Reality Defender red team enrolls fake faces in Google’s biometric account protection

Biometric liveness checks do not effectively protect Google’s selfie biometrics via video for account enrollment, log-in and recovery against synthetic identities, according to a red team assessment by Reality Defender.

Google says its new feature, launched near the end of July, includes active liveness detection to protect against spoof attacks using fake photos or videos and anti-deepfake protections. It is still in the process of rolling out.

The brief announcement referred to “multiple layers of security to help prevent impersonation attempts” and “standard security practices to detect and help prevent suspicious sign-in attempts.”

A post by Reality Defender AI Researcher and Red Team Specialist Dharva Khambholia explains how a team member used synthetic faces with an eligible Google account owned by Reality Defender.

Two enrollments were attempted by a Reality Defender employee using commercially available software for real-time face-swapping and camera manipulation techniques.  Both succeeded, Khambholia says.

“We did not develop a custom deepfake model, exploit Google infrastructure, or build purpose-made attack software,” Khambholia writes.

Au10tix is integrating Reality Defender’s technology into its identity verification platform amidst the projected 53 percent revenue spike for deepfake detection by 2028, according to “The Deepfake Fraud Detection Market 2026: Securing Identity in the AI Era” from Biometric Update and Goode Intelligence.

The deepfake spoofs were carried out using Microsoft Edge on a Windows 11 laptop running an NVIDIA GPU, and the researcher notes that the success does not necessarily establish vulnerability for any system with different variables.

Reality Defender’s success does not mean that Google Accounts are vulnerable to account takeovers (ATOs) or that every risk signal applied by Google has been defeated. The experiment did not involve stealing an account, or credentials, and the process involved is not used for KYC.

Still, Reality Defender suggests that it shows the necessity of deepfake detection and media manipulation detection software, ongoing adversarial testing and analysis of a wide range of signals to assess integrity and risk.

Biometric liveness detection retains a role, but it is a slightly smaller role, within a larger stack.

Article Topics

biometric liveness detection  |  face biometrics  |  Google  |  Reality Defender  |  spoof detection  |  synthetic faces