MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202621075795 A) filed by Mrs. Saltanat Mohammed Abdullah Shaikh; and Dr. Sangeeta Vhatkar on June 18, 2026, for Historical Optimized Gated Graph Attention-Based Distributed Neural Network (hg2adn) Model For Reentrancy Vulnerability Detection In Ethereum Smart Contracts.
Inventors include Mrs. Saltanat Mohammed Abdullah Shaikh; and Dr. Sangeeta Vhatkar.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: Reentrancy is one of the most severe security vulnerabilities in Ethereum smart contracts, often leading to unauthorized fund withdrawals and significant financial losses in decentralized applications. Existing detection techniques, including static analysis and rule-based systems, suffer from high false positive rates, limited ability to capture complex structural patterns, and high computational overhead. To address these limitations, the present invention proposes a Historical Optimized Gated Graph Attention-based Distributed Neural Network (HG2ADN) model for efficient detection and classification of reentrancy vulnerabilities in smart contracts. The proposed system transforms smart contract code into structured graph representations, where nodes represent functional elements such as variables, functions, and external calls, while edges represent control flow and data dependencies. An Autoencoder Distance-based Multi-metric Graph Feature (ADMGF) module is employed to extract compact and informative feature embeddings by reducing dimensionality while preserving critical semantic and structural information. These features are processed using a gated graph attention network that dynamically assigns importance to graph components, enabling precise identification of vulnerability-prone execution paths.To enhance scalability and efficiency, the model is implemented within a distributed neural network framework that enables parallel processing of large-scale blockchain data. Furthermore, a Historical Chip Optimization (HiCO) algorithm is integrated to improve convergence speed and stability by leveraging historical training performance for parameter optimization. The combined framework achieves improved detection performance while reducing computational cost. Experimental evaluation demonstrates that the HG2ADN model achieves an accuracy of 97.10%, precision of 97.62%, and recall of 96.48%, outperforming existing approaches. The proposed invention provides a robust, scalable, and intelligent solution for securing Ethereum smart contracts against reentrancy vulnerabilities.
Disclaimer: Curated by HT Syndication.
