An Intelligent LSTM Based Smart Contract Framework for Detecting Malicious Transactions in Ethereum Blockchain Ecosystem

Abstract

This research proposes a secure and confidential blockchain-based transaction system which would ease blockchain interactions and have high transparency, automation, and mitigation rates. The system is developed using Private Ethereum test network, browser based Ethereum wallet, and Truffle to deploy smart contracts. The transaction processing component enables seamless on-chain asset transfers within the Ethereum blockchain, allowing users to execute transactions without additional technical complexity. To increase the security of the application, a deep learning-based model is deployed for malicious transaction detection by analyzing the user activity patterns to flag and block suspicious transactions. The platform integrates automated wallet provisioning, smart-contract management, currency interoperability, and artificial intelligence into a single architecture, which can be used in financial applications and in digital marketplaces. A comparative experimental analysis of four deep learning models reveals that the LSTM model has the highest accuracy rate of 99.43%, making it the most appropriate model for identifying malicious transactions in the proposed system, which proves the potential of the system as scalable and safe blockchain-based solution.