Incode announced the launch of Deepsight, a breakthrough defense that detects and blocks deepfakes, injected virtual cameras, and synthetic identity attacks with unmatched accuracy. Validated by Purdue University, this breakthrough AI system protects enterprises from deepfakes, synthetic identities, and AI-generated fraud at scale.
The Threat of Deepfakes
Due to the rapid evolution of technology and AI, new AI-driven fraud tools and tactics are constantly emerging. Each one is designed to exploit weaknesses in biometric and identity verification systems to bypass even advanced identity verification processes.
The threat deepfake technology poses to cybersecurity continues to grow, with hyper-realistic spoofing attempts and deepfake scams increasingly outsmarting traditional liveness checks. Incode’s 2025 Customer Survey found that 96.4% of fintech professionals consider deepfake and synthetic identity fraud to be a top-of-mind concern. Additionally, our customers have experienced a 7x increase in deepfake-driven impersonations over the last 2 years.
Meanwhile, the use of virtual cameras, emulated devices, and automated bots to create and inject manipulated content at scale is helping cybercriminals to bypass basic liveness detection tools, and increasing the success rates of fraud. Deloitte’s Center for Financial Services predicts that generative AI could enable fraud losses to reach US$40 billion in the United States by 2027, up from US$12.3 billion in 2023.
With fraudsters often being the first to instrumentalize new technology for their purposes, staying ahead of threats requires continuous innovation to protect against the bleeding edge of technology. Our mission to fight fraud and protect trust has led us to launch our most ambitious product to date. Designed to combat fraud in the era of AI-generated IDs and deepfake injections, we’re excited to introduce Incode Deepsight.
What is Incode Deepsight?
Incode Deepsight is the world’s most accurate deepfake detection system for identity verification. It uses advanced AI to detect the most sophisticated AI-powered fraud, such as hyper-realistic deepfakes, virtual cameras, tampered devices, and suspicious user behaviors.
Deepsight protects the entire identity verification process with deepfake fraud protection and multi-layer deepfake detection technology, blocking AI-driven impersonation, camera injections, and device tampering with unmatched accuracy validated by independent academic benchmarks and rapid enterprise adoption.
Incode Deepsight confirms the presence of a real live person, a real camera, an untampered device, and natural behavior. Users feel no added friction because it works quietly in the background without extra steps or longer waits. It leverages AI and ML to analyze minute, subtle details such as image depth and camera movements to ensure no fraud signal goes undetected.
Incode Deepsight identifies potential risk signals and threats that would remain unflagged in isolated detection methods.
Independent Tests by Purdue University
In an independent evaluation led by Purdue University, Deepsight was benchmarked against leading commercial, academic, and government systems. The results show a clear gap in performance, especially where it matters most for real-world identity verification:
Purdue University independently validated Deepsight as the most accurate deepfake detector:
- 2.5× lower false-acceptance rate across all deepfake samples.
- Outperformed 24 government, academic, and commercial detection systems.
In practice, this means fewer good users being blocked, more fraudulent sessions being caught, and a much stronger foundation of trust for every identity verification flow.
What is a Multi-Layered Deepfake Detection System?
Deepsight is a multi-layered fraud prevention and deepfake detection solution. It defends the identity verification process by looking at multiple fraud attack angles at the same time, holistically, to identify whether a session is likely fraudulent.
The opposite of a holistic solution is a spot solution, a single check that gives an isolated result, leaving it to manual or static rules to weigh their relevance to make a decision. In the worst case, this could mean that even though signs of a deepfake are present, a basic liveness check itself may be fooled and does not cause an alarm, and the fraudster is falsely approved.
In contrast, a holistic solution like Deepsight flags the presence of deepfake-relevant risk signals and incorporates them in its final pass or fail result. This holistic approach significantly increases the effectiveness of fraud detection.
Multi-Layered Fraud. How Fraudsters Use Deepfakes
AI-powered fraud targets different steps of the identity verification process to improve the fraudsters’ chance of success in bypassing security measures.
Here is an example of how fraud would be carried out with a deepfake: the fraudster prepares a rooted or jailbroken device and installs unauthorized applications that can hijack the device’s camera during the use of other applications.
This is called an injection attack and is a fundamental step in hacking the verification process and inserting a deepfake in lieu of live camera footage.
If an injection goes undetected, whether the attack fails or succeeds depends entirely on whether the detection models in the verification step can detect a deepfake.
As the quality of deepfakes is advancing rapidly, and because injected video can also consist of pre-recorded footage of live humans that weren’t actually present during verification, these complex multi-angle fraud attacks are among the most concerning fraud trends in 2026.
Incode Deepsight detects different fraud signals in one holistic AI-powered fraud prevention suite, cross-validating trust checks and advanced biometric checks biometric, behavioral, and device signals in real time.
How Does Incode Deepsight Work?
Incode Deepsight is a multi-layered deepfake detection system and fraud prevention tool that blocks fraud across multiple key attack points.
