Quantum computing is transforming fraud detection by analyzing complex transaction data faster and more accurately than ever. Financial institutions are using hybrid quantum-classical systems to reduce false alerts, improve detection rates, and save millions annually. Here's how:
- Intesa Sanpaolo: Achieved a 92% fraud detection rate and reduced false positives by 55%, saving €22 million annually.
- PayPal: Partnered with IBM to improve feature selection using quantum algorithms, addressing challenges like data imbalance.
- HSBC: Collaborated with Quantinuum to enhance anti-money laundering efforts and use quantum systems as a "second opinion" for fraud detection.
These systems combine quantum processors' ability to analyze high-dimensional data with classical systems for tasks like preprocessing. While enterprise-wide adoption is years away, pilot projects show quantum computing's growing role in tackling financial fraud effectively.
Quantum Machine Learning in Fraud Detection [QCT21/22, Seminar #2]
Case Study: Intesa Sanpaolo's Quantum Fraud Detection System
Quantum vs Classical Machine Learning for Fraud Detection Performance Comparison
Implementation of Variational Quantum Computing
In 2020, Intesa Sanpaolo took a bold step into the quantum realm by establishing its Quantum Competence Center. Fast forward to early 2023, and they were already piloting a live production system designed to process real-time transaction data.
The system's architecture is built around two key components: a Quantum Feature Mapping stage and a Variational Quantum Classifier. The Quantum Feature Mapping stage transforms transaction data into a high-dimensional Hilbert space using parameterized rotation gates and entangling operations. Meanwhile, the Variational Quantum Classifier is fine-tuned to identify fraud patterns effectively.
A hybrid training approach powers this system. Classical systems handle preprocessing and optimization tasks, while quantum processors manage the feature mapping and classification processes on NISQ (Noisy Intermediate-Scale Quantum) devices. The system continuously learns and adapts to new fraud patterns, ensuring it stays ahead of emerging threats. These design choices have led to impressive performance gains, as detailed below.
Performance Metrics and Business Impact
This advanced quantum system has delivered tangible business results. With a 92% fraud detection rate and an average risk scoring time of just 38 milliseconds, it has significantly improved operational efficiency. The system generates annual savings of €22 million in fraud prevention and reduces manual review costs by €8.4 million. Additionally, it has cut down false alerts by 2.8 million each year.
The solution also requires 47% fewer training examples, achieving a remarkable return on investment (ROI) within just 18 months. Beyond financial metrics, the system has enhanced customer experience, boosting satisfaction scores by 3.2 points and reducing fraud-related customer service inquiries by 14%.
| Performance Metric | Classical Machine Learning | Variational Quantum Computing (VQC) |
|---|---|---|
| Fraud Detection Rate | 83% | 92% |
| False Positive Rate | 5.1% | 2.3% |
| Risk Scoring Time | Millisecond-level | 38ms (average) |
| Training Efficiency | Baseline | 47% fewer examples required |
| Annual Fraud Savings | Baseline | +€22 Million |
| Manual Review Savings | Baseline | +€8.4 Million |
This case study highlights how quantum computing can revolutionize fraud detection, delivering both financial benefits and improved customer satisfaction. Intesa Sanpaolo's system stands as a testament to the potential of integrating cutting-edge quantum technology into financial operations.
Case Study: PayPal's Collaboration with IBM Quantum Network
Quantum Algorithms in Real-Time Detection
PayPal joined forces with IBM to tackle a persistent challenge in fraud detection: feature selection. This process involves pinpointing key data points to uncover threats like identity theft, social engineering, and data breaches. Traditional methods often fall short when dealing with high-dimensional datasets, but quantum computing steps in by analyzing complex feature spaces more effectively.
The collaboration is built on the IBM Safer Payments platform, which integrates IBM Quantum Computers through the Qiskit software stack. Central to this system are Quantum Support Vector Machines (QSVM), which excel at exploring feature spaces in ways classical methods cannot. The setup uses a hybrid approach: classical computing handles data preparation, while quantum processors take on feature mapping. This partnership highlights how quantum computing can enhance fraud detection, using advanced algorithms to identify threats in real time.
The focus here is on combating high-risk fraud, where criminals deploy sophisticated techniques. Quantum algorithms enable PayPal to detect subtle patterns in transaction data - patterns that traditional machine learning might overlook - especially in cases involving stolen identities or social engineering. These advancements were put to the test in PayPal's proof-of-concept implementation, paving the way for practical applications.
Insights from Proof-of-Concept Implementation
During its proof-of-concept phase, PayPal encountered challenges familiar to many in the financial sector. One major issue was data imbalance - fraudulent transactions make up only a tiny fraction of the total volume, which complicates model training.
