Bots have been a consistent thorn in fraud teams’ side for years. But since the advent of generative AI (genAI), what used to be just one more fraud type has become a fraud tsunami. This surge in fraud bot attacks has brought with it:
- A 108% year-over-year increase in credential stuffing to take over accounts1
- A 134% year-over-year increase in carding attacks, where stolen cards are tested1
- New account opening fraud at more than 25% of businesses in the first quarter of 2024
While fraud professionals rush to fight back the onslaught, they’re also reckoning with the ever-evolving threat of genAI. A large factor in fraud bots’ new scalability and strength, genAI was the #1 stress point identified by fraud teams in 2024, and 70% expect it to be a challenge moving forward, according to Experian’s U.S. Identity and Fraud Report.
This fear is well-founded. Fraudsters are wasting no time incorporating genAI into their attack arsenal. GenAI has created a new generation of fraud bot tools that make bot development more accessible and sophisticated. These bots reverse-engineer fraud stacks, testing the limits of their targets’ defenses to find triggers for step-ups and checks, then adapt to avoid setting them off.
How do bot detection solutions fare against this next generation of bots?
The evolution of fraud bots
The earliest fraud bots, which first appeared in the 1990s2, were simple scripts with limited capabilities. Fraudsters soon began using these scripts to execute basic tasks on their behalf — mainly form spam and light data scraping. Fraud teams responded, implementing bot detection solutions that continued to evolve as the threats became more sophisticated.
The evolution of fraud bots was steady — and mostly balanced against fraud-fighting tools — until genAI supercharged it. Today, fraudsters are leveraging genAI’s core ability (analyzing datasets and identifying patterns, then using those patterns to generate solutions) to create bots capable of large-scale attacks with unprecedented sophistication. These genAI-powered fraud bots can analyze onboarding flows to identify step-up triggers, automate attacks at high-volume times, and even conduct “behavior hijacking,” where bots record and replicate the behaviors of real users.
How next-generation fraud bots beat fraud stacks
For years, a tried-and-true tool for fraud bot detection was to look for the non-human giveaways: lightning-fast transition speeds, eerily consistent keystrokes, nonexistent mouse movements, and/or repeated device and network data were all tell-tale signs of a bot. Fraud teams could base their bot detection strategies off of these behavioral red flags.
Stopping today’s next-generation fraud bots isn’t quite as straightforward. Because they were specifically built to mimic human behavior and cycle through device IDs and IP addresses, today’s bots often appear to be normal, human applicants and circumvent many of the barriers that blocked their predecessors. The data the bots are providing is better, too3, fraudsters are using genAI to streamline and scale the creation of synthetic identities.4 By equipping their human-like bots with a bank of high-quality synthetic identities, fraudsters have their most potent, advanced attack avenue to date.
Skirting traditional bot detection with their human-like capabilities, next-generation fraud bots can bombard their targets with massive, often undetected, attacks. In one attack analyzed by NeuroID, a part of Experian, fraud bots made up 31% of a business’s onboarding volume on a single day. That’s nearly one-third of the business’s volume comprised of bots attempting to commit fraud. If the business hadn’t had the right tools in place to separate these bots from genuine users, they wouldn’t have been able to stop the attack until it was too late.
Beating fraud bots with behavioral analytics: The next-generation approach
Next-generation fraud bots pose a unique threat to digital businesses: their data appears legitimate, and they look like a human when they’re interacting with a form. So how do fraud teams differentiate fraud bots from an actual human user?
NeuroID’s product development teams discovered key nuances that separate next-generation bots from humans, and we’ve updated our industry-leading bot detection capabilities to account for them. A big one is mousing patterns: random, erratic cursor movements are part of what makes next-generation bots so eerily human-like, but their movements are still noticeably smoother than a real human’s. Other bot detection solutions (including our V1 signal) wouldn’t flag these advanced cursor movements as bot behavior, but our new signal is designed to identify even the most granular giveaways of a next-generation fraud bot.
Fraud bots will continue to evolve. But so will we. For example, behavioral analytics can identify repeated actions — down to the pixel a cursor lands on — during a bot attack and block out users exhibiting those behaviors. Our behavior was built specifically to combat next-gen challenges with scalable, real-time solutions. This proactive protection against advanced bot behaviors is crucial to preventing larger attacks.
For more on fraud bots’ evolution, download our Emerging Trends in Fraud: Understanding and Combating Next-Gen Bots report.
Sources
1HUMAN Enterprise Bot Fraud Benchmark Report
2Abusix
3 NeuroID
4 Biometric Update
