Independent AI Inspection: Fraunhofer IPA examines Experian's machine learning system

An audit of the Fraunhofer IPA (second step in the graphic) confirms that the ML component for Experian's fraud detection solution has been developed correctly and produces traceable results. © Experian

The Fraunhofer Institute for Production Technology and Automation IPA has conducted an audit of the machine learning component "Transaction Miner" from the information service provider Experian. The Transaction Miner is a core component of Experian's fraud prevention solution AI:drian. The audit confirmed that the development of the Transaction Miner was carried out correctly and in accordance with relevant best practices, and that the model produces comprehensible results.

Machine Learning (ML) is the most widely used method of Artificial Intelligence (AI). ML systems largely learn independently from examples to recognize certain patterns or regularities. However, the causal relationships within a trained model are not always transparent, meaning the logic of how an outcome is derived from the initial information is often hardly visible. Therefore, in public discourse, the metaphor of the black box is frequently used. This also means that ML systems are evaluated not only positively but also critically, which poses an additional hurdle for broader adoption of the technology.

Detailed Analysis of the ML Solution

As with any use of models, especially with ML systems, a responsible and regulation-compliant approach is necessary. Technological advances have led to new possibilities, prompting the EU Commission to work on a draft law to regulate AI deployment. To ensure that customers are safe when using the fraud prevention solution AI:drian, Experian commissioned the Fraunhofer IPA to assess whether the ML component Transaction Miner was developed following all best practices and legal requirements, and whether it delivers comprehensible results. The review was based on the EU draft legislation, a white paper from TÜV Austria, and an AI testing catalog from the Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS, which are guiding standards for future market practices. After a very thorough four-month testing phase, this investigation resulted in a positive outcome.

“We observe that more and more ML systems are being used in areas that the EU Commission has defined as critical, for example in self-driving cars or — as in this case — in verifying consumers' fraud intentions and creditworthiness,” explains Professor Marco Huber, head of the Cyber Cognitive Intelligence (CCI) department at Fraunhofer IPA. “Not all details of the upcoming EU legislation are already known. However, how algorithms should look to produce fair, explainable, and comprehensible results is not trivial but also not magic. The review of Experian's Transaction Miner gave us the opportunity to look under the hood of a product already in practical use and to work on a standardized procedure that can be used once EU legislation becomes legally binding.”

Workload Reduction of Over 95 Percent

The Transaction Miner, audited by Fraunhofer IPA, is the central component of the fraud prevention solution AI:drian, which Experian recently launched in the DACH region. Using the ML algorithms employed, AI:drian is able to detect online fraud attempts more reliably than previous systems. In a pilot with a customer from the e-commerce sector, 99.9 percent of transactions identified by the Transaction Miner as legitimate based on its training were not fraudulent and were accordingly not rejected. Instead of around 7,000 order requests per month, only 300 had to be checked manually — a workload reduction of over 95 percent.

“New developments like Machine Learning are being critically observed by the public,” comments Martina Neumayr, Senior Vice President Credit Risk & Fraud Services at Experian DACH. “There are also examples where ML systems deliver questionable results. With the Transaction Miner, we have developed a particularly powerful ML component that can be used even where the EU Commission sees a strong need for regulation, such as in finance. That is why it was especially important for us to act transparently and have the technology independently audited. We believe that the deployment of new technologies can only gain acceptance if there are independent controls like those by Fraunhofer IPA. It’s even better if this pioneering effort paves the way for a future standardized procedure.”

Diverse Offerings Around "Reliable AI"

Audits like the one described are one of the core offerings of the CCI department at Fraunhofer IPA. One of their research focuses is “Reliable AI.” In this context, IPA experts develop methods for explainability and validation of AI systems. This also includes conducting audits and validating or qualifying AI systems as independent and legally compliant. The services are industry-independent and suitable for companies from startups to large corporations.

About Fraunhofer IPA

The Fraunhofer Institute for Production Technology and Automation IPA, known as Fraunhofer IPA, is one of the largest institutes of the Fraunhofer Society with nearly 1,200 employees. The total budget amounts to €82 million. Organizational and technological tasks related to production are research priorities of the institute. Methods, components, and devices up to complete machines and systems are developed, tested, and implemented. Nineteen specialized departments work interdisciplinarily, coordinated through six business units, mainly with industries such as automotive, machinery and plant engineering, electronics and microsystems technology, energy, medical and biotechnologies, and process industry. Fraunhofer IPA's research is oriented towards the economically sustainable production of personalized products.

About Experian DACH

Experian is a global leader in information and IT services. Experian helps its customers to use data confidently. It supports individuals in gaining control over their financial matters and accessing financial services. For businesses, Experian assists in making smart, growth-promoting decisions. It guides borrowers in responsible credit use. Organizations receive the necessary tools to protect themselves against identity fraud and crime. In the DACH region, Experian offers companies of all sizes and industries, such as e-commerce operators, insurers, financial service providers, and telecommunications companies, the best foundation for risk, fraud, and identity management.