Cổng tri thức PTIT

Trang chủ

Giới thiệu

AI Cộng đồng

Kho tri thức

Tin tức

Liên hệ

Bài báo quốc tế

Kho tri thức

Bài báo quốc tế

When datasets deceive: Exposing overlap in smart contract vulnerability detection

Trần Tiến Công

Existing smart contract vulnerability datasets exhibit over 34% train–test overlap due to repeated function-level code, causing models to favor structural memorization over semantic generalization. To mitigate this issue, we construct a benchmark dataset with zero function overlap between the training and test partitions. Furthermore, we introduce GraphFusionDetect (GFD), a novel approach that integrates fine-tuned CodeBERT embeddings with Graph Neural Networks (GNNs) to capture inter-function dependencies. GFD achieves F1-scores of 80% for detecting reentrancy vulnerabilities and 89% for timestamp dependency vulnerabilities, surpassing baseline methods and enabling more robust and generalizable vulnerability detection.

Xuất bản trên:

When datasets deceive: Exposing overlap in smart contract vulnerability detection

Ngày đăng:

2026

DOI:

https://www.sciencedirect.com/science/article/pii/S2405959526000238

Nhà xuất bản:

ICT Express

Địa điểm:

Từ khoá:

Smart contracts; Vulnerability detection; Graph Neural Networks; CodeBERT; Dataset curation

Bài báo liên quan

A Two-Stage Agent-based Framework for Network Attack Detection And Categorization in IoT

Nguyễn Huy Trung

Anomaly-based intrusion detection leveraging optimized firewall log analysis: a real-time machine learning solution

Tran Cong Hung

Anomaly-based intrusion detection leveraging optimized firewall log analysis: a real-time machine learning solution

Tran Cong Hung

Improving the Web Crawling Accuracy with Machine Learning Based on Parsers Using Linguistic Structures

Nguyễn Minh Tuấn

ENHANCING THE EFFECTIVENESS OF ENCRYPTED TRAFFIC CLASSIFICATION THROUGH DATA PRESERVATION AND INPUT ALIGNMENT WITH DEEP NEURAL NETWORKS

Nguyễn Hồng Sơn

Context-Enriched Sentiment Analysis for Short Vietnamese Restaurant Reviews Using Large Language Models

Nguyễn Ngọc Điệp

Virtual Fusion with Contrastive Learning for Single Sensor-based Activity Recognition

Nguyễn Đức Anh