“We were very happy with the quality of the data that we purchased. It will assist our efforts to improve our product. The team at Unidata was very responsive to our needs, and provided the data format needed to complete our tasks. We are happy to state that Unidata delivers a high quality product in an effective and efficient manner.”
Commercial
iBeta Level 2 Dataset
The dataset validates biometric systems against advanced presentation attacks and supports successful compliance with iBeta Level 2 Presentation Attack Detection certification for liveness detection in security applications
-
- Videos
- 54,590
-
- iBeta Level 2 Completion
- 100%
-
- Types of Attacks
- 10
- Facial Recognition
- iBeta
- Liveness Detection
- Security
- Anti-spoofing
- Computer Vision
- Facial Recognition
- iBeta
- Liveness Detection
- Security
- Anti-spoofing
- Computer Vision
The dataset validates biometric systems against advanced presentation attacks and supports successful compliance with iBeta Level 2 Presentation Attack Detection certification for liveness detection in security applications
- Facial Recognition
- iBeta
- Liveness Detection
- Security
- Anti-spoofing
- Computer Vision
-
- Videos
- 54,590
-
- iBeta Level 2 Completion
- 100%
-
- Types of Attacks
- 10
iBeta Level 1, 2 with 100% Accuracy
What is iBeta?
iBeta is a leading biometric testing lab, accredited with ISO/IEC 17025 and certifying compliance with ISO/IEC 30107.
Since 2018, iBeta Lab has certified over 200 companies, 17% of all of these organizations were Unidata’s clients.
Dataset Info
| Characteristic | Data |
| Description | Videos of people for training algorithms to detect attempts to hack biometric systems. |
| Data types | Video |
| Tasks | Face recognition, Computer Vision |
| Total number of files | 54 590 |
| Labeling | Only technical characteristics and metadata (age, gender, ethnicity) |
| Gender | Male, Female |
| Ethnicity | Caucasian (90%), African (10%) |
| Number of attributes | 31 |
| Type of attack | Real Person, 2D mask, 2D mask with eyeholes, latex mask, wrapped 3D mask, silicone mask |
| People | 101 |
Fintech hit by 100K biometric spoofing attacks monthly
Discover how top fintechs prevent fraud — download our free white paper
iBeta Level 2 Dataset
Real Videos
Outlined 2D mask
3D mask
Real
background
background
Real background
Printed background
Real background
On a real person
On a stand
On a stand
On a real person
Real attributes
(put on a person)
(put on a person)
Real attributes
(put on a person)
(put on a person)
Attributes are
printed on a photo
printed on a photo
Real attributes
(put on a person)
(put on a person)
Attributes are
printed on a photo
printed on a photo
Attributes are printed
on a photo
on a photo
Real attributes
(put on a person)
(put on a person)
Attack 1
9,000 videos
Attack 3
9,600 videos
Attack 4
1,600 videos
Attack 2
8,000 videos
Attack 5
5,000 videos
Attack 6
4,000 videos
Attack 7
4,000 videos
Attack 8
2,500 videos
Attack 9
2,500 videos
Statistics
Statistics
| i | Attribute | Count | Percent |
| 1 | Wigs | 39826 | 72.95% |
| 2 | Glasses | 20403 | 37.37% |
| 3 | Beard | 10322 | 18.91% |
| 4 | Hat | 9635 | 17.65% |
| 5 | Scarf | 8508 | 15.59% |
| 6 | Hood | 8426 | 15.44% |
| 7 | Without attributes | 4470 | 8.19% |
| 8 | Cap | 728 | 1.33% |
| i | Attack Explanation | Count | Percent |
| 1 | 2D mask on a person, photo from real video without accessories, accessories added later, cut along the outline | 9690 | 17.75% |
| 2 | People in the cloth mask | 9090 | 16.65% |
| 3 | People without attacks | 9000 | 16.49% |
| 4 | 2D mask, photo from real video without attributes, attributes are added later, on the table, the lower boundary of the camera is limited by the photo, cropped along the contour | 7999 | 14.65% |
| 5 | 2D mask, photo from real video with attributes, cropped along the outline, on a table, the lower edge of the camera is limited by the photo | 3994 | 7.32% |
| 6 | 2D mask, photo from real video with accessories, not cropped along the outline, on the table, the lower camera boundary is limited by the photo | 3940 | 7.22% |
| 7 | 2D mask, photo from real video with accessories, not cropped along the contour, on a table, the lower boundary of the camera limited by the photo (second device) | 3940 | 7.22% |
| 8 | Silicone mask | 2475 | 4.53% |
| 9 | Silicone mask (second device) | 2474 | 4.53% |
| 10 | 2D mask on a person, a photo from real video without attributes, cropped along the outline | 1616 | 2.96% |
| 11 | Latex mask | 186 | 0.34% |
| 12 | Latex mask (second device) | 186 | 0.34% |
Statistics
- Background Type
- Phone Model Distribution
- View Distribution
- Resolution Distribution
- Gender Distribution
Before You Book iBeta Certification
iBeta is a paid, pass-or-fail exam. We run your model on 31 attack classes first and report error rates per class, segment & condition, so you buy only the 5–30% that fixes the classes it missed.
Audit first with unidata, then the lab
01
We audit your model
02
See what fails, and who gets rejected
03
You buy only the 5–30% you need
04
You book the lab once
Technical
Characteristics
| Characteristic | Data |
| Video Extensions | mp4, MOV |
| Video Resolutions | Min = 1920х1080, Max = 3840х2160 |
| Video duration | 4 second |
| Number of backgrounds | 9 |
| Video overlap | No more than 5% |
| Devices | Mi10s, Google Pixel 4, Samsung Galaxy A03s, iPhone 11, iPhone SE 2, etc. |
Unidata iBeta Datasets in Numbers
- 80+
- times purchased by enterprise clients
- 1.5
- datasets purchased per client
- 30+ companies
- passed iBeta with our data
Fraud prevention & iBeta Certification success
~127K videos
~17K images
8 datasets
- 17%
- of all iBeta certified companies are our clients
Dataset Use Cases
FAQs
What is this dataset used for?
The dataset is designed for biometrics testing, face recognition, and computer vision tasks, especially for developing detection solutions against presentation attacks. It supports IBeta certification workflows, independent testing, and biometric evaluations for biometric security and authentication systems.
What formats and specifications are provided in this dataset?
Videos are delivered in MP4 and MOV formats with resolutions ranging from 1920×1080 to 3840×2160. Each video is around 4 seconds long, captured across 9 backgrounds using devices like iPhone 11, Google Pixel 4, Mi10s, and Samsung Galaxy A03s, ensuring independent quality and broad device coverage.
Is this dataset synthetic data or real-world data?
The iBeta Level 2 Certification Dataset contains real-world biometric data collected from actors under controlled conditions. It provides authentic facial images and biometric data for presentation attack detection, ensuring reliable quality assurance and compliance with accredited biometrics standards.
How are the Unidata datasets licensed?
Unidata datasets follow a dual-access model: free samples are provided for trial and testing, while complete datasets are available exclusively through purchase.
Do Unidata datasets follow GDPR or other data privacy regulations?
Yes. All datasets comply with GDPR and relevant data protection regulations. Information is sourced from lawful channels to guarantee ethical and responsible use.
How are Unidata datasets stored?
Unidata hosts all datasets on AWS cloud infrastructure, designed for reliability, scalability, and strong security. Our data handling complies with ISO 27001 and ISO 27701 standards, ensuring globally recognized levels of information security and privacy protection. This guarantees a secure and trustworthy environment for dataset storage and management.
Does the iBeta Level 2 Dataset meet SLA compliance standards?
Yes. The iBeta Level 2 Dataset is fully SLA compliant, ensuring it meets strict standards for security, reliability, and biometric data integrity. It supports IBeta certification, biometric evaluations, liveness detection, and presentation attacks testing in biometric systems and authentication solutions, aligning with industry best practices for the biometrics industry.
Why is the iBeta Level 2 Dataset valuable for biometric security?
The iBeta 2 dataset helps developers examine how biometric systems respond to advanced spoofing attempts. This can strengthen authentication systems used for digital identity, access control, remote verification, and other security-sensitive applications.
How can mask-based attack data improve facial recognition security?
Mask-based attack examples help liveness detection algorithms learn characteristics associated with artificial facial presentations. This can improve the ability of biometric systems to identify suspicious presentations before granting authentication.
Still have questions about using Unidata datasets?
Read our user-guides
Unidata Cases
Digital Tree Passport Annotation for Forest Mapping
- Forestry Monitoring & GIS
- 200,000 trees, 10 species classes
- 2 months
License Plate Annotation for Vehicle Recognition System
- 100,000 images with detailed license plate markup (bounding boxes, digits, regional symbols)
- 3 weeks
Sentiment Annotation for Brand Monitoring
- Marketing & Consumer Insights
- 12,000 text samples
- 7 weeks
Surveillance Video Annotation for Entrance Monitoring
- Surveillance & Security
- 90 minutes of video from three cameras, approximately 50-60 thousand frames
- 2 week
Similar Datasets
-
Commercial
- Identity Verification
- Computer Vision
- Machine Learning
- Biometric Recognition
- Security
Human Iris Images Biometric Dataset
High-quality iris dataset with 5,000+ JPG images from 5,000+ individuals, containing left and right eye captures for iris recognition and computer vision tasks. This eye iris dataset includes labeled images collected under visible light, with metadata (ID, eye, sex) to support training data for biometric recognition systems, iris segmentation, and iris pattern analysis.
5,000+ Images
5,000+ People -
Commercial
- Computer Vision
- Machine Learning
- Image Processing
- Security
- Anti-Spoofing
Multi-Material Fingerprint Spoofing Dataset
This fingerprint spoofing dataset contains fingerprint images, captured with a ZKTeco ZK9500 optical scanner and including real fingerprints and spoofing attacks created with alginate, plasticine, and silicone materials. The fingerprint dataset includes metadata (gender, age, finger, hand, device) and supports biometric security research, presentation attack detection, spoof detection, and fingerprint recognition model training.
100 People
4000+ Photos -
Commercial
- Computer Vision
- Machine Learning
- Image Processing
- Security
- Anti-Spoofing
Biometric Fingerprint Spoofing Dataset
Contains 5,000+ high-quality fingerprint images capturing real fingerprints and multiple fingerprint spoofing attack types, including print and replay scenarios. Designed for spoofing detection and liveness detection tasks, the fingerprint dataset provides labeled biometric data from different devices and fingers to train and evaluate biometric security and fingerprint recognition systems.
100 People
5000+ Photos -
Commercial
- Facial Recognition
- Liveness Detection
- Security
- Anti-spoofing
- Computer Vision
Anti-Spoofing Replay PC Videos Dataset
High-quality replay attack dataset containing 4,714 PC-recorded video clips of real faces, designed for training and evaluating face recognition and liveness detection systems. This anti-spoofing videos dataset includes diverse attack scenarios, technical metadata (age, gender, ethnicity), and MP4/MOV formats to support spoofing detection, biometric security, and computer vision model development.
4,714 Videos
4,714 People -
Commercial
- Facial Recognition
- Liveness Detection
- Security
- Anti-spoofing
- Computer Vision
Anti-Spoofing Replay Phone Videos Dataset
This anti-spoofing dataset contains over 40,036 live facial video recordings captured on mobile devices to support replay attack detection and biometric anti-spoofing research. With paired video sets, MP4/MOV formats, and rich metadata such as age, gender, and ethnicity, it provides reliable training data for face antispoofing, liveness detection, and secure biometric authentication systems.
40,036 Videos
20,018 Sets -
Commercial
- Speech Analysis
- ASR
- Machine learning
- Data generation
- Audio Processing
Real vs Fake Human Voice – Deepfake Audio Dataset
Audio dataset containing 5,000 audio files featuring both genuine human recordings and AI-generated voice samples. Each set includes four speakers with multiple clips across M4A and MP3 formats. The dataset supports research in deepfake detection, generated speech analysis, and real vs fake human voice recognition tasks.
5,000 Audio
-
Commercial
- Facial Recognition
- Security
- Anti-spoofing
- Computer Vision
- Machine Learning
Kids Anti-Spoofing Dataset
The dataset provides 6,000 high-quality facial images of children aged 7–15 for face anti-spoofing and liveness detection tasks. This child safety dataset supports research in biometric systems, helping improve facial recognition accuracy, detect spoofing attacks, and build safer AI models for protecting kids in digital and identification environments.
6 000 Images
300 people -
Commercial
- Image Processing
- Machine Learning
- Hand Recognition
- Forensics
- Computer Vision
Open Palm Hand Images Dataset
This high-quality open palm dataset includes 179,052 annotated images collected from 29,842 people, with each set containing six palm photos, two printed-hand images, and two replay videos. Designed for hand recognition and computer vision research, it provides detailed metadata - age, gender, ethnicity, profession, device type, dominant hand, and jewelry status.
179,052 Images
29,842 People -
Commercial
- PII
- Data generation
- Security
- Anti-spoofing
- Computer Vision
Synthetic Printed Turkish Passports Dataset
It is a synthetic Turkish passports dataset containing 5,000 high-quality, AI-generated images. Labeled with detailed metadata - including passport ID, class, gender, and lighting - this dataset supports PII extraction, identity verification, and biometric recognition system training while maintaining strict data protection standards.
5000 Images
-
Commercial
- Facial Recognition
- iBeta
- Liveness Detection
- Security
- Anti-spoofing
- Computer Vision
iBeta Kids Dataset
It is a child safety dataset featuring over 46,000 short video samples of children across different age groups, recorded under varied lighting, devices, and conditions. It includes four main attack types: Real Person, 2D Mask, 3D Mask, and Replay, helping develop biometric systems that detect spoofing and ensure safe, accurate child identification.
45 600 Videos
60 People
Why Companies Trust Unidata's Datasets
Share your project requirements, we handle the rest. Every service is tailored, executed, and compliance-ready, so you can focus on strategy and growth, not operations.
Our Clients Love Us
Ready to get started?
Tell us what you need — we’ll reply within 24h with a free estimate
- Andrew
- Head of Client Success
— I'll guide you through every step, from your first
message to full project delivery
Thank you for your
message
We use cookies to enhance your experience, personalize content, ads, and analyze traffic. By clicking 'Accept All', you agree to our Cookie Policy.