Dark Mode
High-Fidelity Biometric Attack and Security Verification Dataset
Data Science and Analytics
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
Safeguarding biometric systems against sophisticated physical attacks is the primary focus of these records. By providing a vast library of over 2,800 videos, the collection documents genuine facial presentations alongside fraudulent attempts using silicone masks and 2D printed overlays. The recordings are captured under diverse lighting conditions—ranging from dark rooms to bright daylight—and include subjects wearing various accessories such as glasses, wigs, and hats. This diversity allows for the development of robust anti-spoofing algorithms capable of distinguishing between real human skin and complex synthetic materials.
Columns
- video: A direct link or path used to access the specific video file within the repository.
- type: The classification of the video, identifying it as a real person (real), a person wearing a printed 2D mask (outline/mask), or a person wearing a silicone mask (silicone).
Distribution
The information is primarily organised within a CSV file, such as data.csv or dataset_info.csv, with a file size of approximately 1.3 kB for the sample listing. The full dataset consists of over 5,700 unique video files. These records maintain a perfect usability score of 10.00, with 100% validity across the video and type fields. No future updates are expected for this collection.
Usage
This resource is perfectly suited for training deep neural networks for liveness detection and presentation attack detection (PAD). It can be used to develop security systems for banking, mobile unlocking, and border control that must defend against high-fidelity 3D mask attacks. Researchers can also utilise the various lighting scenarios and character attributes to test the generalisability of their computer vision models in challenging real-world environments.
Coverage
The scope includes 5,792 videos in total, featuring a significant demographic spread of 3,107 men and 2,685 women. It covers three main attack vectors: real-life captures, 2D printed masks, and 3D silicone masks. The recordings account for four distinct lighting environments: dark room, daylight, light room, and nightlight, ensuring broad environmental applicability.
License
Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
Who Can Use It
Biometric security engineers can use these videos to bolster the accuracy of face recognition systems against spoofing. AI researchers focused on computer vision can leverage the diversity of masks and lighting to benchmark new detection algorithms. Additionally, developers of surveillance and identity verification platforms will find the structured attack scenarios essential for creating secure, fraud-resistant user authentication workflows.
Dataset Name Suggestions
- Silicone Mask Biometric Spoofing and Anti-Attack Repository
- Global Face Liveness Detection and 3D Mask Attack Database
- Biometric Presentation Attack Detection (PAD) Video Archive
- Enhanced Anti-Spoofing Dataset: Silicone and 2D Mask Variants
- High-Fidelity Biometric Attack and Security Verification Dataset
Attributes
Original Data Source: High-Fidelity Biometric Attack and Security Verification Dataset
QUALITYLoading...
Free
Download Dataset in ZIP Format
Recommended Datasets
Loading recommendations...