Unmasking Cybercrime Strengthening Digital Identity Verification against Deepfakes 2026

TABLE 1 Evaluation criteriaEvaluation of deepfake tools Legitimate content creation tools are being adapted for criminal purposes.03 Numerous face-swapping methods and camera injection tools are discussed or advertised on the internet, with some claiming the capability to bypass digital KYC systems. To assess whether these claims are well-founded, and to characterize the current deepfake landscape, a set of widely discussed face-swapping and camera injection tools was collected from darknet forums, Telegram channels and other social media platforms for the period 1 July 2024 to 30 April 2025.The evaluation process followed the criteria listed in Table 1 and was applied separately to the face-swapping applications and camera injection toolsets. The evaluation covered 17 face- swapping tools and eight camera injection tools. To reduce the potential for misuse, vendor identities, discovery methods and step-by-step exploitation techniques have been redacted from this paper. High-level technical observations, aggregated metrics and non-exploitable examples are presented.Overview of evaluation process Category Evaluation criteria 1 Face-swapping tools (fake media generators) Tool basics Tool name, version, source (GitHub, Telegram, .onion), price/licensing Media type Static image swap/real-time video/3D avatar/animation Real-time capability Can it operate in live mode (e.g. front cam selfie KYC)? Facial expression sync Smile, blink, head turn, eye tracking supported? Lighting adaptation Does the fake face adapt to real lighting changes? Voice integration Built-in voice clone? Compatible with external voice clone tools? Target use cases KYC, adult content, livestreaming, impersonation Quality metrics Artefact presence (e.g. flicker, blending), resolution, smoothness Model typeGenerative adversarial network (GAN)-based, encoder-decoder, neural radiance fields (NeRF), StyleGAN, Avatarify, etc. Device compatibility Android, iOS, PC; rooted/jailbroken required? Resource demands Graphics processing unit (GPU)/central processing unit (CPU)/memory use during real-time rendering Spoof detection bypassCan it fool blink/liveness/selfie video detection? Output format Export as video/image or directly stream to camera (VCam)? Toolchain dependenciesFFmpeg, TensorFlow, OpenCV, face landmarks library, etc. Network behaviour Connects to remote server? Downloads model weights? Customization Can the fraudster upload and replace specific face ID (victim photo/video)? Unmasking Cybercrime 8