and duplicative submissions. Widely used proprietary tools to identify fraud offer little transparency about the methods used, effectiveness, or representativeness of resulting data sets. Robust, reproducible, and context-specific methods of accurately detecting fraudulent responses are needed to ensure integrity and maximize the value of web-based survey research. OBJECTIVE: This study aims to describe a multilayered fraud detection system implemented in a large web-based survey about COVID-19 attitudes, beliefs, and behaviors; examine the agreement between this fraud detection system and a proprietary fraud detection system; and compare the resulting study samples from each of the 2 fraud detection methods. METHODS: The PhillyCEAL Common Survey is a cross-sectional web-based survey that remotely enrolled residents ages 13 years and older to assess how the COVID-19 pandemic impacted individuals, neighborhoods, and communities in Philadelphia, Pennsylvania. Two fraud detection methods are described and compared: (1) a multilayer fraud detection strategy developed by the research team that combined automated validation of response data and real-time verification of study entries by study personnel and (2) the proprietary fraud detection system used by the Qualtrics (Qualtrics) survey platform. Descriptive statistics were computed for the full sample and for responses classified as valid by 2 different fraud detection methods, and classification tables were created to assess agreement between the methods. The impact of fraud detection methods on the distribution of vaccine confidence by racial or ethnic group was assessed. RESULTS: Of 7950 completed surveys, our multilayer fraud detection system identified 3228 (40.60%) cases as valid, while the Qualtrics fraud detection system identified 4389 (55.21%) cases as valid. The 2 methods showed only "fair" or "minimal" agreement in their classifications (κ=0.25; 95% CI 0.23-0.27). The choice of fraud detection method impacted the distribution of vaccine confidence by racial or ethnic group. CONCLUSIONS: The selection of a fraud detection method can affect the study's sample composition. The findings of this study, while not conclusive, suggest that a multilayered approach to fraud detection that includes conservative use of automated fraud detection and integration of human review of entries tailored to the study's specific context and its participants may be warranted for future survey research.

Beyond the individual: An improved telecom fraud detection approach based on latent synergy graph learning.

The development of telecom technology not only facilitates social interactions but also inevitably provides the breeding ground for telecom fraud crimes. However, telecom fraud detection is a challenging task as fraudsters tend to commit co-fraud and disguise themselves within the mass of benign ones. Previous approaches work by unearthing differences in calling sequential patterns between independent fraudsters, but they may ignore synergic fraud patterns and oversimplify fraudulent behaviors. Fortunately, graph-like data formed by traceable telecom interaction provides opportunities for graph neural network (GNN)-based telecom fraud detection methods. Therefore, we develop a latent synergy graph (LSG) learning-based telecom fraud detector, named LSG-FD, to model both sequential and interactive fraudulent behaviors. Specifically, LSG-FD introduces (1) a multi-view LSG extractor to reconstruct synergy relationship-oriented graphs from the meta-interaction graph based on second-order proximity assumption; (2) an LSTM-based calling behavior encoder to capture the sequential patterns from the perspective of local individuals; (3) a dual-channel based graph learning module to alleviate the disassortativity issue (caused by the camouflages of fraudsters) by incorporating the dual-channel frequency filters and the learnable controller to adaptively aggregate high- and low-frequency information from their neighbors; (4) an imbalance-resistant model trainer to remedy the graph imbalance issue by developing a label-aware sampler. Experiment results on the telecom fraud dataset and another two widely used fraud datasets have verified the effectiveness of our model.

CTCN: a novel credit card fraud detection method based on Conditional Tabular Generative Adversarial Networks and Temporal Convolutional Network.

Credit card fraud can lead to significant financial losses for both individuals and financial institutions. In this article, we propose a novel method called CTCN, which uses Conditional Tabular Generative Adversarial Networks (CTGAN) and temporal convolutional network (TCN) for credit card fraud detection. Our approach includes an oversampling algorithm that uses CTGAN to balance the dataset, and Neighborhood Cleaning Rule (NCL) to filter out majority class samples that overlap with the minority class. We generate synthetic minority class samples that conform to the original data distribution, resulting in a balanced dataset. We then employ TCN to analyze transaction sequences and capture long-term dependencies between data, revealing potential relationships between transaction sequences, thus achieving accurate credit card fraud detection. Experiments on three public datasets demonstrate that our proposed method outperforms current machine learning and deep learning methods, as measured by recall, F1-Score, and AUC-ROC.

