Fragile watermarking using Karhunen–Loève transform: the KLT-F approach

Abstract

The paper presents a fragile watermarking technique that may be used to discover image manipulations at block level. The proposed algorithm, based on the Karhunen-Loève linear transform (KLT), modifies some of the KLT coefficients obtained from a secret base (defined by a secret key image) so that they contain a binary pseudo-random message. A genetic algorithm (GA) is used to compensate for rounding errors introduced by inverse transforming in the integer pixel domain. An extensive experimentation has been performed to test the effectiveness of the method and to show the sensitivity of the algorithm to single pixel modifications (also as a function of the number of modified coefficients). A comparison with other fragile watermarking methods is reported. It should be noted that the proposed approach results in both a high PSNR (more than 53 dB on average) and a high subjective quality. The system may be tested online by submitting images to be watermarked and subsequently verifying the presence of modifications in a previously marked image.

Automatic Selection of GA Parameters for Fragile Watermarking

“…In a previous study [4], we also focused on the analysis of the watermarking algorithm properties, by setting the GA parameters to default values (population size=100, mutation probability p m =0.05, crossover probability p c =0.8, terminate if best individual does not change in the last 10 generations or 2000 generations reached) and obtaining good quality results (PSNR between 53 and 54 dB).…”

A corpus of debunked and verified user-generated videos

“…a. Multimedia forensics A large part of related research concerns tampering detection and image/video forensics algorithms. Proposed algorithms attempt to detect and localize image modifications, either actively by embedding watermarks in multimedia content and monitoring their integrity (Dadkhah et al, 2014) (Botta et al, 2015), or passively by searching for telltale self-repetitions (Zandi et al, 2016) (Ferreira et al, 2016) or inconsistencies in the image. Such inconsistencies may appear in the pixel domain or the compressed domain depending on the specific process of tampering.…”

Verification of Web Videos Through Analysis of Their Online Context

“…A large body of research concerns the detection of traces of forgery and tampering in multimedia items (InVID work in this area is further presented in Chapter 6 of this volume). A family of algorithms, known as "active forensics" attempt to deal with image modifications by embedding watermarks in multimedia content and monitoring their integrity [13,5]. Another strand of work focuses on searching for telltale self-repetitions [51,17] or inconsistencies in the image, without making any further assumptions.…”