Quantum-driven feature selection offers a promising solution by identifying transaction attributes that signal fraud, even when examples are scarce. This capability allows for the development of more accurate and efficient fraud detection models. While PayPal's work in this area is still under research, their emphasis on optimizing features demonstrates a clear step toward leveraging quantum computing for real-world fraud detection. By addressing these challenges, PayPal is inching closer to realizing the potential of quantum technology in financial security.
sbb-itb-5174ba0
Case Study: HSBC's Exploration with Quantinuum
Multi-Stage Research and Use Case Development
In May 2023, HSBC launched a multi-phase quantum initiative in partnership with Quantinuum, focusing on areas like cybersecurity, fraud detection, and natural language processing. This structured approach aimed to explore specific applications in cryptography, anti-money laundering (AML), and explainable AI.
The journey began with cryptographic advancements. HSBC used Quantinuum's Quantum Origin platform, integrating it with hardware security modules (HSMs) to generate highly secure, quantum-hardened keys. These keys, designed to be unpredictable, provided an extra layer of protection for sensitive data. Importantly, this step showcased how quantum technology could seamlessly merge with existing classical systems.
By April 2025, HSBC extended its quantum research to tackle AML challenges. Collaborating with the UK's National Quantum Computing Centre (NQCC), HSBC applied quantum analysis to identify subtle indicators of money laundering. This work is particularly relevant given that fraud costs the UK economy around £2.6 billion annually. Philip Intallura, HSBC's Global Head of Quantum Technologies, highlighted the importance of this partnership:
Our collaboration provides us a great opportunity to access cutting-edge quantum hardware and take our use cases to a truly transformational level.
Another area of focus for HSBC is Quantum Natural Language Processing (QNLP). By encoding word meanings into quantum states, QNLP enables the creation of explainable AI systems, which are essential in regulated financial markets. HSBC utilizes Quantinuum's TKET software platform for tasks like qubit routing and circuit optimization, improving the efficiency of Quantum Machine Learning techniques on current hardware.
In fraud detection, HSBC is taking a measured approach. The bank uses quantum computing as a "second opinion" to review suspicious cases flagged by classical systems. This method demonstrates a realistic understanding of quantum computing's present capabilities while acknowledging its potential to reshape fraud detection in the future.
Key Lessons and Industry Applications
Hybrid Architectures and Domain Knowledge
Insights from real-world pilots highlight the advantages of blending quantum and classical systems. For instance, Intesa Sanpaolo’s success came from cross-functional teams that paired quantum researchers with fraud experts who deeply understood financial crime patterns. This collaboration allowed the design of quantum circuits tailored to detect fraudulent transactions effectively.
Intesa Sanpaolo’s hybrid system, which required fewer training parameters, achieved impressive results: a 55% reduction in false positives, translating to annual savings of $8.4 million. The quantum layer processed transaction data in high-dimensional spaces, uncovering complex non-linear patterns that classical systems often overlook. Meanwhile, classical components efficiently handled tasks like data preprocessing and optimization. These findings align with earlier results from Deloitte Italy’s pilot program.
In fraud detection, precision matters more than simple accuracy. Since fraudulent transactions make up less than 1% of total volumes, a model that flags all transactions as legitimate could appear “accurate” but fail in practice. Intesa Sanpaolo’s system reduced false alerts by 2.8 million annually, underscoring the importance of precision.
New tools are also simplifying quantum implementation. For example, frameworks like CircuitHunt now reduce the time needed to search for suitable quantum architectures from days to hours by filtering circuits based on qubit and parameter constraints. This automation addresses a common challenge, as manual circuit design often yields meaningful results in only about 33% of cases.
Scalability and Future Directions
The success of hybrid models raises essential questions about scaling quantum solutions from pilot projects to enterprise-wide use. Currently, quantum fraud detection systems are limited to production pilots rather than full-scale operations. Intesa Sanpaolo’s system, for instance, focuses on high-value portfolios where even small improvements in accuracy lead to significant financial returns. This targeted strategy makes sense given the current limitations of quantum hardware.
Scaling these systems poses engineering challenges. Quantum processors are highly sensitive to environmental factors like temperature changes, and maintaining qubit stability requires advanced error correction methods. Cloud-based quantum services, such as Amazon Braket and Azure Quantum, provide a practical workaround by offering pay-as-you-go access, avoiding the need for large upfront investments.
Experts predict that enterprise-scale quantum deployments won’t become viable until around 2035 or later. In the interim, quantum-inspired optimization techniques running on classical hardware offer a practical way to improve fraud detection while quantum technology continues to evolve.
