A context-aware framework for updating dynamic NFT attributes within the tokenization service layer
Year: 2026
Type: article
Abstract: The immutability of blockchains complicates updating dynamic attributes of Real-World Asset (RWA) backed Non-Fungible Tokens (NFTs), often requiring new NFTs to be minted when asset properties change. We present a context-aware framework that updates dynamic (mutable) attributes while preserving static (immutable) ones. The framework employs two machine learning (ML) layers: (i) contextual analysis that identifies static and dynamic attributes from unstructured text, and (ii) context-aware learning that determines when and which dynamic attributes require updates based on real-world data from property listings, market indices, news articles, and social media feeds. An adaptive smart contract (ASC) applies ML predictions to update dynamic attributes while locking static ones. We evaluated four contextual models, using accuracy, precision, recall, F1-score and five models for context aware update using using MSE, RMSE, MAE, and R 2 . Results show that CNN-LSTM achieves 90.5% accuracy in identifying static and dynamic attributes, while DANBoost, a hybrid ensemble combining DARNet and CatBoost attains 92% accuracy in predicting relevant updates. The ASC enforces secure ML-driven updates, preserves immutability, ensures access control and data integrity, and maintains gas usage within normal limits.
Source: Array
Institutions: RMIT University, Data61
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,680
OpenAlex ID: W7220859232