Credit Card Fraud Detection: An Improved Strategy for High Recall Using KNN, LDA, and Linear Regression.

Efficiently and accurately identifying fraudulent credit card transactions has emerged as a significant global concern along with the growth of electronic commerce and the proliferation of Internet of Things (IoT) devices. In this regard, this paper proposes an improved algorithm for highly sensitive credit card fraud detection. Our approach leverages three machine learning models: K-nearest neighbor, linear discriminant analysis, and linear regression. Subsequently, we apply additional conditional statements, such as "IF" and "THEN", and operators, such as ">" and "<", to the results. The features extracted using this proposed strategy achieved a recall of 1.0000, 0.9701, 1.0000, and 0.9362 across the four tested fraud datasets. Consequently, this methodology outperforms other approaches employing single machine learning models in terms of recall.

The fraud loss for selecting the model complexity in fraud detection.

Statistical fraud detection consists in making a system that automatically selects a subset of all cases (insurance claims, financial transactions, etc.) that are the most interesting for further investigation. The reason why such a system is needed is that the total number of cases typically is much higher than one realistically could investigate manually and that fraud tends to be quite rare. Further, the investigator is typically limited to controlling a restricted number k of cases, due to limited resources. The most efficient manner of allocating these resources is then to try selecting the k cases with the highest probability of being fraudulent. The prediction model used for this purpose must normally be regularised to avoid overfitting and consequently bad prediction performance. A loss function, denoted the fraud loss, is proposed for selecting the model complexity via a tuning parameter. A simulation study is performed to find the optimal settings for validation. Further, the performance of the proposed procedure is compared to the most relevant competing procedure, based on the area under the receiver operating characteristic curve (AUC), in a set of simulations, as well as on a credit card default dataset. Choosing the complexity of the model by the fraud loss resulted in either comparable or better results in terms of the fraud loss than choosing it according to the AUC.

A smart secured framework for detecting and averting online recruitment fraud using ensemble machine learning techniques.

With the rise of the Internet and social media, information has become available at our fingertips. However, on the dark side, these advancements have opened doors for fraudsters. Online recruitment fraud (ORF) is one of the problems created by these modern technologies, as hundreds of thousands of applicants are victimized every year globally. Fraudsters advertise bogus jobs on online platforms and target job hunters with fake offerings such as huge salaries and desirable geographical locations. The objective of these fraudsters is to collect personal information to be misused in the future, leading to the loss of applicants' privacy. To prevent such situations, there is a need for an automatic detecting system that can distinguish between real and fake job advertisements and preserve the applicants' privacy. This study attempts to build a smart secured framework for detecting and preventing ORF using ensemble machine learning (ML) techniques. In this regard, four ensemble methods-AdaBoost (AB), Xtreme Gradient Boost (XGB), Voting, and Random Forest (RF)-are used to build a detection framework. The dataset used was pre-processed using several methods for cleaning and denoising in order to achieve better outcomes. The performance evaluation measures of the applied methods were accuracy, precision, sensitivity, F-measure, and ROC curves. According to these measures, AB performed best, followed by XGB, voting, and RF. In the proposed framework, AB achieved a high accuracy of 98.374%, showing its reliability for detecting and preventing ORF. The results of AB were compared to existing methods in the literature validating the reliability of the model to be significantly used for detecting ORF.

Data-Centric AI for Healthcare Fraud Detection.

