The cost of mounting a sophisticated identity-fraud attack has fallen by more than one hundredfold, according to joint research from Unico and Liminal that examines how generative artificial intelligence and reusable criminal infrastructure are changing fraud economics.
The report combines Unico’s network data with Liminal market research. It defines the sophisticated tier as attacks that combine methods such as deepfakes, injection and physical manipulation, and says those attempts now account for 23.3 percent of classified attacks.
The companies argue that falling production costs let criminals test more identities and institutions. The lower costs support higher-volume attacks. A face swap, synthetic document or automated injection workflow can be reused at scale, while each targeted organization still pays for verification systems, investigations and compliance.
One example in the data linked a single fraud entity to 949 identities and as many as 30 businesses. A person rejected by one institution can appear to another as a first-time applicant when fraud signals remain confined to separate systems.
The research also identifies a gap between confidence and infrastructure. Ninety-three percent of surveyed practitioners said they trust their fraud models, yet 92 percent acknowledged that legacy systems allow signals to slip through. Eighty-five percent reported repeat account-takeover incidents.
Device fingerprinting was ranked by 76 percent of practitioners as the most effective defense against AI-enabled fraud, while 47 percent identified stronger partnerships as a rising investment priority. The report recommends combining identity, device and network intelligence around those controls.
Other identity providers are extending fraud controls throughout the customer lifecycle. AU10TIX has added risk monitoring beyond initial onboarding, while deepfake specialists are being integrated directly into operational fraud platforms.
Unico and Liminal propose a five-stage maturity model for assessing how well an institution connects those controls. They argue that point-in-time identity checks are insufficient when the same actor can move across accounts, channels and businesses.
The companies recommend connecting identity, device and network data across onboarding, authentication and transaction monitoring. Shared risk signals can expose recurring actors who otherwise appear as new applicants to each organization.
Sources: Liminal, Liminal and Unico webinar
–
By the ID Tech Editorial Team
