IAD procurement, testing and deployment criteria clarified by expert panel

The injection attacks powering a wave of digital fraud can be stopped, but not with liveness detection or deepfake detection, even though they spoof liveness, often with deepfakes.

Procuring and deploying effective defense against biometric injection attacks means building a comprehensive software stack around data capture processes and selecting injection attack detection (IAD) technology that has been through repeated, independent testing, panelists in a webinar hosted this week by Biometric Update agreed.

Alan Goode, CEO and chief analyst at Goode Intelligence, framed the rise of injection attacks as a shift by the fraud industry in part responding to the effectiveness of presentation attack detection (PAD) introduced to thwart biometric spoofing. Other panelists picked up on this point during the presentation as well. Goode forecasts that there will be more than 300 million injection attacks in 2028. This shift, and recognition of it in regulations around the world, are among key drivers identified in the 2026 Injection Attack Detection Market Report and Buyers Guide from Biometric Update and Goode Intelligence.

Mitek Systems Senior Manager, Product Management Becky Kiichle-Gross emphasized the importance of securing the capture channel with a full stack of technologies that covers the user, device, code, session, payload and image.

Organizations looking to implement IAD in their stack cannot simply look up a trusted ranking of how effective they are, however.

Ted Dunstone, CEO of testing organization BixeLab, explained the current state of IAD assurance. A formal standard (ISO/IEC 25456) is still developing, however labs generally now test in alignment with the European standard CEN/TS 18099. This testing provides critical insight to both buyers and developers of IAD.

Innovatrics’ IAD works by binding the physics characteristics measured by smartphones to the pixels in images, says the company’s Enterprise Architect and Injection Attack Prevention Lead Filip Stiglic.

Panelists agreed that educating the market on the difference between deepfake detection, PAD and IAD and how the three elements fit together remains a priority.

Kiichle-Gross pointed out that not all deepfakes are delivered as injection attacks, and Stiglic noted that injection attacks can be carried out with images that are fraudulent but not deepfaked, such as people who are unwittingly filmed or photographed.

This is why the full stack of different software layers is necessary to defend the integrity of face biometrics.

Article Topics

AI fraud  |  biometric liveness detection  |  biometrics  |  BixeLab  |  deepfake detection  |  injection attack detection  |  Innovatrics  |  Mitek