This paper aims to provide an authorship identification framework that can provide added clues to online auction user profiling systems, the primary purpose of which is early detection and prevention of fraudulent behaviors such as account taken-over. This framework authenticates the authorship of product listing pictures in terms of their image editing styles, and it achieves this goal in three steps: (1) editing style distinction: using a Hough transform based edge clustering algorithm to differentiate between image editing styles created by the same seller, (2) editing style summarization: analyzing and encoding each distinctive image editing style into a template based on the common edge or color features within each style, and (3) authorship authentication: validating the authorship of newly posted images with encoded templates. The experiments show promising results of the proposed framework in identifying authorship of images.
College or School
College of Arts and Sciences
Creative Commons License
Zhang, Chengcui; Chen, Wei-Bang; Zhou, Liping. (2011). "Image Editing Style-Based Authorship Authentication" (2011). 5 pages. https://digitalcommons.library.uab.edu/fw-arts-sciences/3/