Authorised Push Payment (APP) fraud is skyrocketing, with losses in the UK surpassing more than £1 billion over the last 2 years.  

In the relentless battle against APP fraud, the UK Payment Systems Regulator (PSR) has taken a decisive step forward, releasing a report highlighting the performance of the UK’s 14 largest banking groups around levels of APP fraud and reimbursement of victims. In it, the PSR calls for the financial industry to unite and make greater strides in preventing APP fraud and improving outcomes for victims.  

Data is Key  

To answer this call, data is key. Fraud detection models rely on data to identify and prevent fraudulent transactions, and the quality and diversity of that data is critical. Payment Service Providers (PSPs) each possess valuable data, but this is nowhere near enough to combat evolving fraud tactics, and using real data at scale can be a minefield due to privacy considerations, security issues and technological constraints. The industry urgently needs new solutions supported by broader, more sophisticated data models.  

AI-Simulated Synthetic Data: A Game-Changer  

Generative AI, in the form of AI-simulated synthetic data, has the potential to revolutionise the fight against APP fraud. Synthetic data is traditionally generated from existing data and mimics the properties of the source data it is trained on – this approach has benefits, but also inherits the flaws and constraints of real-world data: the quality, completeness, privacy, security, and other issues mentioned above. AI-simulated synthetic data goes a step further by generating data by simulating the actions and reactions of autonomous "agents" – thousands of fully synthetic people, businesses, and other organisations with different goals, ambitions, and incentives, all interacting over multi-year timelines. This enables the generation of synthetic data with a level of safety that is not possible with more traditional methods of synthetic data creation, while emulating the detail and complexity of real-world data.  

AI-simulated synthetic data offers several advantages for APP fraud detection:  

• It can be used to create broader and more diverse datasets. Traditional methods for generating synthetic data are strictly limited by the availability of real data. AI-simulated synthetic data can generate data for scenarios that occur rarely in the real world, providing a valuable window into fraudsters’ behaviours and tactics. Moreover, new scenarios and novel typologies can be developed and implemented, enabling those tackling fraud to stay one step ahead of the scammer.  

• It protects privacy.  AI-simulated synthetic data is entirely synthetic, so there is zero risk of compromising personal information. This makes it a safe and secure way to train and test fraud detection models and innovative solutions, and to collaborate with others who may not be authorised to access the real data.  

• It empowers advanced fraud detection models.  AI-simulated synthetic data can be used to create rich, multi-industry datasets with complex interdependencies that can be used to develop highly accurate and adaptable fraud detection systems. Referential integrity and ground truth are built into the generation process, so there is no data quality loss or censoring associated with pseudonymisation or other traditional privacy protecting techniques.  

Collaboration and Centralised Learning.  

Tackling APP fraud needs a united front, with financial institutions, regulators, and technology providers collaborating to create a safer banking environment. Synthetic data can support this cooperation by providing consortium-level data to power centralised learning and the creation of industry-led solutions.  

The FCA trusts AI-simulated synthetic data.  

The Financial Conduct Authority (FCA) is using AI-simulated synthetic data to create a safer financial landscape. The FCA hosts Smart Data Foundry’s APP Fraud synthetic dataset on the FCA digital sandbox. This provides a valuable resource for innovation around new fraud detection and prevention solutions. The FCA's usage of synthetic data speaks to the reliability and efficacy of AI-simulated agent-based synthetic data in the fight against APP fraud. The FCA also established a Synthetic Data Expert Group (SDEG) in March 2023 to explore the use of synthetic data in financial markets. “As part of the SDEG, we‘ll create a framework that enables organisations to collaborate on synthetic data use cases, for example by using our Digital Sandbox. The framework is in the early stages of development and will likely focus on some of the identified priority use cases, such as fraud” said the group in its most recent update.  

The fight against APP fraud requires a concerted effort, driven by innovation and collaboration. AI-simulated synthetic data is a powerful tool that can help the financial industry forge a more secure future for its customers.