WARWICK IMAGE FORENSICS DATASET

A comprehensive dataset for advancing digital forensic research.

About • Cameras • How To Use • Citations • License

About

The Warwick Image Forensics Dataset is crafted to address the evolving challenges in the field of digital forensics, specifically for device fingerprinting involving sensor pattern noise (SPN). As digital photography advances, particularly in consumer-level mobile devices, the need for sophisticated analysis tools becomes crucial. This dataset responds to the shifting landscape caused by innovations in camera parameter flexibility and the rise of multi-frame photography techniques like high dynamic range (HDR) imaging.

This dataset features over 6,000 images from 14 different digital cameras, each carefully captured under varied exposure settings. It is designed to facilitate in-depth research on device fingerprinting. It provides a rich resource for exploring how different exposure settings affect forensic analysis, making it an ideal tool for developing and testing new computational photography algorithms.

More detailed information can be found in our original paper Warwick Image Forensics Dataset for Device Fingerprinting in Multimedia Forensics.

Cameras

Details of the cameras used in the Warwick Image Forensics Dataset are crucial for understanding the diversity and capability of data acquisition methods employed. The dataset utilizes 14 different digital cameras, each offering unique features that contribute to the depth and variability of the dataset. Below is a summary table of the cameras used:

No.	Camera	Resolution	Sensor Format	Sensor Dimensions	CFA Type	Lens
1	Canon EOS 6D	3648 x 5472	35 mm	35.8 x 23.9 mm²	Bayer Filter	Interchangeable
2	Canon EOS 6D Mark II	4160 x 6240	35 mm	35.9 x 24 mm²	Bayer Filter	Interchangeable
3	Canon EOS 80D	4000 x 6000	APS-C	22.5 x 15 mm²	Bayer Filter	Interchangeable
4	Canon EOS M6	4000 x 6000	APS-C	22.3 x 14.9 mm²	Bayer Filter	Interchangeable
5	Fujifilm X-A10.1	3264 x 4896	APS-C	23.6 x 15.6 mm²	Bayer Filter	Interchangeable
6	Fujifilm X-A10.2	3264 x 4896	APS-C	23.6 x 15.6 mm²	Bayer Filter	Interchangeable
7	Nikon D7200	4000 x 6000	APS-C	23.5 x 15.6 mm²	Bayer Filter	Interchangeable
8	Panasonic Lumix DC-TZ90.1	3888 x 5184	1/2.3"	6.16 x 4.62 mm²	Bayer Filter	Fixed
9	Panasonic Lumix DC-TZ90.2	3888 x 5184	1/2.3"	6.16 x 4.62 mm²	Bayer Filter	Fixed
10	Olympus E-M10 Mark II	3456 x 4608	Four Thirds	17.3 x 13 mm²	Bayer Filter	Interchangeable
11	Sigma Sd Quattro	3616 x 5424	Foveon X3	23.4 x 15.5 mm²	NA	Interchangeable
12	Sony Alpha 68	4000 x 6000	APS-C	23.5 x 15.6 mm²	Bayer Filter	Interchangeable
13	Sony RX100.1	3648 x 5472	1”	13.2 x 8.8 mm²	Bayer Filter	Fixed
14	Sony RX100.2	3648 x 5472	1”	13.2 x 8.8 mm²	Bayer Filter	Fixed

This comprehensive array of cameras allows researchers to study various aspects of digital image forensics, including sensor noise, resolution impacts, and the effects of different lens and sensor configurations on image forensics.

Instructions

Cloning the Repository

To get started with the Warwick Image Forensics Dataset, you will first need to clone the repository. Use the following command to clone it to your local machine:

git clone https://github.com/CSCRC-SCREED/WIFD.git

Dataset Structure

The dataset is divided into device folders, and within each device folder, there are subfolders for reference images, SDR images, and a single scene. The scene folder contains images categorized by ISO settings and further organized by extension type.

dataset/
├── device_1/
│   ├── reference/
│   │   ├── img_0001.jpg
│   │   ├── img_0002.jpg
│   │   └── ...
│   ├── sdr_image/
│   │   ├── img_0001.jpg
│   │   ├── img_0002.jpg
│   │   └── ...
│   └── scene_01/
│       ├── AEB/
│       │   ├── ISO100/
│       │   │   ├── jpg/
│       │   │   │   ├── img_0001.jpg
│       │   │   │   ├── img_0002.jpg
│       │   │   │   └── ...
│       │   │   ├── cr2/
│       │   │   │   ├── img_0001.cr2
│       │   │   │   ├── img_0002.cr2
│       │   │   │   └── ...
│       │   │   └── ...
|       |   └── ...
│       └── Burst/
│           ├── ISO100/
│           │   ├── jpg/
│           │   │   ├── img_0001.jpg
│           │   │   ├── img_0002.jpg
│           │   │   └── ...
│           │   ├── cr2/
│           │   │   ├── img_0001.cr2
│           │   │   ├── img_0002.cr2
│           │   │   └── ...
│           │   └── ...
│           └── ...
│    
└── ...

Naming Conventions

AEB (Auto Exposure Bracketing) and Burst images are named using the following convention:

<device_name>_scene<scene_number>_ISO<iso>_<exposure_time>_<fnumber>_<sequence_number>.<extension>

Reference and SDR (Standard Dynamic Range) images are named using the following convention:

<device_name>_sdr_image_ISO<iso>_<exposure_time>_<fnumber>_<sequence_number>.<extension>

• device_name: The name of the camera device.
• scene_number: The scene number.
• ISO: The ISO setting used.
• exposure_time: The exposure time for the image.
• fnumber: The aperture value used.
• sequence_number: The sequence number for multiple images with the same settings, only occurs when multiples are present.
• extension: The file extension (e.g., jpg, tif).

Citations

We kindly request that if you use the Warwick Image Forensic Dataset in your research or work, please cite our original paper. Here is the Bibtex entry for citation:

@inproceedings{quan2020warwick,
    author={Quan, Yijun and Li, Chang-Tsun and Zhou, Yujue and Li, Li},
    booktitle={2020 IEEE International Conference on Multimedia and Expo (ICME)}, 
    title={Warwick Image Forensics Dataset for Device Fingerprinting in Multimedia Forensics}, 
    year={2020},
    pages={1-6},
    doi={10.1109/ICME46284.2020.9102783}
}

License

The data and code housed in this repository are released under the MIT License. This choice supports open innovation and collaboration by permitting free use, modification, and distribution of the project.

---

Cyber Security Cooperative Research Centre