A Secure Perceptual Hash Algorithm for Image
Content Authentication
Li Weng and Bart Preneel
⋆
Katholieke Universiteit Leuven, ESAT/COSIC-IBBT
li.weng@esat.kuleuven.be,bart.preneel@esat.kuleuven.be
Abstract. Perceptual hashing is a promising solution to image content
authentication. However, conventional image hash algorithms only offer
a limited authentication level for the protection of overall content. In this
work, we propose an image hash algorithm with block level content pro-
tection. It extracts features from DFT coefficients of image blocks. Ex-
p eriments show that the hash has strong robustness against JPEG com-
pression, scaling, additive white Gaussian noise, and Gaussian smooth-
ing. The hash value is compact, and highly dependent on a key. It has
very efficient trade-offs between the false positive rate and the true pos-
itive rate.
1 Introduction
In the Internet era, images are massively produced and distributed in digital
form. Although digital images are easy to store and pro cess, they are also sus-
ceptible to malicious modification. Due to widely available image editing soft-
ware, even non-professionals can perform content modification. Consequently,
people begin to suspect what they see from digital images. Sometimes, public
incidents happen, due to fake images. Therefore, the need for protecting content
authenticity is emerging.
Among various techniques, perceptual hashing is a promising solution. Hash-
ing means to compute a digest value from data. This digest value, typically a
short binary string, is called a hash value. Perceptual hash algorithms are a
particular kind of hash algorithms for multimedia data. They have the special
property that the hash value is dependent on the multimedia content, and it re-
mains approximately the same if the content is not significantly modified. Since a
⋆
This work was supported in part by the Concerted Research Action (GOA) AM-
BioRICS 2005/11 of the Flemish Government and by the IAP Programme P6/26
BCRYPT of the Belgian State (Belgian Science Policy). The first author was sup-
p orted by the IBBT/AQUA project. IBBT (Interdisciplinary Institute for Broad-
Band Technology) is a research institute founded in 2004 by the Flemish Govern-
ment, and the involved companies and institutions (Philips, IPGlobalnet, Vital-
sys, Landsbond onafhankelijke ziekenfondsen, UZ-Gent). Additional support was
provided by the FWO (Fonds Wetenschappelijk Onderzoek) within the project
G.0206.08 Perceptual Hashing and Semi-fragile Watermarking.