Automated methods for detecting fraudulent healthcare providers have the potential to save billions of dollars in healthcare costs and improve the overall quality of patient care. This study presents a data-centric approach to improve healthcare fraud classification performance and reliability using Medicare claims data. Publicly available data from the Centers for Medicare & Medicaid Services (CMS) are used to construct nine large-scale labeled data sets for supervised learning. First, we leverage CMS data to curate the 2013-2019 Part B, Part D, and Durable Medical Equipment, Prosthetics, Orthotics, and Supplies (DMEPOS) Medicare fraud classification data sets. We provide a review of each data set and data preparation techniques to create Medicare data sets for supervised learning and we propose an improved data labeling process. Next, we enrich the original Medicare fraud data sets with up to 58 new provider summary features. Finally, we address a common model evaluation pitfall and propose an adjusted cross-validation technique that mitigates target leakage to provide reliable evaluation results. Each data set is evaluated on the Medicare fraud classification task using extreme gradient boosting and random forest learners, multiple complementary performance metrics, and 95% confidence intervals. Results show that the new enriched data sets consistently outperform the original Medicare data sets that are currently used in related works. Our results encourage the data-centric machine learning workflow and provide a strong foundation for data understanding and preparation techniques for machine learning applications in healthcare fraud.

A distributed multiple sample testing for massive data.

When the data are stored in a distributed manner, direct application of traditional hypothesis testing procedures is often prohibitive due to communication costs and privacy concerns. This paper mainly develops and investigates a distributed two-node Kolmogorov-Smirnov hypothesis testing scheme, implemented by the divide-and-conquer strategy. In addition, this paper also provides a distributed fraud detection and a distribution-based classification for multi-node machines based on the proposed hypothesis testing scheme. The distributed fraud detection is to detect which node stores fraud data in multi-node machines and the distribution-based classification is to determine whether the multi-node distributions differ and classify different distributions. These methods can improve the accuracy of statistical inference in a distributed storage architecture. Furthermore, this paper verifies the feasibility of the proposed methods by simulation and real example studies.

Health insurance fraud detection by using an attributed heterogeneous information network with a hierarchical attention mechanism.

BACKGROUND: With the rapid growth of healthcare services, health insurance fraud detection has become an important measure to ensure efficient use of public funds. Traditional fraud detection methods have tended to focus on the attributes of a single visit and have ignored the behavioural relationships of multiple visits by patients. METHODS: We propose a health insurance fraud detection model based on a multilevel attention mechanism that we call MHAMFD. Specifically, we use an attributed heterogeneous information network (AHIN) to model different types of objects and their rich attributes and interactions in a healthcare scenario. MHAMFD selects appropriate neighbour nodes based on the behavioural relationships at different levels of a patient's visit. We also designed a hierarchical attention mechanism to aggregate complex semantic information from the interweaving of different levels of behavioural relationships of patients. This increases the feature representation of objects and makes the model interpretable by identifying the main factors of fraud. RESULTS: Experimental results using real datasets showed that MHAMFD detected health insurance fraud with better accuracy than existing methods. CONCLUSIONS: Experiment suggests that the behavioral relationships between patients' multiple visits can also be of great help to detect health care fraud. Subsequent research fraud detection methods can also take into account the different behavioral relationships between patients.

TFD-IIS-CRMCB: Telecom Fraud Detection for Incomplete Information Systems Based on Correlated Relation and Maximal Consistent Block.

Telecom fraud detection is of great significance in online social networks. Yet the massive, redundant, incomplete, and uncertain network information makes it a challenging task to handle. Hence, this paper mainly uses the correlation of attributes by entropy function to optimize the data quality and then solves the problem of telecommunication fraud detection with incomplete information. First, to filter out redundancy and noise, we propose an attribute reduction algorithm based on max-correlation and max-independence rate (MCIR) to improve data quality. Then, we design a rough-gain anomaly detection algorithm (MCIR-RGAD) using the idea of maximal consistent blocks to deal with missing incomplete data. Finally, the experimental results on authentic telecommunication fraud data and UCI data show that the MCIR-RGAD algorithm provides an effective solution for reducing the computation time, improving the data quality, and processing incomplete data.

Class balancing framework for credit card fraud detection based on clustering and similarity-based selection (SBS).

Credit card fraud is a growing problem nowadays and it has escalated during COVID-19 due to the authorities in many countries requiring people to use cashless transactions. Every year, billions of Euros are lost due to credit card fraud transactions, therefore, fraud detection systems are essential for financial institutions. As the classes' distribution is not equally represented in the credit card dataset, the machine learning trains the model according to the majority class which leads to inaccurate fraud predictions. For that, in this research, we mainly focus on processing unbalanced data by using an under-sampling technique to get more accurate and better results with different machine learning algorithms. We propose a framework that is based on clustering the dataset using fuzzy C-means and selecting similar fraud and normal instances that have the same features, which guarantees the integrity between the data features.

