AI-Powered Visual Similarity Detection and Blockchain-Based Copyright Verification Framework for Preventing Art Appropriation

Abstract

Artistic content, especially in the digital era, is increasingly susceptible to unauthorized appropriation, imitation, and misuse, often with little to no attribution to original creators. In this paper, an unified AI-assisted visual similarity-detecting system with a blockchain-based provenance-checking base is introduced to prevent art appropriation and securely and in interpretable and real-time apply digital copyright. With the help of a hybrid deep learning framework that consists of Vision Transformer (ViT) and ResNet-based architectures, the system compares the stylistic and structural characteristics of the artworks with those of cosine similarity, Euclidean distance, and style loss measurements. The triplet loss-trained visual encoder allows the model to be useful in distinguishing between direct copies, derivative works, and inspired originals. In an image-based art collection (designed as a WikiArt replica, though not a real WikiArt collection) of more than 10,000 examples, the framework obtained a high classification performance of 94.3, and exact copy performance (precision) of 96.4 and inspired work performance (precision) of 88.6. Artwork metadata and visual fingerprints are hashed with SHA-256 and stored on a permissioned blockchain (Hyperledger Fabric) to provide provenance records and immutable and to enforce copyright permissioned via smart contracts. The proposed system is more accurate in detection and explainable than the available tools like Google Reverse Search and the TinEye. It is also supplemented with Grad-CAM-based visual heatmaps and real-time smart contract execution as the dispute resolution. Further development of the model will focus on extending it to video art and 3D designs, as well as generative AI art (e.g., GANs, Diffusion), and the global copyright databases will be integrated to be universal. This system is a strong move towards the protection of the integrity and rights of Web3 digital artists.