Boosting Fraud Detection in Mobile Payment with Prior Knowledge

During the past decade, cybersecurity threats have drawn everyone’s attention and it’s becoming a national priority in many leading countries. With the development of sophisticated mobile technology, mobile (contactless) payment insecurity, which may cause huge financial losses, is now becoming a serious threat to our daily life. During the holiday season in 2013, China’s most welcome mobile payment system provider - Alipay - lost over 20 GB worth of customer data in a security breach, which affected at least 15 million customers. Even though the company has promised to evaluate the security of the system and to take necessary measures to protect customer’s data, are we still safe with the payment? In this paper, we investigate several security vulnerabilities for Alipay wallet, which may cause individual’s personal data and financial losses. This is due to not only less regulation by authorities but also the failure of enabling secure proximity authentication during mobile payment. By going through these surprising vulnerabilities, we come up with some ideas on how to combat them and show how to enhance the mobile payment security by enabling proximity authentication before monetary transactions.

In Japan, cashless is not yet popular but government and companies are devoted to the development of mobile payment methods. This research collected 241 Japanese users and applied decision trees algorithm. Six types of perceived risks (financial, privacy, performance, psychological, security, and time) were used and the categorized class is intention to use mobile payment (low, medium, and high). We also compared different competitive models to examine the performance, including decision trees, kNN, Naïve Bayes, SVM, and logistic regression and decision trees outperformed among all models. The findings indicated that privacy and performance risks are import to Japanese users. Safe, secured, reliable, and fast mobile payment environment are more important to low intention users (less concerns about financial risk). Financial loss, safe, secured, reliable, and fast mobile payment environment are more important to medium intention users (less concerns about time and security risk). Monetary loss, safe, reliable, and fast mobile payment environment are more important to high intention users (less concerns about security risk and psychological risk). The results can help Japanese companies unlock the perceived risk on mobile payment and furnish appropriate strategies to improve usage.

In this digital world, numerous credit card-based transactions take place all over the world. Concomitantly, gaps in process flows and technology result in many fraudulent transactions. Owing to the spurt in the number of reported fraudulent transactions, customers and credit card service providers incur significant financial and reputation losses respectively. Therefore, building a powerful fraud detection system is paramount. It is noteworthy that fraud detection datasets, by nature, are highly unbalanced. Consequently, almost all of the supervised classifiers, when built on the unbalanced datasets, yield high false negative rates. But, the extant oversampling methods while reducing the false negatives, increase the false positives. In this paper, we propose a novel data oversampling method using Generative Adversarial Network (GAN). We use GAN and its variant to generate synthetic data of fraudulent transactions. To evaluate the effectiveness of the proposed method, we employ machine learning classifiers on the data balanced by GAN. Our proposed GAN-based oversampling method simultaneously achieved high precision, F1-score and dramatic reduction in the count of false positives compared to the state-of-the-art synthetic data generation based oversampling methods such as Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic Sampling (ADASYN) and random oversampling. Moreover, an ablation study involving the oversampling based on the ensemble of SMOTE and GAN/WGAN generated datasets indicated that it is outperformed by the proposed methods in terms of F1 score and false positive count.

This study investigates the impact of cybersecurity incidents on financial losses and user exposure in the global financial sector, analyzing data from 2015 to 2024. The study utilizes machine learning models, specifically Random Forest and XGBoost, to predict the extent of financial damage and the number of affected users based on various cybersecurity incident characteristics, including attack type, source, and resolution time. The findings indicate that XGBoost outperforms Random Forest, with higher predictive accuracy (R² = 0.74) and lower error metrics (RMSE = 14.52, MAE = 6.08). Key features influencing financial loss include Incident Resolution Time, Country, and Year, with Incident Resolution Time emerging as the most significant predictor. Phishing, social engineering, and DDoS attacks were identified as the most financially damaging, emphasizing the need for robust defense mechanisms. Feature importance analysis further revealed that defense mechanisms, particularly VPNs and AI-based detection, play critical roles in mitigating losses. Despite the strong performance of both models, the study highlights challenges such as class imbalance in the data, which affects minority class detection. Recommendations include investing in AI-driven detection systems, enhancing employee awareness, adopting multi-factor authentication, and updating legacy systems. These measures are crucial for reducing financial and reputational damage. The study concludes that machine learning techniques, particularly XGBoost, can significantly improve cybersecurity practices in the financial sector, providing accurate, real-time predictions that enhance risk management and defense strategies.