Adopting quantum-ready cloud platforms now can smooth the transition to full quantum systems as hardware improves. Federica Marini from Deloitte Italy highlights the importance of building workflows that can integrate quantum components seamlessly as they become available. This approach allows businesses to leverage quantum advancements without waiting for hardware perfection.
While quantum technology is still developing, combining quantum innovation with classical expertise is shaping the future of fraud detection. This strategic blend offers a powerful way to tackle financial crime with increasing sophistication.
Conclusion
The examples from Intesa Sanpaolo, PayPal, and HSBC highlight how quantum computing is already making a difference in fraud detection. These real-world applications show measurable improvements in both detection accuracy and operational efficiency, leading to notable cost savings every year.
At the heart of this success lies the use of hybrid systems - integrating quantum processors with classical computing. Quantum processors shine when it comes to analyzing transaction data in high-dimensional spaces, uncovering subtle fraud patterns that classical systems might miss. Meanwhile, classical components handle tasks like preprocessing and optimization. This combination has proven to be highly effective, as seen in Deloitte's hybrid model, which outperformed traditional systems in efficiency.
These advancements highlight the importance of adopting quantum-ready strategies sooner rather than later. Financial institutions that start building quantum capabilities today will be better positioned to capitalize on future opportunities. According to McKinsey, quantum computing applications in finance could generate $622 billion in value by 2035. For organizations hesitant about large investments, cloud-based quantum platforms like Amazon Braket and IBM Quantum Network provide accessible options to begin experimenting and developing workflows.
Building these capabilities requires collaboration. Cross-functional teams that bring together quantum researchers and fraud detection experts are key. Tools like CircuitHunt, which can cut architecture design time from days to mere hours, are already making quantum implementation more manageable. While enterprise-wide adoption of quantum systems is still a few years away, today’s pilot projects are laying the groundwork for the fraud detection systems of the future.
"Quantum's payoff is less about replacing everything classical and more about unlocking the last, stubborn 5% of problems that have resisted classical methods for decades."
- PYMNTS
FAQs
How does quantum computing enhance fraud detection compared to traditional methods?
Quantum computing is making strides in fraud detection, thanks to its quantum-enhanced data processing and advanced optimization capabilities that surpass traditional systems. A standout example is Intesa Sanpaolo, which implemented a quantum computing pipeline. This system slashed false-positive rates from 5.1% to under 2%, all while managing an impressive 2.7 billion transactions annually with decision-making speeds measured in milliseconds.
Research backs these advancements with compelling results. For instance, a quantum support vector machine (QSVM) outperformed top classical models like Random Forest and XGBoost in accuracy, recall, and false-positive rates. Quantum anomaly detection models also delivered up to 15% higher precision compared to their classical counterparts.
Speed is another area where quantum systems shine. Hybrid quantum models, such as quantum-LSTM architectures, have dramatically cut training times - reducing them from several minutes to under a minute per epoch - all while boosting critical metrics like precision and recall. These developments underscore quantum computing's potential to process complex, high-volume transaction data faster and more effectively than traditional methods.
What challenges do businesses face when integrating quantum computing into fraud detection systems?
Integrating quantum computing into fraud detection systems isn't without its hurdles. A key obstacle lies in hardware limitations. Current quantum processors are restricted by the small number of stable qubits they can maintain, making it tough to scale these systems for analyzing large volumes of transactions. Because of this, many organizations turn to hybrid quantum-classical systems, where quantum computing tackles specific tasks while traditional systems handle the bulk of the workload.
Another significant challenge is noise sensitivity and high error rates. Quantum bits, or qubits, are extremely delicate and can be thrown off by factors such as temperature fluctuations or electromagnetic interference. To combat this, systems require error-correction protocols, which not only add layers of complexity but also drive up costs.
On top of that, transforming existing fraud detection platforms to incorporate quantum computing demands major overhauls in infrastructure, talent acquisition, and workflows. Companies need to redesign their data pipelines, invest in specialized software, and bring in experts with quantum computing knowledge. These adjustments can be expensive and may slow down the rate of adoption. Despite these challenges, the potential of quantum computing in fraud detection remains an exciting frontier.
When will quantum computing be widely adopted for fraud detection in enterprises?
Quantum computing is expected to become a game-changer for fraud detection in enterprises by the mid-2030s. This projection aligns with predictions that fault-tolerant quantum machines will be available around 2035. Early efforts, like Intesa Sanpaolo’s planned deployment in 2025, are setting the stage for industry-wide adoption in the years to come.
As this technology evolves, it promises to deliver faster and more precise identification of fraudulent activities, surpassing the capabilities of traditional methods. However, achieving widespread adoption will hinge on progress in quantum hardware and software, as well as businesses' ability to integrate these advanced systems smoothly into their operations.
