Incode Launches “Deepsight” Deepfake Defense, Validated by Purdue Study

Incode Technologies has introduced Deepsight, a real-time AI system designed to detect and block deepfakes, injected virtual cameras, and synthetic identity attacks in under 100 milliseconds. The company positions Deepsight as a core defense layer for enterprises facing AI-generated fraud across onboarding, authentication, workforce access, and age verification.

Deepsight analyzes video, motion, and depth signals to differentiate real users from synthetic media, operating across three main layers: a behavioral layer that looks for bot-like interaction patterns, an integrity layer that validates camera and device authenticity, and a perception layer that spots visual inconsistencies in faces and scenes. Incode says the system can even identify the fingerprint of the generative model used to produce a fake, giving security teams more context on the threat source.

The platform’s performance has been independently evaluated in Purdue University’s study “Fit for Purpose? Deepfake Detection in the Real World,” which benchmarked 24 detection systems from commercial, government, and academic providers. According to the paper and Incode’s summary of the results, Deepsight’s detector achieved the highest accuracy and lowest false acceptance rate among commercial tools and outperformed government and academic models on the real-world political deepfakes used in the benchmark. In internal tests, Incode reports that Deepsight was ten times more accurate than trained human reviewers.

Deepsight is already integrated into the Incode Identity Platform and is being used by organizations including TikTok, Scotiabank, and Nubank, protecting millions of users across more than six million live identity sessions to date. Experian is also partnering with Incode to bring Deepsight into its identity and fraud solutions, citing its own research that 72 percent of business leaders expect AI-generated fraud, including deepfakes, to be a top operational challenge by 2026.

Deepsight is notable on several fronts: it offers an off-the-shelf detector that has been tested against real-world political deepfakes rather than lab-generated content alone; it integrates directly into identity workflows (KYC onboarding, step-up checks, workforce SSO); and it explicitly targets emerging attack vectors such as virtual camera injection, synthetic identities, and AI-driven bot farms. Organizations evaluating their exposure to deepfake-enabled account takeover, social engineering, and synthetic account creation can treat Deepsight as a candidate control for both customer-facing and internal verification journeys, with Purdue’s benchmark providing a relatively rare independent data point on commercial tool performance.

Sources: Business Wire, arXiv, ResearchGate, iProfesional

—

By the ID Tech Editorial Team