Bank fraud is an increasingly prevalent issue, causing substantial financial losses for both financial institutions and customers. Also, with the increasing prevalence of bank fraud, especially facilitated by online banking and digital payments, there is a pressing need for potential expertise deployment and advanced fraud detection techniques in real-world banking systems. This study specified the Bayesian Network models for detecting fraudulent bank transactions. Bayesian Networks offer a probabilistic graphical modeling approach that can effectively capture complex relationships and dependencies within financial data. Thus, the research aimed at developing a custom Bayesian network model trained on a large dataset of bank transactions comprising over one million bank transactions to classify instances as fraudulent or non-fraudulent. Python 3.11.10 was used for the graphical representations. Down-sampling was employed to reduce the dataset from its initial size of 1 million observations to 17,650 observations, ensuring a balanced representation of both fraud and non-fraud instances. Various estimation techniques: Maximum Likelihood, Bayesian Networks, and Expectation Maximization were employed and evaluated to learn the model parameters. The model's performance was measured using metrics: accuracy, precision, recall, F1-score, and ROC-AUC. The Bayesian Networks estimator achieved an overall accuracy of 66.18%, precision of 67.73%, F1-score of 64.06%, ROC-AUC of 66.13%, Recall of 60.77%, Sensitivity of 60.77% and Specificity of 71.50%. Also, the Maximum Likelihood achieved an overall Accuracy of 66.77%, Precision of 69.22%, F1-score of 63.96%, ROC-AUC of 66.71%, Recall of 59.45%, Sensitivity of 59.45% and Specificity of 73.97%. Likewise, the Expectation Maximization achieved an overall Accuracy of 66.83%, Precision of 69.31%, F1-score of 64.00%, ROC-AUC of 66.77%, Recall of 59.45%, Sensitivity of 59.45% and Specificity of 74.09%. On the confusion matrix, the model correctly classified 1272 instances as Non-Fraudulent transactions (True Negative). Also, it was observed that the model incorrectly classified 507 instances as fraudulent transactions when they were Non-Fraudulent (False Positive). The model similarly incorrectly classified 687 instances as Non-Fraudulent transactions when they were Fraudulent (False Negative). Finally, it correctly classified 1064 instances as fraudulent transactions (True Positive). These results demonstrated the Bayesian Network's ability to identify fraudulent transactions while minimizing false alarms accurately. The findings highlight the potential of Bayesian Networks as a robust framework for fraud detection in the banking sector, contributing to enhanced security and reduced financial losses. The developed model can be introduced into existing banking systems to strengthen fraud prevention strategies.

Chapter 6 - Security Issues of In-Store Mobile Payment

Background: Fraud in financial transaction is at the root of corruption issues recorded in organization. Detecting fraud practices has become increasingly complex and challenging. As a result, auditors require precise analytical tools for fraud detection. Grouping financial transaction data using K-Means Clustering algorithm can enhance the efficiency of applying Benford Law for optimal fraud detection. Objective: This study aimed to introduce Multiple Benford Law Model for the analysis of data to show potential concealed fraud in the audited organization financial transaction. The data was categorized into low, medium, and high transaction values using K-Means Clustering algorithm. Subsequently, it was reanalyzed through Multiple Benford Law Model in a specialized fraud analysis tool. Methods: In this study, the experimental procedures of Multiple Benford Law Model designed for public sector organizations were applied. The analysis of suspected fraud generated by the toolkit was compared with the actual conditions reported in audit report. The financial transaction dataset was prepared and grouped into three distinct clusters using the Euclidean distance equation. Data in these clusters was analyzed using Benford Law, comparing the frequency of the first digit’s occurrence to the expected frequency based on Benford Law. Significant deviations exceeding ±5% were considered potential areas for further scrutiny in audit. Furthermore, the analysis were validated by cross-referencing the result with the findings presented in the authorized audit organization report. Results: Multiple Benford Law Model developed was incorporated into an audit toolkit to automated calculations based on Benford Law. Furthermore, the datasets were categorized using K-Means Clustering algorithm into three clusters representing low, medium, and high-value transaction data. Results from the application of Benford Law showed a 40.00% potential for fraud detection. However, when using Multiple Benford Law Model and dividing the data into three clusters, fraud detection accuracy increased to 93.33%. The comparative results in audit report indicated a 75.00% consistency with the actual events or facts discovered. Conclusion: The use of Multiple Benford Law Model in audit toolkit substantially improved the accuracy of detecting potential fraud in financial transaction. Validation through audit report showed the conformity between the identified fraud practices and the detected financial transaction. Keywords: Fraud Detection, Benford’s Law, K-Means Clustering, Audit Toolkit, Fraudulent Practices.

