Abstract Financial fraud represents one of the most costly and persistent threats to the global banking sector, with annual losses exceeding $485 billion USD. Machine learning-based fraud detection has become the industry standard, with classical algorithms such as Support Vector Machines (SVM) and Random Forests achieving near-perfect performance on well-structured datasets. As quantum computing hardware matures, the question arises: can variational quantum classifiers (VQCs) approach or exceed classical performance on banking fraud detection, and how does realistic hardware noise affect their viability? In this work, we address this question through a systematic, noise-aware benchmarking study executed on PKTRON v3.7.3, using the PK Falcon 27Q and PK NoisyLab 8Q virtual quantum processors. We generate a 200-sample synthetic banking transaction dataset with six clinically motivated fraud features — transaction amount, time of day, merchant category, transaction velocity, geographic distance, and account age — and evaluate a 4-qubit, 2-layer Variational Quantum Circuit against classical SVM and Random Forest baselines across four experimental phases. Our results reveal that the ideal VQC achieves 88.00% accuracy and 0.6940 Banking Quantum Readiness Score (BQRS), compared to 100.00% and 98.00% for classical SVM and Random Forest respectively. Under noise, amplitude damping at p = 0.10 achieves the highest noisy BQRS of 0.5189, while depolarizing noise is most destructive, collapsing recall and F1 to zero across all tested strengths. Phase damping produces a counter-intuitive pattern of high AUC with zero recall, indicating preserved ranking ability but collapsed decision boundary. We introduce the Banking Quantum Readiness Score (BQRS) — a recall-weighted composite metric (40% recall, 30% F1, 30% AUC-ROC) — as a standardized benchmark for evaluating quantum classifier readiness for banking deployment. Our findings establish the current performance gap between quantum and classical classifiers on fraud detection and define the noise thresholds that must be overcome for practical quantum advantage in financial cybersecurity.
Abstract Financial fraud represents one of the most costly and persistent threats to the global banking sector, with annual losses exceeding $485 billion USD. Machine learning-based fraud detection has become the industry standard, with classical algorithms such as Support Vector Machines (SVM) and Random Forests achieving near-perfect performance on well-structured datasets. As quantum computing hardware matures, the question arises: can variational quantum classifiers (VQCs) approach or exceed classical performance on banking fraud detection, and how does realistic hardware noise affect their viability? In this work, we address this question through a systematic, noise-aware benchmarking study executed on PKTRON v3.7.3, using the PK Falcon 27Q and PK NoisyLab 8Q virtual quantum processors. We generate a 200-sample synthetic banking transaction dataset with six clinically motivated fraud features — transaction amount, time of day, merchant category, transaction velocity, geographic distance, and account age — and evaluate a 4-qubit, 2-layer Variational Quantum Circuit against classical SVM and Random Forest baselines across four experimental phases. Our results reveal that the ideal VQC achieves 88.00% accuracy and 0.6940 Banking Quantum Readiness Score (BQRS), compared to 100.00% and 98.00% for classical SVM and Random Forest respectively. Under noise, amplitude damping at p = 0.10 achieves the highest noisy BQRS of 0.5189, while depolarizing noise is most destructive, collapsing recall and F1 to zero across all tested strengths. Phase damping produces a counter-intuitive pattern of high AUC with zero recall, indicating preserved ranking ability but collapsed decision boundary. We introduce the Banking Quantum Readiness Score (BQRS) — a recall-weighted composite metric (40% recall, 30% F1, 30% AUC-ROC) — as a standardized benchmark for evaluating quantum classifier readiness for banking deployment. Our findings establish the current performance gap between quantum and classical classifiers on fraud detection and define the noise thresholds that must be overcome for practical quantum advantage in financial cybersecurity.
Home banking and digital payments diffusion has greatly increased in recent years. As a result, fraud has also dramatically grown, resulting in the loss of billions of dollars worldwide every year. Therefore, banks and financial institutions are required to offer clients increasingly effective and sophisticated services for illegal transaction detection. Machine learning strategies are commonly employed for this crucial application. However, classical models are not satisfactory enough in highly unbalanced classification tasks like fraud detection. Quantum machine learning, working intrinsically in a higher dimensional computation space thanks to superposition and entanglement, can express more complex models than its classical counterpart and thus could provide a significant advantage in identifying potential illegal transactions. This work aims to analyze the potential of quantum classifiers in the fraud detection context, focusing on the Variational Quantum Circuit (VQC) model. The study has been led by exploiting a dataset based on real transactions provided by Intesa Sanpaolo of 500000 items with 15 features. Considering the limitations of contemporary Noisy Intermediate-Scale Quantum (NISQ) computers and quantum simulators, the dataset has been reduced in the number of transactions and features, exploiting Principal Component Analysis (PCA). The results obtained have been compared on equal terms with those of the most commonly employed classical methods, such as Logistic Regression, Random Forest, XGBoost, Support Vector Machine, and Neural Networks, obtaining a better classification quality in terms of recall. Even though this work is preliminary, the results are encouraging and prove the quantum models’ potential in highly unbalanced classification tasks.
