---
title: "2D Print Attacks and Silicone Masks Attacks Dataset"
description: "The dataset contains images and videos of real faces and spoof attempts, including printed photos, silicone masks, and varied facial attributes. Designed for liveness detection…"
url: "https://unidata.pro/datasets/2d-print-attacks-and-silicone-masks-attacks-dataset/"
date_modified: "2025-11-25T15:58:23+03:00"
language: "en-US"
---
The dataset contains images and videos of real faces and spoof attempts, including printed photos, silicone masks, and varied facial attributes. Designed for liveness detection and facial recognition research, it functions as a face anti-spoofing dataset for training and evaluating AI models that distinguish genuine identity features from presentation attacks under real-world conditions.

## Dataset Structure

### The Numbers Section

**Numbered list:**

- **Number:** 16,867 — **Text:** images
- **Number:** 2,929 — **Text:** videos

### Tooltips Section

**Tooltip items:**

- **Name:** iBeta
- **Name:** Computer Vision
- **Name:** Machine learning
- **Name:** Security
- **Name:** Anti-spoofing
- **Name:** Liveness Detection

### Dataset Information

**Table with data:**

| Characteristic | Data |
| --- | --- |
| Description | Videos of individuals wearing or holding printed/silicone masks |
| Data types | Video, Image |
| Tasks | Liveness Detection, Computer Vision, iBeta |
| Number of image | 16,867 images (12,656 2D-print, 4,211 silicone) |
| Number of video | 2,929 videos (2,480 2D-print, 449 silicone) |
| Labeling | Metadata (age, gender, ethnicity) |
| Gender | Male, Female |
| Name of attributes | Wigs, glasses, beards/mustaches, masks, clothing, etc |

**Media Slider:**

- **Image in the slider:** ![](https://unidata.pro/wp-content/uploads/2025/07/2d-print-attacks-and-silicone-masks-attacks-dataset0a-primerfoto2.webp)
- **Image in the slider:** ![](https://unidata.pro/wp-content/uploads/2025/07/2d-print-attacks-and-silicone-masks-attacks-dataset0a-primerfoto1.webp)

**Link to the sample:** [Download sample](https://drive.google.com/drive/folders/1WKK3gF-ZASd82kfJq7zgo3UunxFNhOUr?usp=sharing)

### Technical Specifications

**Table with data:**

| Characteristic | Data |
| --- | --- |
| Video extension | mp4 |
| Image extension | JPG |

**Source and data collection methodology:** Source and collection methodology: Data was collected by a partner of Unidata.

### Dataset Use Cases - Slider

**Industry Cards:**

- **Industry:** Biometric Security — **Title:** Strengthening Face Anti-Spoofing Systems — **Text:** This dataset provides training data for recognizing presentation attacks such as 2D print attacks, silicone masks, and paper masks. It helps biometric systems distinguish real faces from fake faces under varied lighting conditions and capture settings. Researchers can use these samples to refine face antispoofing pipelines, reduce spoofing attacks, and improve identity verification workflows in real environments.
- **Industry:** Facial Recognition Technology — **Title:** Improving Attack Detection in Recognition Systems — **Text:** Facial recognition teams can use this dataset to study how different attack types – printed photos, replay attacks, and 3D masks – impact recognition algorithms. Since the dataset contains videos and face images with clear facial features, it supports detecting fake identities with better precision. The material helps developers tune detection algorithms for stable performance across recognition systems and devices.
- **Industry:** Digital Forensics — **Title:** Analyzing Spoofing Methods and Biometric Attacks — **Text:** Forensic analysts rely on datasets containing real faces and spoofed samples to examine biometric attacks in detail. This dataset provides varied examples of silicone masks, fake faces, and printed photos that allow teams to trace how spoofing attempts bypass weaker models. Its structure helps specialists evaluate deep models under challenging scenarios and document evidence for security audits.
- **Industry:** AI & Machine Learning Research — **Title:** Training Deep Models for Anti-Spoofing Detection — **Text:** Machine learning teams use this dataset as a training set for detecting fake identities and building stronger antispoofing models. Since the dataset includes videos across multiple presentation attacks, it supports deep learning experiments focused on robust detection. The samples help refine recognition algorithms, improve generalization, and test new architectures for preventing biometric fraud in diverse conditions.

### Fact

**FAQs Heading:** FAQs

**List of Questions:**

- **Question:** Can I request a sample of the dataset before downloading or purchasing it? — **Answer:** Yes, Unidata provides free dataset samples so you can evaluate the data quality, attack types, and annotation structure. These samples help you confirm compatibility with your biometric systems, deep learning pipelines, or facial recognition models before purchase.
- **Question:** Which presentation attack types are included in the dataset? — **Answer:** The dataset includes two major biometric attack categories: 2D print attacks and silicone mask attacks. Specifically, it contains 12,656 printed attack images, 4,211 silicone mask images, 2,480 printed attack videos, and 449 silicone mask videos, enabling comprehensive evaluation of presentation attack detection systems.
- **Question:** How does this dataset improve face recognition models? — **Answer:** By exposing machine learning models to diverse spoofing attacks and genuine facial samples, the dataset improves the robustness of face recognition, presentation attack detection (PAD), and biometric authentication systems. The combination of multiple attack types helps reduce false acceptance rates in real-world deployments.
- **Question:** What types of annotations or metadata are provided? — **Answer:** Each sample includes metadata describing demographic attributes and presentation attack attributes such as mask type, accessories, and other visual conditions.
- **Question:** How is the data in this dataset collected? — **Answer:** Data was collected through controlled acquisition sessions performed by an official Unidata partner, ensuring consistency across lighting, pose variations, and attack types. The collection process captures realistic biometric attacks, including printed photos and silicone masks, to support robust antispoofing research.
- **Question:** How are Unidata datasets licensed? — **Answer:** Unidata datasets follow a dual-licensing model: free samples are available for evaluation, while full datasets require purchase. This ensures that organizations can test compatibility before committing to a complete dataset.
- **Question:** Do Unidata datasets comply with GDPR and other privacy regulations? — **Answer:** Yes. All data is curated in full compliance with GDPR and relevant data protection laws, using only ethically sourced and legally captured biometric data to support responsible AI development.
- **Question:** How are Unidata datasets stored? — **Answer:** Unidata stores all datasets on secure AWS cloud infrastructure compliant with ISO 27001 and ISO 27701 standards. This ensures high reliability, privacy protection, and scalable access to large biometric datasets.
- **Question:** Does the dataset include both 2D print attacks and silicone mask attacks? — **Answer:** Yes, the dataset includes a large number of both attack types: 12,656 2D-print images, 4,211 silicone mask images, 2,480 print-attack videos, and 449 silicone-mask videos. This variety supports research in detecting multiple forms of biometric spoofing.

[Full list of this site's AI-readable pages](https://unidata.pro/llms.txt)