Robust Financial Fraud Alerting System Based in the Cloud Environment.

The digitalisation of finance influenced the emergence of new technological concepts for existing user needs. Financial technology, or fintech, provides improved services for customers and new economic value for businesses. As such, fintech services require on-demand availability on a 24/7 basis. For this reason, they are often deployed in cloud environments that allow connectivity with ubiquitous devices. This allows customers to perform online transactions, which are overseen by the respective financial institutions. However, such cloud-based systems introduce new challenges for information security. On one hand, they represent attractive targets for cyberattacks. On the other, financial frauds can still go unnoticed by the financial institutions in charge. This paper contributes to both challenges by introducing the concept for a cloud-based system architecture for fraud detection and client profiling in the banking domain. Therefore, a systematic risk assessment was conducted in this context, and exploitation probabilities were inferred for multiple attack scenarios. In addition, formal verification was accomplished in order to determine the effects of successful vulnerability exploits. The consequences of such security violations are discussed, and considerations are given for improving the resilience of fintech systems.

A behaviour biometrics dataset for user identification and authentication.

As e-Commerce continues to shift our shopping preference from the physical to online marketplace, we leave behind digital traces of our personally identifiable details. For example, the merchant keeps record of your name and address; the payment processor stores your transaction details including account or card information, and every website you visit stores other information such as your device address and type. Cybercriminals constantly steal and use some of this information to commit identity fraud, ultimately leading to devastating consequences to the victims; but also, to the card issuers and payment processors with whom the financial liability most often lies. To this end, we recognise that data is generally compromised in this digital age, and personal data such as card number, password, personal identification number and account details can be easily stolen and used by someone else. However, there is a plethora of data relating to a person's behaviour biometrics that are almost impossible to steal, such as the way they type on a keyboard, move the cursor, or whether they normally do so via a mouse, touchpad or trackball. This data, commonly called keystroke, mouse and touchscreen dynamics, can be used to create a unique profile for the legitimate card owner, that can be utilised as an additional layer of user authentication during online card payments. Machine learning is a powerful technique for analysing such data to gain knowledge; and has been widely used successfully in many sectors for profiling e.g., genome classification in molecular biology and genetics where predictions are made for one or more forms of biochemical activity along the genome. Similar techniques are applicable in the financial sector to detect anomaly in user keyboard and mouse behaviour when entering card details online, such that they can be used to distinguish between a legitimate and an illegitimate card owner. In this article, a behaviour biometrics (i.e., keystroke and mouse dynamics) dataset, collected from 88 individuals, is presented. The dataset holds a total of 1760 instances categorised into two classes (i.e., legitimate and illegitimate card owners' behaviour). The data was collected to facilitate an academic start-up project (called CyberSignature1) which received funding from Innovate UK, under the Cyber Security Academic Startup Accelerator Programme. The dataset could be helpful to researchers who apply machine learning to develop applications using keystroke and mouse dynamics e.g., in cybersecurity to prevent identity theft. The dataset, entitled 'Behaviour Biometrics Dataset', is freely available on the Mendeley Data repository.

GC-IMS data on the discrimination between geographic origins of olive oils.

Gas chromatography hyphenated with ion mobility spectrometry (GC-IMS) is an emerging benchtop technique for sensitive and selective detection of volatile organic compounds. It is commonly used for non-target screening (NTS) of complex sample materials, such as food products. Resulting spectra are used as "fingerprints" for multivariate chemometric data analysis to extract information. This has been successfully applied in the field of food fraud detection in several published studies. The presented dataset contains GC-IMS measurements of extra virgin olive oil samples from Spain, Italy, and Greece. It allows classification and class modelling to differentiate geographic origins and was used in the associated publication gc-ims-tools, a new Python package for chemometric analysis of GC-IMS data (https://doi.org/10.1016/j.foodchem.2022.133476) as an example to demonstrate the functionality.