During the course of past decade, our company has been developing custom solution to address various needs of our clients. None had presented so much challenge as those which dealt with omnipresent problem of online fraud. This biggest obstacle surprisingly turned out to be the disruptive nature of AI. On one hand it enables us, the developers, to create new, innovative solutions, while on the other, it is making the same technology readily available to the fraudsters. A good example is the face recognition capability. When it first became available for commercial integration it was viewed as the "final solution" only to be soon challenged by AI powered morphing algorithms. We have all seen videos on YouTube of famous people delivering speeches and looking themselves only in face. Right now, a teenager can download a free app into a phone, upload a face scan picture of someone else and then shoot a video wearing the "mask". The "similarity" has become so indistinguishable that it would take an AI to tell which is fake and which is true.

Banks have been in a perpetual arms race with the fraudsters. Since most existing fraud detection solutions are tailored for internet and app-based banking systems, they are not suited to detecting fraud in real-time. Online fraud detection must occur in real-time, with constant monitoring. Not just at sign in, but across all transactions. New online channels such as API-based banking services present additional fraud risks and require cutting edge approaches. Regulators require banks to absorb the costs of fraud—for real-time transactions and processes, not the consumer. If regulators find systemic issues in a bank’s Know Your Customer (KYC) processes, they will require a look back. Banks over $10BB in assets are treated like a “big” bank and will be subject to regulatory orders/fines over and above the fraud losses.

Last year, a big shift occurred with the introduction of the European Union’s revised Payment Services Directive (PSD2). Though the PSD2 regulation is set to revolutionize the banking industry, it’s also introduced new fraud risks for financial institutions. These risks raise the question once again of how banks can increase security while still delivering excellent customer experiences.

The growth of risk management and risk-related regulatory compliance technology spending in 2019 is expected to hit $72 billion at 10.1% CAGR. Fraud detection will constitute a significant part of total technology spending in banking and capital markets in North America, Europe, Sub-Sharan Africa and the Asia-Pacific region, shaped by the market and regulatory dynamics. This "arms race" is expected to continue for years to come. Some believe only digitalization of our DNA can stop it.

One of the biggest challenges in fighting fraud for banks is the fraud paradox. This paradox occurs in false positive results when banks try to drastically lower fraud detection thresholds in attempt to identify “probable” fraudulent transactions which leads to increase of false positives. Meantime, banks try to tighten fraud detection models to lower the number of false positives which results in likelihood of missing fraud.

The only mitigating approach in this equation is AI which brings the promise of breaking the tradeoff between false positive and false negative errors. Building such solution lays through integration logic within existing bank’s systems and aggregation logic within data management. All sizes of retail and commercial banks can accelerate transaction monitoring processes and drastically reduce false positives. AI also reduces the chance of missing true positives or compromising the institution's risk profile.

If you are interested in this topic, please contact us and let us know your point of view.