Abstract: Financial fraud detection is a persistent challenge due to the dynamic, adversarial nature of fraudulent activities. This study investigates the potential of Quantum Machine Learning (QML) algorithms for pattern recognition in financial datasets characterized by high dimensionality and imbalance. Quantum Support Vector Machines (QSVM) and Variational Quantum Classifiers (VQC) were trained and tested using the Qiskit platform on a representative fraud dataset. Performance metrics were evaluated against classical benchmarks and analyzed using statistical and regression models. The findings highlight that QML offers meaningful performance enhancements, particularly in F1 score and recall, suggesting its viability for future adoption in financial surveillance systems. Keywords: Quantum Machine Learning, Financial Fraud Detection, Pattern Recognition, Quantum Computing, QSVM, VQC, Predictive Analytics
Credit Card Fraud Detection (CCFD) is a major concern for the financial industry, requiring rapid and accurate identification of fraudulent transactions amidst vast datasets. The rapid evolution of quantum computing presents transformative opportunities for enhancing Machine Learning (ML) applications, particularly in the domain of financial security. This paper explores the implementation of Quantum Machine Learning (QML) techniques for CCFD, a critical challenge in the financial industry characterized by imbalanced datasets. Traditional ML models, while effective, often struggle with scalability and computational efficiency when processing large-scale transactional data. QML leverages quantum phenomenon namely entanglement and superposition to potentially outperform classical algorithms in feature extraction and pattern recognition. Our study proposes a hybrid quantumclassical framework, integrating Variational Quantum Circuits (VQC) with classical Neural Networks (NN) to classify fraudulent transactions. Using a simulated quantum environment, we preprocess a publicly available credit card transaction dataset, encoding features into quantum states for processing through a parameterized quantum circuit. The model is trained to improves its ability to detect fraud accurately with fewer false positives, addressing the critical issue of customer inconvenience.
Financial fraud in loan transactions has emerged as a critical challenge for banking institutions, with global losses exceeding $30 billion annually. Traditional machine learning approaches, while achieving moderate success (85-92% accuracy), struggle with the high-dimensional, non-linear patterns characteristic of sophisticated fraud schemes and face limitations in processing the exponential growth of transaction data. This paper introduces a novel Quantum Machine Learning (QML) framework for detecting fraudulent transactions in loan systems, implementing a Quantum Support Vector Machine (QSVM) within a Python-based web application. Unlike classical SVMs that operate in Hilbert spaces with polynomial complexity, QSVMs leverage quantum feature maps to project data into exponentially larger Hilbert spaces, enabling the identification of complex fraud patterns that remain invisible to classical algorithms. Our architecture integrates Qiskit for quantum circuit execution, scikit-learn for classical preprocessing, and Flask for web deployment, creating an end-to-end fraud detection pipeline accessible through a responsive web interface. The system processes transaction datasets containing 1.2 million labeled samples (fraudulent and legitimate) from real-world loan applications, extracting 47 distinct features including transaction amounts, temporal patterns, geolocation data, device fingerprints, and behavioral biometrics. Experimental results demonstrate that the QSVM implementation achieves 98.3% detection accuracy with 1.7% false positive rate, outperforming classical SVM (92.1%), Random Forest (93.5%), and XGBoost (94.2%) by significant margins. The quantum advantage becomes particularly pronounced for complex fraud patterns (improvement of 8-12%) and high-dimensional feature spaces (>30 dimensions). The web application processes transactions in real-time (average latency 340ms), making it suitable for production deployment. This work represents the first production-ready integration of quantum machine learning for loan fraud detection, demonstrating practical quantum advantage on near-term quantum hardware.
The increasing number of fraudulent transactions and spoofed URLs is a major issue of security in financial systems online. The traditional machine learning models are unable to address the evolving trends of cybercriminals. The present paper will be a comparative analysis of the Quantum Support Vector Classifiers (QSVC), Variational Quantum Classifiers (VQC), and Quantum Neural Networks (QNN) models in detecting fraud in three datasets, malicious URLs, credit card transactions, and phishing emails. We trained the QML models on the individual datasets and tested them based on the accuracy, precision, and the F1-score metrics, and the confusion matrices of the models were obtained. We discovered that, QSVC is highly accurate in classifying benign URLs, but it performs poorly with fraud cases, compared to VQC which is very balanced and has room to optimize. QNN was consistent in learning but needed more complicated circuit architectures to reduce misclassifications. We have found that QML models are promising in the detection of fraud, and more studies are required in the area of ensemble, extended training, and feature engineering. The interactive predictions showed that there were different decision boundaries in the models and hence taking the strengths of both models and combining them.
The increasing complexity and volume of financial transactions pose significant challenges to traditional fraud detection systems. This technical report investigates and compares the efficacy of classical, quantum, and quantum-hybrid machine learning models for the binary classification of fraudulent financial activities. As of our methodology, first, we develop a comprehensive behavioural feature engineering framework to transform raw transactional data into a rich, descriptive feature set. Second, we implement and evaluate a range of models on the IBM Anti-Money Laundering (AML) dataset. The classical baseline models include Logistic Regression, Decision Tree, Random Forest, and XGBoost. These are compared against three hybrid classic quantum algorithms architectures: a Quantum Support Vector Machine (QSVM), a Variational Quantum Classifier (VQC), and a Hybrid Quantum Neural Network (HQNN). Furthermore, we propose Fraud Detection for Quantum Computing (FD4QC), a practical, API-driven system architecture designed for real-world deployment, featuring a classical-first, quantum-enhanced philosophy with robust fallback mechanisms. Our results demonstrate that classical tree-based models, particularly \textit{Random Forest}, significantly outperform the quantum counterparts in the current setup, achieving high accuracy (97.34%) and F-measure (86.95%). Among the quantum models, \textbf{QSVM} shows the most promise, delivering high precision (77.15%) and a low false-positive rate (1.36%), albeit with lower recall and significant computational overhead. This report provides a benchmark for a real-world financial application, highlights the current limitations of quantum machine learning in this domain, and outlines promising directions for future research.