Unveiling fake artworks: using automatic classification based on freely available information

Résumé fourni par la source

Purpose In the dynamic art market, fake artworks pose a significant threat to cultural heritage preservation. This article aims to introduce a novel approach to enhancing art fraud prevention through the application of computer-based art classification techniques. Design/methodology/approach The study analyzes attributes such as subject matter, size, date, and signature drawn from documented sources, including sales receipts, auction records, and online descriptions, and examines how art buyers can assess the risk of fakes based on these features. We conducted a comparative analysis of Félix Vallotton's paintings. The methodology involved compiling a detailed database, including characteristics from 1,704 original pieces and 210 fakes, to establish a robust classification and regression tree (CART) model. Findings Our results reveal that specific combinations of characteristics effectively distinguish original from fake artworks. The findings demonstrate the effectiveness of our approach for Vallotton's works and suggest its potential for standardization and application to other artists, providing a reliable and independent preventive measure against art counterfeiting. Originality/value Our study is original for at least two reasons. First, we were able to analyze a database comprising not only authentic works by Félix Vallotton but also more than 200 fakes, which gave us an excellent point of comparison. Second, we based our analysis on characteristics of artworks that are accessible to everyone, not just specialists, making our work directly accessible to the general public.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.