By analyzing the behavioral, device and camera integrity, and perception layers in real time, it ensures only real users are verified – and AI-driven impersonation fraud is stopped in its tracks.
- Perception Layer
Using a world-class, multi-modal AI that examines thousands of data points across multiple frames, motion, and depth data to detect deepfakes and physical spoofs in the captured selfie. It also protects ID captures from AI-generated content by running its own proprietary machine-learning model that was trained on thousands of AI-generated ID cards. As a result, Deepsight is able to accurately detect >99.99% of deepfakes and synthetic ID images. - Behavioral Layer
Deepsight monitors user behavior to identify fraudulent patterns and ensure authentic interactions. This includes analyzing motion dynamics, detecting suspicious bot-like behavior, and verifying natural user activity. - Integrity Layer – Device
Deepsight detects fraudulent attempts by analyzing device signals. This includes identifying suspicious devices, virtual emulators, and fingerprint anomalies.
- Integrity Layer – Camera
Deepsight safeguards camera integrity by identifying fraudulent attempts through analysis of camera interactions. This includes detecting virtual cameras and tampering to ensure live, unaltered video feeds and authenticity.
Incode Deepsight delivers a seamless user experience, enhancing security without adding any additional steps.
Single-Frame Liveness Vs. Multi-Frame Video Liveness
Single-Frame Liveness
- Based on a single frame (a static picture).
- Only RGB modality, where the image is analyzed to find subtle variations in skin texture, reflections and lighting to detect live subjects from spoof attacks (e.g., 2D & 3D masks, screen and paper spoofs).
- APCER (attack presentation classification error rate) ≈ 1.0% for physical check.
Multi-Frame Video Liveness with Deepsight only
- Utilizes multiple RGB frames in digital and physical spoof checks.
- Motion uses an accelerometer and frame comparison to detect injections and deepfakes.
- Depth uses hardware and software to build a face map with information on the camera’s distance.
- Digital Check detects >99% deepfakes, face swaps, and other digital alterations.
- APCER (attack presentation classification error rate) ≈ 0.02% for physical check.
- APCER ≈ 0.6% for digital check.
- APCER ≈ 0.7% for evasion check.
Deepsight’s state-of-the-art deepfake detection can identify pixel-level fingerprints left by genAI tools and identify the source of the deepfake attacking your organization.
As we can see by the color-coded clusters, each synthetic image generation tool has a unique UMAP profile that Deepsight’s AI can recognize.
In other words, it means that even if FaceStudio (the purple cluster at [5,5]) would suddenly generate a completely convincing deepfake, its clear fingerprint would indicate a significant likelihood that the image originated from this tool, and is very probably a deepfake, regardless of how perfect it looks on the surface.
Does Incode Deepsight Ensure Regulatory Compliance?
The growing use and increasing sophistication of deepfakes poses a significant threat to the effectiveness of traditional KYC processes in ensuring compliance with AML regulations.
These new fraud tactics make it easier for unauthorized fraudsters to bypass cybersecurity protocols, potentially access sensitive data, and carry out account takeovers.
To ensure contingency in a time of rapidly evolving fraud vectors, Deepsight’s multi-layered fraud detection achieves a new level of fraud prevention accuracy, outpacing traditional methods that are more or less a sum of their parts.
How Does Incode Deepsight Change the User Experience?
There are no changes whatsoever to the end-user experience. We tested Deepsight extensively to ensure we achieve a design that is unnoticeable to the user: No extra steps are added, and automatic capture of selfies and IDs remains just as fast and intuitive.
This was crucial for maintaining a great experience for good users while blocking fraudsters without letting them know why.
5 Reasons Why Deepsight is essential to protect you from fraud in the AI era
- Better Fraud Detection, Zero Friction
No extra steps, no increase in processing time —security improves without user disruption. - Contingency in Compliance
Multi-layered fraud prevention increases confidence in meeting compliance standards by staying ahead of innovative fraud techniques that outpace regulatory requirements. - Zero-commitment testing
Test Deepsight in shadow mode in any existing verification flow, allowing you to see real results without needing to commit to changing your flows like you would in a proof of concept. - Instant Activation
For current Incode customers, embedding Deepsight is as simple as switching on a toggle in the verification flow. - Reduced Fraud Costs
AI-driven fraud detection minimizes losses and prevents sophisticated fraud in the act.
Interested in Setting Up Incode Deepsight?
Deepsight is available now through the Incode Identity Platform, protecting enterprises across KYC onboarding, step-up verification, authentication, workforce access, and age verification.
If you’re new to Incode, book a call with our team of product experts to learn how Incode Deepsight can help you prevent deepfakes from slipping through your onboarding process.
Interested in implementing Incode Deepsight into your existing IDV workflow? Contact your Incode partner today to get set up with a free trial, including the option for zero-commitment testing in shadow mode. Once set up, you can activate Incode Deepsight simply by clicking a switch on your dashboard. It’s that easy.
Learn more about Incode Deepsight here.
Incode was named a Leader in the 2025 Gartner® Magic Quadrant™ for Identity Verification. Download the report.
