Disinformation detection and source tracking using semi-fragile watermarking and blockchain
Tanya Koohpayeh Araghi, David Megías, Víctor Garcia-Font, Minoru Kuribayashi, Wojciech Mazurczyk · 2024
A blind semi-fragile watermarking scheme for disinformation detection in online social media (OSM) is presented. The proposed scheme can be used for both images and videos. We also propose a blockchain-based framework for traceability and propagation curtailment of fake images, even if they are republished in different OSMs. When an image is uploaded to an OSM, a token is created in a blockchain including a perceptual hash of the image, and, if required, information to identify the OSM and the publisher. In case of fake news detection, users can request the OSM to extract the watermark. Tampered areas can be identified by a masking file stemming from the extracted watermark to provide proof of the fake content. Additionally, information about the originator and the propagators of the image can be extracted from the blockchain and the token image will be included in a public blacklist to avoid further dissemination. The proposed scheme has been tested on more than 100 images, and experimental results show a high level of accuracy for fake content detection, localization, and curtailment with a range of attacks like copy, move, swapface, and collage attack in the swapface.