Printed 3D Mask Attack - 3,500+ videos

The 3D Mask Attack Detection Dataset comprises over 3,500 videos captured across five different devices. Each video features individuals holding a photo fixed on a cylinder, designed to simulate potential presentation attacks against facial recognition systems. This dataset is instrumental in advancing research on attack detection and enhancing spoofing detection techniques, particularly for fraud prevention and compliance with iBeta Level 1 certification standards.

Dataset Details:

• Purpose: Developed to support research in attack detection and improve spoofing detection techniques, this dataset is crucial for fraud prevention and ensuring compliance with iBeta Level 1 certification standards.

• Funding: The creation of this dataset was not funded by any external organization.

• Instances Represent: Each instance corresponds to a video recording of an individual simulating a presentation attack using a 3D mask.

• Data Splits: The dataset does not include predefined data splits; users are encouraged to partition the data as per their research requirements.

• Sensitive Data: The dataset does not contain sensitive information, ensuring compliance with privacy standards.

• Missing Values: The dataset is complete, with no missing values reported.

Key Features:

• Diverse Capture Devices: Videos are captured using five different devices, providing a varied dataset that enhances the robustness of detection models.

• Simulated Presentation Attacks: The dataset includes videos of individuals holding photos fixed on cylinders, simulating potential presentation attacks against facial recognition systems.

• Large-Scale Collection: With over 3,500 videos, the dataset offers a substantial amount of data for training and evaluating detection algorithms.