WARWICK IMAGE FORENSICS DATASET FOR DEVICE FINGERPRINTING IN MULTIMEDIA FORENSICS Yijun Quan∗, Chang-Tsun Liy, Yujue Zhou∗ and Li Liz ∗Department of Computer Science, University of Warwick, UK E-mail: fy.quan,
[email protected]; ySchool of Information Technology, Deakin University, Australia E-mail:
[email protected]; z School of Computer Science and Technology, Hangzhou Dianzi University, China E-mail:
[email protected] ABSTRACT are very important for the study of device fingerprint analysis and the development of relevant techniques. Device fingerprints like sensor pattern noise (SPN) are widely As the digital forensic community is gaining more under- used for provenance analysis and image authentication. Over standing of image device fingerprinting, digital and computa- the past few years, the rapid advancement in digital photog- tional photography has undergone huge development as well. raphy has greatly reshaped the pipeline of image capturing Driven by the need for consumer-level devices to produce bet- process on consumer-level mobile devices. The flexibility ter images, we witness significant advances in both hardware of camera parameter settings and the emergence of multi- and software development. As far as hardware is concerned, frame photography algorithms, especially high dynamic the improvement in the design of electronic components like range (HDR) imaging, bring new challenges to device fin- complementary metal-oxide-semiconductor (CMOS) brings gerprinting. The subsequent study on these topics requires better noise immunity. Such improvements allow cameras a new purposefully built image dataset. In this paper, we to have greater flexibility in camera parameter settings, es- present the Warwick Image Forensics Dataset, an image pecially for using high signal gain (commonly known by dataset of more than 58,600 images captured using 14 digital the name of ISO speed in photography) without introduc- cameras with various exposure settings.