Financial fraud has emerged as one of the most critical challenges in the digital economy, with increasingly sophisticated attack strategies threatening the security of online transactions. This study examined the application of artificial intelligence (AI) and machine learning (ML) models for enhancing fraud detection accuracy and operational efficiency. Six algorithms—Logistic Regression, Decision Tree, Random Forest, XGBoost, Stacking Ensemble, and Graph Neural Network (GNN)—were evaluated using a large, imbalanced dataset of financial transactions. Model performance was assessed based on accuracy, precision, recall, F1-score, and AUC-ROC, with results demonstrating the superiority of ensemble-based and graph-oriented methods. The GNN achieved the highest performance, leveraging its ability to model complex relational structures in transactional data. Feature importance analysis via SHAP values indicated that transaction amount, frequency, and previous fraud history were the most influential predictors. Computational cost analysis revealed a trade-off between model complexity and inference latency, with lightweight models offering faster response times but lower detection capability. The findings suggested that integrating explainable AI and ensemble learning could significantly improve fraud detection while ensuring regulatory compliance and operational transparency. This research contributes to both academic literature and practical financial security systems by offering empirical evidence and actionable insights for the deployment of advanced fraud detection technologies.

The paper presents a contactless mobile payment method based on security TF card and NFC technology, with which we can use the security TF card, instead of a special SIM card that support SWP protocol, as a security module to store the user’s important information (e.g. transaction data) in mobile payment. And because security TF card can be distributed by non-telecom operators, the method can obtain a more extensive support by units and institutions other than telecom operators. then it is meaningful to the development of NFC technology.

Nowadays, credit card fraud has increased because of all the traditional payment modes being converted into online payment modes. Individuals and financial organizations face heavy losses every year due to fraud. The customer is very scared because of the scam. Frauds and heavy losses destroy the business of financial institutions; hence, researchers give an overview of who detects fraudulent activities. The second researcher used machine learning algorithms for scam detection. This research focuses on machine-learning algorithms that detect scams and scammers. It reveals that a machine learning approach can predict whether the transaction is a scam. In this research, the machine learning algorithms were applied using the European dataset gathered from Kaggle. The imbalance is because there are many genuine and very few fraudulent transactions. These algorithms are also used to identify genuine and fraudulent transactions. All models have the same accuracy of 99%. However, compared to the other models, a random forest has better recall, precision, and an F1 score in its evaluation metrics. Therefore, a random forest comparatively shows better results than logistic regression and a decision tree.

The study presents an approach to diagnosing mechanical faults in an induction motor drive based on phase current analysis. Experimental data were obtained on a dedicated test bench for the induction motor drive under two operating conditions-no-load and 50% of nominal load-and two technical conditions: healthy, faulty with a “load misalignment” defect, and faulty with a “worn bearing” defect. The preprocessing of current signals included filtering, envelope extraction, and normalization. Statistical features such as root mean square value, kurtosis and skewness coefficients were extracted from the original current waveforms to reflect changes in the waveform shape and distribution under different fault conditions. Additionally, spectral characteristics and integral indicators were calculated from the current envelopes to describe the relative contribution of harmonic components and the fundamental harmonic, allowing for the analysis of signal modulation features associated with various mechanical faults. A classification model,based on the extracted features andusing the Random Forest (RF) algorithm with an average accuracy of 85% under cross-validation,was trained. Feature importance analysis was performed using the SHAP(Shapley Additive exPlanations) method, followed by physical interpretation to enable establishing the relationships between specific features and fault types and confirm the validity of the selected diagnostic indicators. The obtained results confirmed the effectiveness of the classical Random Forest algorithm for diagnosing mechanical faults in induction motor drives, identified the most sensitive features to changes in motor technical condition, and demonstrated the potential for application in condition monitoring and equipment health assessment systems.

