Multilabel Vulnerability Classification in Decentralized Blockchain–Based Reputation System

Abstract

Smart contracts serve as decentralized applications essential for extensive utilization of blockchain technology across various contexts that have transitioned from the blockchain, characterized primarily by digital currency systems that emphasize the financial systems. Blockchain operates as a distributed ledger that securely records transactions using cryptographic techniques to establish a unique, chain-like data structure managed collectively by miners within the network. However, current methods for analyzing smart contracts often demand substantial processing time and face challenges in accurately detecting vulnerabilities in complex contracts. To address these limitations, this research introduces the Updated Wave search Graph Bidirectional Convolutional Neural Network (UWGBCNN), a novel approach designed to enhance smart contract security. UWGBCNN integrates a multilabel vulnerability classification mechanism, utilizing the Updated Wave Search Algorithm (UWSA) to efficiently analyze and identify patterns in smart contracts by adapting network parameters to detect vulnerabilities with speed and precision. Additionally, feature extraction is enhanced through the Bidirectional Encoder Representations from Transformer (BERT) language model, incorporating supplementary word embedding features. The proposed technique achieves superior performance, reaching a precision of 98.5%, recall of 98.6%, and an F1-score of 99.6%, surpassing current methods. This approach contributes significantly to blockchain security by minimizing financial risks associated with vulnerabilities in decentralized applications.