AI in Finance: Technologies for fraud detection

Reading Time: 4:30 Minutes

The fintech industry is confronting an increasingly complex landscape of fraud. In the US, fintech firms lose an average of $51 million yearly to fraudulent activities, translating to approximately 1.7% of their total revenue. Given the magnitude of this financial drain, it's unsurprising that many fintech firms regard the cost of fraud as their principal challenge in conducting business.

“One needs to consider the emerging costs from a two-dimensional perspective: Firstly, there are costs arising directly from fraudulent attacks. Secondly, costs may also arise if authorities such as our European Federal Financial Supervisory Authority (BaFin) identifies deficiencies in preventing fraudulent attacks. In such cases, they possess the authority to curtail daily business operations, initiate additional investigations into ongoing cases, and impose limitations on the company's daily functions," says Miriam Wohlfarth, Co-CEO and Founder at Banxware. "The security and integrity of transactions conducted online are not only crucial for us at Banxware. Considering the rapid pace at which AI technologies are being developed, one can also recognize significant opportunities in utilising them to prevent fraudulent activities."

The increase in fraud activities does not only impact businesses but consumers as well. The Federal Trade Commission (FTC) reported that American consumers lost more than $5.8 billion to fraud in 2021, up from $3.4 billion the year before, with a 70% increase in just twelve months.

Furthermore, according to a report by SEON, card fraud incidents experienced a significant surge in 2022, with around 36% of all financial institutions falling victim. This marked a substantial year-on-year increase of 26%, highlighting the growing severity of this issue within the sector. It's also worth noting that the prevalence of fraud in neobanks, or fintech banks, can be almost twice as high as in traditional banks. Specifically, neobanks have an average fraud rate of 0.3% as opposed to 0.15% in more conventional financial institutions.

Europe is unfortunately not exempt from these threats. If it’s true that implementing stricter security measures under the revised payment services directive (PSD2) has led to noteworthy reductions in fraud, fraudsters continue to adapt, exploiting the digitalization of retail payments and the increasing use of mobile phones for day-to-day payments.

Here we will explore how AI technologies can serve as advanced, adaptive tools for fraud detection and mitigate such pervasive threats. 

Common frauds reported by fintech firms

The sophistication of financial fraud has increased substantially over the past decade, with stolen data readily traded by crime syndicates offering "fraud as a service" (FaaS) on the dark web. This challenging landscape has seen a 13% uptick in fintech fraud, marking small US fintech firms, especially those in the buy-now-pay-later (BNPL) space, as the most impacted entities in the financial sector. In 2022, fraud attacks on BNPL platforms have gone up 54% year-over-year. In the same year, just considering identity fraud, the estimated losses amounted to $20 billion.

To further illustrate the problem, the most common frauds reported by fintech firms include:

1. Phishing: fraudsters deceive users into revealing sensitive information, like usernames, passwords, or credit card details, by posing as a trustworthy entity, typically via email or text messages.

2. Account Takeovers (ATO): illegitimate access to genuine user accounts, often using stolen credentials or exploiting vulnerabilities.

3. Presentation Attacks: fraudsters trick biometric systems using fake traits like images or voice recordings. The advent of deepfakes, AI-generated audio or video content, has exacerbated the problem, as fraudsters can convincingly mimic a user's face or voice.

4. Identity Fraud and Theft: fraudsters use stolen or synthetic identities to open accounts, carry out transactions, or otherwise impersonate a legitimate user.

5. Card-Not-Present (CNP) Fraud: transactions made using stolen card information without needing the physical card.

6. Pharming and Hacking: pharming attacks redirect users from legitimate to fraudulent websites, often by exploiting the Domain Name System (DNS). Users unwittingly input their credentials into these fake sites, enabling fraudsters to steal sensitive information. On the other hand, hacking involves unauthorized access to computer systems, potentially leading to significant data breaches or theft.

According to the Global Economic Crime and Fraud Survey 2022 carried out by PwC, 44% of companies from the financial services industry found customer fraud to be their foremost concern.

Artificial Intelligence Technologies for Enhanced Fintech Fraud Detection: A Closer Look

AI technologies, including machine learning and natural language processing, can be crucial in offering robust defenses against escalating fintech fraud. Machine learning excels at adapting to evolving fraud patterns and enhancing detection capabilities. Natural language processing counters phishing by identifying suspicious language patterns in communications.

Furthermore, real-time fraud detection powered by AI is instrumental in scanning vast volumes of data to identify fraudulent activities as they happen. Advanced machine learning algorithms are used to uncover anomalies and suspicious patterns in user behavior, transactions, or data changes that could indicate fraudulent activity.

User behavior analytics, powered by AI, can alert to unusual activities indicative of ATOs. Meanwhile, advanced biometric analysis using deep learning algorithms can differentiate between real and fake traits, mitigating presentation attacks. Additionally, AI's ability to detect unusual network patterns bolsters defenses against pharming and hacking attacks.

Regarding account security, AI-assisted multi-factor authentication (MFA) adds an additional security layer, which could involve the use of biometrics, like fingerprints or voice recognition, in addition to traditional password methods. AI algorithms can help to evaluate behavioral patterns to differentiate between legitimate users and potential imposters.
Meanwhile, in the context of onboarding new customers, AI streamlines the verification process, significantly reducing the risk of identity fraud. Machine learning algorithms cross-check the information provided by new customers against multiple databases to authenticate their identity. AI also proves helpful in detecting synthetic identity fraud, where fraudsters blend real and fictitious information to forge new identities.

It’s also worth to highlight that the use of AI in finance can be extended to traditional banks and financial institutions - in this sense, Cognizant detailed a practical case study demonstrating the efficacy of AI in mitigating fraud risks for banks. In their case, an AI-powered machine learning solution was developed and implemented for a prominent global bank to scrutinize scanned images of handwritten checks. With the solution implemented, they have been able to flag potential fraudulent activity and provide a confidence score in less than 70 milliseconds for each processed check. The forecasted results here are a 50% reduction in fraudulent transactions, resulting in $20 million saved by the banks.

Finance AI: What Lies Ahead

As we look to the future, it is clear that the conversation about AI's role in fraud detection and its prevention is still at an early stage. Still, the potential is already clear, with use cases and actual data available to confirm that AI in finance is not a fading trend, as these technologies are becoming necessary to combat increasingly sophisticated fraudulent activities within the whole financial sector landscape. 

And yes, this will inevitably be one of the key topics discussed at FIBE, the new international fintech festival we look forward to celebrating next April in Berlin! Don’t miss out on more news related to AI technologies in finance and sign up to our newsletter.