Abstract:

Abstract Smart Contract Attack Detection Using Graph Convolution Network (GCN) is a research area that focuses on identifying and preventing malicious activities within smart contracts deployed on blockchain platforms. Smart contracts are self-executing digital agreements that run on decentralized networks, such as Ethereum. While smart contracts provide transparency and automation, they can also be vulnerable to various attacks, leading to financial losses or system disruptions. To address this challenge, the concept of Graph Convolution Network is leveraged. GCN is a deep learning technique that operates on graph-structured data, where nodes represent entities, and edges represent relationships between them. In the context of smart contracts, a graph can be constructed to capture the dependencies between different functions, variables, and transactions within the contract. The goal of utilizing GCN in smart contract attack detection is to learn patterns and detect anomalies in the graph structure. By training the model on a large dataset of known secure and malicious smart contracts, it can learn to identify suspicious patterns that might indicate an ongoing attack. The GCN model can consider features such as function calls, control flow, and data dependencies to detect potential vulnerabilities or abnormal behavior. In this paper we are going to address the detection of reentrancy attack, timestamp dependence attack and infinite loop attack using Graph Convolution Network. Smartbugs wild dataset is used for performing the attack detection. By using GCN we are able to detect these attacks accurately and our model is compared with the existing models and it shows that our model is better than the existing models in terms of performance metrics.

Keywords: