IDNet: Identity Analysis Image Dataset

In the digital age, safeguarding against identity theft is paramount. To bolster security measures, we present IDNet, a meticulously curated benchmark dataset designed to enhance privacy-preserving fraud detection systems. This extensive collection comprises 597,900 images, totaling approximately 400 gigabytes, of synthetically generated identity documents. The dataset features driver’s licenses from 10 U.S. states and passports/ID cards from 10 European countries.

Key Features of the IDNet Dataset:

• Diverse Document Types: Includes a wide array of identity documents, such as driver’s licenses and passports, representing various U.S. states and European nations.

• Synthetic Data Generation: All portrait images are synthetically produced, ensuring privacy and eliminating the risk of exposing personal information. These images are provided by.

• Comprehensive Fraud Patterns: Each document is accompanied by multiple fraud scenarios, including text field replacement, face morphing, portrait substitution, and mixed fraud patterns, providing a robust foundation for developing and testing fraud detection algorithms.

• Accessible Format: Due to its substantial size, the dataset is divided into multiple parts, each containing the original synthetic document, various fraud patterns, and associated metadata. All parts are readily available for download.