FaceLivenessDetection-SDK Overview
FaceLivenessDetection-SDK is an open-source software development kit (SDK) developed by DoubangoTelecom, designed to address critical challenges in biometric security through advanced anti-spoofing and deepfake detection. Released in April 2021, the SDK employs a 3D passive liveness detection approach, requiring only a single image to compute a liveness score.
The SDK’s accuracy is a defining feature, achieving a 99.67% success rate on DoubangoTelecom’s proprietary dataset. It also attains perfect scores on multiple public benchmark datasets, including NUAA (National University of Defense Technology Anti-Spoofing), CASIA FASD (Chinese Academy of Sciences Institute of Automation Face Anti-Spoofing Database), and MSU (Michigan State University Mobile Face Spoofing Database).
The project is actively maintained, with the latest updates pushed to its GitHub repository as recently as September 2025. The SDK is implemented in C++, a language known for its performance and efficiency in computationally intensive tasks such as deep learning and real-time processing.
How It Works
The SDK’s core functionality revolves around 3D passive liveness detection, a technique that analyzes facial images to determine whether they originate from a live human or a spoofing attempt. Unlike active liveness detection methods, which often require user interaction (e.g., head movements or specific lighting conditions), the passive approach relies solely on static image analysis.
Once the face is identified, it applies a series of deep learning-based algorithms to assess various biometric cues. The combination of these techniques enables the system to distinguish between genuine liveness and synthetic or replayed content.
A key innovation of the SDK is its ability to perform deepfake detection within the same pipeline. The SDK addresses this by incorporating convolutional neural networks (CNNs) and other machine learning models trained on datasets containing both authentic and synthetic facial images. The result is a unified system capable of handling both traditional spoofing attacks and emerging threats from AI-generated content.
This flexibility is further supported by the SDK’s compatibility with frameworks such as Keras and TensorFlow, which are widely used in the machine learning community.
Use Cases
FaceLivenessDetection-SDK is particularly well-suited for applications where secure identity verification is paramount. One of the most prominent use cases is in electronic Know Your Customer (eKYC) processes, where financial institutions and regulatory bodies require robust authentication to prevent fraud. The SDK’s ability to detect spoofed or deepfake identities in a single image streamlines the verification process, reducing the risk of identity theft or synthetic media exploitation.
Another significant application lies in Anti-Money Laundering (AML) systems. Its passive nature also aligns with the need for seamless user experiences, as it does not require additional hardware or user interaction beyond a standard image capture.
The SDK’s ability to detect deepfakes also addresses the growing threat of AI-generated impersonation, where attackers may use synthetic media to bypass security measures.
The SDK’s deepfake detection capabilities enable real-time identification of such manipulations, ensuring the integrity of video-based authentication processes.
Key Features and Capabilities
- Single-Image Processing: The SDK operates on a single static image, eliminating the need for multi-frame analysis or active user prompts.
- 3D Passive Liveness Detection: By simulating 3D analysis from 2D images, the SDK assesses depth and texture cues that are indicative of liveness. This approach is effective against a wide range of spoofing attacks, including printed photos, video replays, and silicone masks.
- Deepfake Detection: The integration of deep learning models trained on synthetic media datasets enables the SDK to identify AI-generated facial manipulations.
- This is crucial for maintaining the integrity of video-based authentication and communication systems.
- Face Recognition Integration: Beyond liveness detection, the SDK can be used in conjunction with face recognition systems to ensure that the identified face is indeed live.
- Its compatibility with Keras and TensorFlow ensures seamless integration into existing machine learning workflows.
- High Accuracy on Benchmark Datasets: The SDK has demonstrated perfect scores on multiple public datasets, including NUAA, CASIA FASD, and MSU. This validates its effectiveness in real-world scenarios and against diverse spoofing techniques.
- Open-Source and Customizable: Available under a custom or multiple licenses, the SDK offers flexibility for both commercial and research applications.