Sleep apnea syndrome (SAS) is regarded as one of the most common sleep-related breathing disorders, which can severely affect sleep quality. Since SAS is usually accompanied with the cyclical heart rate variation (HRV), many studies have been conducted on heart rate (HR) to identify it at an earlier stage. While most related work mainly based on clinical devices or signals (e.g., polysomnography (PSG), electrocardiography (ECG)), in this paper we focus on the ballistocardiographic (BCG) signal which is obtained in a non-invasive way. Moreover, as the precision and reliability of BCG signal are not so good as PSG or ECG, we propose a fine-grained feature extraction and analysis approach in SAS recognition. Our analysis takes both the basic HRV features and the breathing effort variation into consideration during different sleep stages rather than the whole night. The breathing effort refers to the mechanical interaction between respiration and BCG signal when SAS events occur, which is independent from autonomous nervous system (ANS) modulations. Specifically, a novel method named STC-Min is presented to extract the breathing effort variation feature. The basic HRV features depict the ANS modulations on HR and Sample Entropy and Detrended Fluctuation Analysis are applied for the evaluations. All the extracted features along with personal factors are fed into the knowledge-based support vector machine (KSVM) classification model, and the prior knowledge is based on dataset distribution and domain knowledge. Experimental results on 42 subjects in 3 nights validate the effectiveness of the methods and features in identifying SAS (90.46% precision rate and 88.89% recall rate).

Fraudulent transactions are becoming an important concern for financial systems because of the increase in digital payments. The imbalance in the transaction data, where fraudulent activities comprise a negligible amount compared to legitimate ones, is an essential obstacle to developing a functional detection model. Traditional classifiers become biased toward the majority class as a result of this imbalance, and as a result, they frequently overlook instances of fraud. The work presented here addresses this by applying the Synthetic Minority Oversampling Technique to a number of traditional machine learning classifiers. To balance the dataset and help the classifier better identify hidden fraud patterns, SMOTE generates synthetic minority samples. Models with machine learning algorithms were run both with and without SMOTE. The results confirm the decrease in false negatives by demonstrating notable improvements in recall and F-measure following the use of SMOTE. The paper finally concludes that machine learning along with SMOTE provides a straightforward yet powerful framework for fraud detection on imbalanced datasets.

Since its inception in 2009, Bitcoin has become and is currently the most successful and widely used cryptocurrency. It introduced blockchain technology, which allows transactions that transfer funds between users to take place online, in an immutable manner. No real-world identities are needed or stored in the blockchain. At the same time, all transactions are publicly available and auditable, making Bitcoin a pseudo-anonymous ledger of transactions. The volume of transactions that are broadcast on a daily basis is considerably large. We propose a set of features that can be extracted from transaction data. Using this, we apply a data processing pipeline to ultimately cluster transactions via a k-means clustering algorithm, according to the transaction properties. Finally, according to these properties, we are able to characterize these clusters and the transactions they include. Our work mainly differentiates from previous studies in that it applies an unsupervised learning method to cluster transactions instead of addresses. Using the novel features we introduce, our work classifies transactions in multiple clusters, while previous studies only attempt binary classification. Results indicate that most transactions fall into a cluster that can be described as common user transactions. Other clusters include transactions made by online exchanges and lending services, those relating to mining activities as well as smaller clusters, one of which contains possibly illicit or fraudulent transactions. We evaluated our results against an online database of addresses that belong to known actors, such as online exchanges, and found that our results generally agree with them, which enhances the validity of our methods.

Credit card fraud is a common occurrence in today’s society because the majority of us use credit cards as a form of payment more frequently. This is the outcome of developments in technology and an increase in online transactions, which have given rise to frauds that have caused significant financial losses. In order to detect fraud in credit card transactions, efficient and effective approaches are needed. In this study, we developed a hybrid CNN‐SVM model for detecting fraud in credit card transactions. The effectiveness of our suggested hybrid CNN‐SVM model for detecting fraud in credit card transactions was tested using real‐world public credit card transaction data. The architecture of our hybrid CNN‐SVM model was developed by replacing the final output layer of the CNN model with an SVM classifier. The first classifier is a fully connected layer with softmax that is trained using an end‐to‐end approach, whereas the second classifier is a support vector machine that is piled on top by deleting the final fully connected and softmax layer. According to experimental results, our hybrid CNN‐SVM model produced classification performances with accuracy, precision, recall, F1‐score, and AUC of 91.08%, 90.50%, 90.34%, 90.41, and 91.05%, respectively.