AiThority.com Featured Post of the Week: Strengthening Fraud Prevention with AI
Business leaders often feel like they’re facing an uphill struggle when it comes to fraud and other financial crime. In Kroll’s recent survey of 400 executives, it was revealed that 69% expected to see a rise in financial crime risks during 2023. Technology has a huge part to play – but it’s both part of the problem and part of the solution. In the same study, 62% of US respondents believed that “evolving technology is going to be the biggest challenge for governments in the fight against financial crime”. But in response, 74% plan technology investments of their own.
Unsurprisingly, artificial intelligence is increasingly part of the story. Fraudsters are already using it for activities such as generating phishing emails and creating deepfake videos and voice recordings.
Thankfully, banks and fintechs also now have many ways to deploy AI to fight back. Let’s looks at five of them.
Spotting Patterns with Machine Learning
Fraud prevention with AI is a perfect application for machine learning. Spotting patterns and trends is something that machines can do far more effectively than humans. In fact, they can outclass humans on speed, accuracy, and (especially) scalability.
Machine learning permits the analysis of vast amounts of data. It allows banks to build models that go far beyond simple rules-based systems (such as automatically blocking transactions over a certain value threshold). It’s literally part of a fraudster’s job to find ways around such rules.
Just as fraudsters constantly adapt their methods, so too must financial institutions if they wish to stay one step ahead. A machine learning model doesn’t stand still: It continues to learn from new activity and new patterns of fraudulent behavior.
Humans cannot do this at scale. They’re only likely to recognize a new fraud trend after significant time has elapsed and plenty of money has already been spirited away. More than half the respondents in a KPMG study said they typically recover less than 25% of their fraud losses. As such, speed is everything in reducing financial loss.
Real-Time Transaction Verification
Another thing that humans cannot do is monitor transactions in real time. With over one billion card transactions alone – every single day – it’s simply not feasible.
According to our analysis, transaction monitoring is something that can be done on a periodic or live basis. The trend is towards a real-time model, allowing suspicious transactions to be flagged or even automatically blocked.
The use of AI facilitates seamless and instantaneous checks of a wide range of data points – such as the user’s location, their digital footprint, their usual patterns of activity, and even the configuration of their browser.
Recommended AI ML Story: AI Sales Assistant Software: Unleashing Sales Superpowers in 2023
From all of this, a risk score is generated that blocks or flags transactions, based on thresholds that can be set by the business and continually refined. Due to their automated nature, such checks can also help to reduce the friction experienced by genuine customers while also decreasing “false positives”.
Similarly, AI can play a part in the admin associated with verifying transactions that may be suspicious. Most banks have moved beyond the days where a suspicious transaction is dealt with via a lengthy phone call with a fraud department. SMS messages, voice alerts and even chatbots can now be involved in checking transactions with account holders.
Voice and Video AI
As covered earlier, voice and video AI has a part to play in financial fraud. It’s well documented that criminals are already actively using deepfakes to fool automated identity checks. The banking industry also has a significant problem with synthetic or “Frankenstein” identities, where fraudsters use a combination of both real and fake personal information to create new identities and open accounts.
The use of AI to defend against such threats is a true cat and mouse game between banks and fraudsters. Many businesses use liveness detection to check that they are dealing with a genuine customer during automated ID checks and logins. Such tests, however, are themselves very vulnerable to deepfake attacks.
In response, the fightback is underway with yet more AI! One project by the University at Buffalo uses light reflections from the corneas to detect deepfakes. The researchers say it is 94% effective.
Biometric Authentication
iPhone users have, perhaps unknowingly, been using AI since 2017. It is at the heart of FaceID, which allows users to unlock their phones just by looking at them.
Banks are increasingly using AI-based biometrics to replace passwords, and even, in some cases, two-factor authentication (2FA) login systems. NatWest in the UK has been a fast mover in this regard, already offering biometric approval via its app and online banking system. Furthermore, it’s also been looking into behavioral biometrics, which focuses on how a customer uses their device(s).
Indeed, no multi-factor authentication (MFA) technology is invulnerable to fraud; but the fact remains that biometric authentication offers benefits in terms of accessibility, ease-of-use, and overall security. It is, after all, not possible to steal a face or fingerprint no matter how successful an attack can be.
Banks have plenty of options in biometric tech, such as fingerprint, retina and iris scanners, along with speech and facial recognition systems.
One way that banks can bolster their biometric security is to combine two types of authentication in tandem. For example, the use of fingerprint readers combined with iris scanners can effectively work as a biometric form of 2FA. Such an approach to MFA makes life very complicated for fraudsters – and AI does much of the legwork in increasing the options. For example, machine learning algorithms are often used to enhance fingerprint reading capabilities, helping the technology to cope with common challenges like scars and skin conditions.
Augmented Intelligence to Help with Investigations
AI isn’t just valuable for preventing fraud: It also has a role in investigations. Once again, this is about scale and pattern analysis. Present an AI system with a known pattern of fraud and it can quickly scan through historical data to see if it has occurred elsewhere. This is not something a human can do quickly.
This use of AI is often described as “augmented intelligence”. Such “AI-enhanced” investigations can be completed more quickly and more accurately. This technology can be used not just by banks, but also by insurance companies. One example is the use of algorithms to flag cases of car insurance fraud.
It’s worth emphasizing that in many of the cases above, AI doesn’t just help to prevent fraud. It also has the potential to enhance the customer experience. When transactions aren’t blocked due to arbitrary thresholds, and legitimate customers aren’t subjected to unnecessarily arduous KYC procedures, they are inevitably more satisfied. This is especially owing to the resultant lack – or even absence – of friction.
Comments are closed.