Fraud Decisioning Pulls Ahead In A Tight Race

The cat-and-mouse game between law enforcement and code-abusing felons is entering a new year, and a new phase. The world’s biggest social media platforms are cracking down like never before. The latest iterations of third-party solutions are potent hybrids of machine learning and artificial intelligence (AI) — paired with actual humans — to make the tougher calls. Companies are getting much better at fraud detection and prevention, partly in response to its rapid spread.

Digital fraudsters aren’t taking this lightly. One analysis of more than 1.3 billion transactions found that between July and September 2019, about 20 percent of accounts opened were the result of massive bot attacks, not humans. The robot army marched on eCommerce, financial services, gaming and travel sites mostly — a 70 percent rise in bot-driven registrations in Q3 2019 alone. Then there’s the mobile advertising situation. Brands will have spent roughly $77 billion on in-app ads when 2019 is over, and it’s estimated that phonies will dip out with $26.5 billion of that. Such is “bundle ID spoofing” that makes false apps look real to ad networks.

Then there’s the disturbing rise in loyalty program scams. A leading index of digital theft found that loyalty fraud exploded by 89 percent over 2018, opening a vast new front in the battle.

For their part, the anti-fraud community is hitting back hard. Facebook is going deep into device data like battery charge and GPS coordinates to determine if it’s you or someone else making that purchase. FinTechs and merchants have formed a posse of sorts, with validation solutions provider Service Objects recommending using application programming interfaces (APIs) to verify emails, while retailers such as Costco, Morrisons and Tesco tell customers not to fall for social media notifications asking for personally identifiable information. It’s all in the latest PYMNTS Fraud Decisioning Playbook.

Fighting Fakes With Fire

War against digital fraud uses live ammo or, in some cases, recently live. Fishing fraud (not to be confused with “phishing”) is big business, for example. According to the European Union’s Food Fraud Network, fraudsters love seafood so much that it has seriously interfered with supply chain integrity. What’s an example of fish fraud? Selling chemically treated tuna intended for canning as “fresh” and fit for restaurants is a $220 million a year scandal in the U.S. Into the fray steps IBM to partner with Raw Seafoods of Fall River, Massachusetts, on the blockchain-powered Food Trust mobile app. IBM calls it a “… permissioned, permanent and shared record of food system data.”

Fraudsters like travel even more than seafood, and travel booking site TripAdvisor has had it. The platform’s recent transparency report tells of how TripAdvisor anti-fraud detection stopped roughly 1 million false and misleading reviews from ever being made live. Each interactive makes the TripAdvisor AI smarter, guarding content integrity and preserving trust in the brand.

Meanwhile in China, the Alibaba Anti-Counterfeiting Alliance (AACA) used AI to scan for fake accounts, which in turn led Chinese authorities to shut down a reported 500 knockoff shops.

The common denominator in these far-flung cases is AI and machine learning engineered for rapid decisioning on millions of possible fraud attacks while simultaneously providing a delightfully seamless experience for your customers. Easier said than done.

But it is getting done, with innovative systems that leverage human and artificial intelligence.

Data-First to the Last

Data-first approaches are winning right now, where smart AI scans impossibly large datasets making split-second decisions, while organizing and visualizing the rest for human analysts to ingest — an incredibly important stage that is now getting the attention it deserves. The brave new world of the machines exposing fraudulent activity is surprisingly human after all.

It’s all a moving target. When the FBI bobs, cybercrooks weave, and so on. But with new capabilities like device recognition, augmented analytics and data-lake enrichment, plus the intuition of human analysts, the cats are winning their eternal fight with kleptomaniacal cyber-mice.