Insider Brief
- Deloitte Italy partnered with Amazon Braket to explore the power of Quantum Machine Learning (QML) to enhance fraudulent transaction detection.
- Deloitte's quantum-enhanced fraud detection system uses a hybrid quantum neural network.
- A hybrid approach leverages the strengths of both classical and quantum computing.
Deloitte Italy reports on a partnership with Amazon's quantum computing service, Amazon Braket, to explore the power of quantum machine learning (QML) to enhance the detection of fraudulent transactions, an investigation that may offer a glimpse into the future of financial security.
Writing in the AWS Machine Learning blog, the team said that as e-commerce continues to experience explosive growth, there is an increasing need for robust and adaptable fraud detection mechanisms. Traditional machine learning (ML) algorithms have helped analyze vast amounts of transaction data in real time and quickly identify suspicious activity. However, integrating quantum computing could further enhance these capabilities.
Quantum Investments
Quantum computing, which is still in its early stages, has the potential to transform many areas of computing, including finance, the researchers wrote.
Unlike classical computers, which use bits as the smallest unit of information, quantum computers use qubits. These qubits exploit quantum phenomena such as superposition and entanglement and can theoretically process complex calculations at unprecedented speeds. This makes them particularly suitable for optimization problems and simulations that are infeasible on classical computers. These types of calculations are critical to solving many of the challenges facing the financial industry today.
The research team writes that approaching quantum is a marathon, not a sprint.
“Quantum computers have the potential to fundamentally reshape the financial system, enabling much faster and more accurate solutions. In the long term, quantum computers are expected to have advantages over classical computers in the areas of simulation, optimization and ML. Whether quantum computers can provide meaningful speedups to ML is currently a topic of active research,” the report said.
Deloitte's quantum-enhanced fraud detection system employs a hybrid quantum neural network. Built using the Keras library, the neural network incorporates quantum components implemented in PennyLane, a software framework for quantum ML. This hybrid approach leverages the strengths of both classical and quantum computing, providing a path to more accurate and efficient fraud detection, according to the researchers.
Overcoming challenges
Despite its promise, quantum computing faces significant hurdles, some of which the researchers outlined below.
Quantum computers are highly sensitive to external disturbances such as temperature fluctuations, which can introduce noise and errors into computations. Techniques such as parametric compilation, which allows a circuit to be compiled once and reused with new parameters, can help mitigate some of these issues. Amazon Braket automates this process, making quantum computation more reliable.
The team adds that the inherent complexity and probabilistic nature of qubits pose further challenges.
“By utilizing the principles of superposition and entanglement, the design of qubits is highly complex. This complex architecture makes the assessment of their computational power a challenging task, as the multidimensional aspect of qubits requires a more nuanced approach to assess their computational power,” the team writes.
Increased computational errors are also a common problem, but ongoing research into error mitigation and suppression techniques aims to address these concerns, the team acknowledges. Further refinements to these methods are expected to improve the accuracy and reliability of quantum computing, which, if successful, could pave the way for more widespread adoption.
Future outlook
Preliminary results of Deloitte's QML-based fraud detection system are promising, showing superior performance in identifying fraudulent transactions compared to traditional ML methods. As quantum computing technology matures, its role in transforming fraud detection and other critical financial processes will become more evident.
To get ahead of the curve, the research team advises organizations to start incorporating quantum-enabled solutions into their operations. While timelines for when quantum will arrive vary, the researchers say this preparation will better ensure a smooth transition to quantum computing when the hardware becomes commercially viable.
For more technical information than this overview can provide, including details about the data sets and quantum circuitry, please see the post.
A little background on the research team: Federica Marini is Manager of AI & Data at Deloitte Italy and is an expert in AI, Gen AI, ML and Data, providing customized data-driven solutions with a human-centric approach. Matteo Capozi, also at Deloitte Italy, specializes in advanced AI, GenAI models and quantum computing, driving innovation and strategic objectives across industries. Kasi Muthu is a Senior Partner Solutions Architect at AWS in Dallas, focusing on generative AI and data, helping partners and customers accelerate their cloud journey with scalable and resilient solutions. Kuldeep Singh is Principal Global AI/ML Leader at AWS with combined experience in deep AI, ML and cybersecurity knowledge.
