D-ARTNET22 V1: a neural network framework against stolen digital artworks getting Non-Fungible Token (NFT) labels

Abstract

"Since NFT sites gained fame for selling digital arts, NFT crimes have taken a toll on excessive amounts of digital content creators as stolen digital artworks get their ownership changed permanently in the name of the thief, further getting sold on humongous fortunes. Due to NFT sites not having any user or content verification system before registration, thieves tend to take the chance of scamming even more by adopting various forgery protocols. Artworks from social media or NFT sites are stolen, forged, and then registered under different names. On the contrary, since blockchains are immutable, the thief remains the owner of the stolen NFT forever which implies that NFT sites fail to provide a secure space for hardworking digital content creators. According to what has been researched, it is discovered that there exists no such work relating to digital media. Despite connecting some certain dis- joint fields, the results were not promising and thus they were not thought to be implemented in real life. Besides, digital artwork datasets are not available online for the purpose of this field to be served. A possible methodology can be doing exten- sive image scraping on selective digital media platforms to extract digital artworks that may then be modified to create a fabricated artwork dataset. This dataset can subsequently be used to train deep learning or neural network models to distinguish between actual and false entities. As no verification system for NFT sites has been proposed before, it is crucial to develop a system to check the authentication of dig- ital artworks and the user before the NFT transaction is passed into the blockchain. Therefore, for the very first time, this paper will present a framework that will check the originality of digital artworks before accepting them as an NFT permanently."