Listen to the article
Entrust reports a 165 per cent rise in deepfake detections, while injection attacks and social engineering show why spotting synthetic media alone may not stop identity fraud.
Entrust says document fraud remained the largest category of identity attack in its 2027 Identity Fraud Report, even as its rate fell by 19 per cent year on year. The company puts the rate at 2.5 per cent in 2026, the lowest level in its dataset. Yet the figures point to a changing threat: nearly 39 per cent of fraudulent document attempts were classed as medium sophistication, while physical counterfeits made up 61.1 per cent of submissions. Entrust’s analysis draws on identity-verification data from more than 20 industries and 195 countries.
Deepfake detections rose by 165 per cent, according to Entrust. Fraudsters may use AI-generated identification, photographs or video to pose as customers, but creating convincing media is only one part of the process. The material must also get through the verification system. That distinction matters because an attacker may use genuine images or recordings stolen from someone else, rather than synthetic media. Persona’s analysis of injection attacks likewise warns that checking whether an image is artificial will not necessarily reveal whether it was illicitly supplied.
Injection methods can feed video or images directly into verification software, bypassing the device camera. Virtual-camera tools and emulated devices can make fabricated footage appear to have been captured live. Entrust reported a 42 per cent annual rise in fraud identified through device-intelligence signals, which assess the camera, device and session involved. NIST guidance also treats the capture pathway as a security concern, alongside checks for forged media: a real-looking image may still have been delivered through a manipulated channel.
Financial services are a particular draw because a fraudulently opened account can be used to access funds or support further criminal activity. FinCEN has warned that accounts opened using identities suspected of being generated or altered with AI may be used to receive and launder proceeds from other schemes, including through funnel accounts. Identity fraud can also continue after onboarding, making the integrity of an initial document check only one part of the risk. Entrust’s report describes threats across the wider customer relationship.
Fraud does not always depend on counterfeit documents or covert technical methods. Entrust says social engineering can persuade customers to complete checks or approve transactions themselves, while employees may be manipulated into actions that expose their organisations. ID.me has also described how generative AI can make deceptive messages more convincing and help attackers assemble matching combinations of documents, personal information and biometric material. In that kind of operation, the victim’s genuine participation can make an identity check harder to interpret.
The patterns vary by region. Entrust recorded a 462 per cent year-on-year increase in European deepfake detections; physical counterfeits accounted for 64.2 per cent of document fraud there, while digital forgeries made up 26.4 per cent. The Americas had the highest document-fraud rate and the fastest growth in deepfake attacks, with driver’s licences prominent among targeted credentials. Asia-Pacific had the largest proportion of digital forgeries, although most attempts there were rated easy. Entrust’s findings underline how criminals can adapt both the documents they target and the methods they use to different markets.
Source Reference Map
Inspired by headline at: [1]
Sources by paragraph:
Source: Fuse Wire Services


