[London, March 12, 2024] - Keyless, the leader in privacy-preserving biometrics, has today launched its new injection attack detection capability. This advanced security feature leverages machine learning to identify risky device behavior and analyze biometric inconsistencies, protecting businesses from deepfake-driven attacks. Designed for banks, fintechs, cryptocurrency platforms, and other high-risk industries, this solution is now available to all Keyless customers as generative AI makes producing deepfaked images and videos easier and faster than ever before.
The Growing Threat of Injection Attacks
Injection attacks involve feeding deepfake content directly into a device’s camera to bypass biometric security. As generative AI tools become more sophisticated, fraudsters are increasingly using these attacks to spoof authentication systems. According to a
2025 Forrester report, most senior fraud leaders now believe generative AI has permanently altered the fraud landscape and view AI/ML-based security solutions as essential.
Keyless’ new injection attack detection strengthens its deepfake defense suite, serving as a third line of defense against AI-generated fraud. This system combines system-side resilience—identifying compromised devices with unusual activity—and content-based resilience, which detects biometric anomalies.
Paolo Gasti, Keyless CTO and Co-Founder, commented: “Deepfake technology is evolving very quickly and security companies must keep pace. By identifying compromised devices as well as manipulated biometric data, our injection attack protection offers strong protection against AI-based threats without compromising user privacy.”
Keyless Deepfake Detection Capabilities
Keyless has three lines of defense against deepfake attacks:
Genuine Identity Assurance: Deepfakes can spoof a person’s appearance but not their original device. Keyless requires both a user’s face and their enrolled device for authentication, so without access to the original device, even the most realistic deepfake is ineffective.
Presentation Attack Detection: Presentation attacks—where fake images or videos are shown to a device’s camera—can be countered with passive liveness detection. Stronger than active liveness detection, which deepfakes can mimic, passive liveness typically assesses natural indicators such as light reflection and vein recognition to differentiate real faces from deepfakes.
Injection Attack Detection: Injection attacks, where fraudsters replace a real camera feed with a deepfake or manipulate authentication software, require a more sophisticated approach. Keyless detects compromised devices displaying risky behavior and uses machine learning to analyze biometric inconsistencies.
The Future of Deepfake Defense
Deepfakes are evolving rapidly, and businesses must stay ahead of fraudsters. Keyless remains committed to developing the strongest and most seamless security solutions that protect organizations against continuously evolving AI-driven threats.