Aviram Ganor, General Manager EMEA at Riskified
Online forums on the dark web are the modern fraudster’s equivalent to meeting your friends at your local coffee shop. In other words, it’s the ideal place to meet like-minded individuals, share updates and discuss future plans. Real-life comments from these forums are a fascinating insight into the fraudster’s mindset, with perspectives ranging from “We’re a modern day Robin Hood,” to “Fraud is an art form” and “The big corporations have it coming”.
It is these online forums which are helping an ever-growing community to defraud retailers by sharing tips and tricks on personal data theft, social engineering scams, and how to best abuse different merchant policies.
Meanwhile, merchants are simultaneously battling a long economic slump, high competition, and slim margins. To make matters worse, there is a rapidly evolving technology which is making it even easier for bad actors to commit fraud: Artificial Intelligence (AI).
AI has accelerated almost every industry – and fraud is no exception. Generative AI makes it easier to create deepfakes to bolster identity theft attempts, while tools such as WormGPT (think ChatGPT but for fraudsters!) are making it far more accessible to scam retailers and consumers alike.
Thankfully, AI has also advanced the ecommerce space and is helping merchants take back control. As fraudsters level-up their sophistication with AI, merchants who employ AI for fraud prevention stand the best chance of neutralising threats.
Addressing grey areas with AI
Abuse of a merchants’ terms and conditions is a rising trend, given extra fuel by this dark web community. It is also a highly complex area, categorised as policy abuse and referred to as ‘friendly fraud’ as it is often perpetrated by historically ‘good’ customers. Return fraud is one of the most common – for example, a customer wears an item once and returns it (known as ‘wardrobing’) or even replaces the original item with an empty box or a substitute item that weighs the same.
AI models are helping merchants become more dynamic in this area. Running risk assessments in real-time, merchants can set more agile policies that adapt to the individuals. For example, a loyal customer may be rewarded with free and flexible returns, while a customer that has behaved suspiciously or has a bad track-record may be asked to pay a fee.
Throwing out the rule book
Traditional fraud detection methods rely on human-generated rules that determine which transactions should be declined and which are considered legitimate. These rules are cumbersome, inflexible, and often inaccurate. Setting stricter rules for authorising transactions might reduce losses to fraud, but it invariably rejects genuine customers, hurting revenues and reducing customer satisfaction. The lack of flexibility also means it’s harder for merchants to keep pace with changes in fraud – by the time a pattern of abuse is identified, the damage is often done and manual rule updates are null.
By contrast, AI-based fraud detection methods have been transformative. They calculate what ‘normal’ looks like based on patterns of behavior in the merchant data and the algorithm learns to detect anomalies and suspicious behavior without the need to set ‘rules’ beforehand. It can process multiple patterns of risk at once to protect merchants from even more sophisticated fraud attacks, delivering insights in real time.
AI-powered risk management technology also benefits from processing collective network data sets. Sophisticated fraud platforms develop their algorithms using large data sets from multiple retailers and millions of transactions. This identifies trends outside of a merchant’s domain and can help stop bad actors before they reach more unsuspecting retailers.
As a result, more retailers can match the data-driven prowess of giants like Amazon to deliver secure yet seamless transactions for genuine customers.
Taking charge of chargebacks with AI
Managing and disputing chargebacks has historically been a highly resource-intensive and manual process but the need for AI and automation has never been more urgent. Since chargeback rates skyrocketed during the pandemic they have continued to rise – the last year alone three out of four customers in the UK and US filed a chargeback. Merchants know many of these are fraudulent claims but struggle without the resources and tools to dispute them, forcing them to forfeit revenues.
New models are helping merchants automate the process of identifying and categorising claims, as well as processing different data points, making it easier to compile the compelling evidence needed to win a claim. Merchants can analyse unusual behaviour right from the start of session to checkout, processing metadata like IP addresses and cookie deletion to compare with previous orders in the retailer network. Better categorisation and automated analysis also helps inform root cause analysis and prevention strategies.
Fighting fire with fire
The reality is that fraud and policy abuse will never go away. Scammers will always share their secrets, hone their MOs on the dark web and use new technologies to steal your goods.
The good news is that retailers can fight back, and the tools are already at their disposal. The phrase ‘fighting fire with fire’ has a lot of application to AI: using AI is the secret weapon for AI-led fraud. When merchants are proactive and harness AI to its full potential to become both more efficient and identify suspect patterns and behaviours, it transforms the fight entirely.
