Aligned Bytecode Vulnerability Pattern Dataset
收藏资源简介:
该数据集由哈尔滨工程大学烟台研究院、西电杭州研究院、深圳北理莫斯科大学工程系和浙江大学计算机科学与技术学院的研究人员创建。数据集通过将源代码模式映射到操作码级别的对应物来构建,是首个与源代码定义对齐的漏洞模式数据集。它为智能合约漏洞检测提供了细粒度的监督,并指导模型关注漏洞关键区域。数据集的创建旨在解决智能合约漏洞检测中字节码语义不透明、粗粒度图嵌入、字节码中缺乏漏洞模式的正式化和注释,以及图级对齐的局限性等问题。
This dataset was created by researchers from Yantai Research Institute, Harbin Engineering University, Hangzhou Research Institute, Xidian University, Department of Engineering, Shenzhen MSU-BIT University, and College of Computer Science and Technology, Zhejiang University. Constructed by mapping source code patterns to their opcode-level counterparts, this is the first vulnerability pattern dataset aligned with source code definitions. It provides fine-grained supervision for smart contract vulnerability detection and guides models to focus on critical vulnerability regions. This dataset was developed to address the limitations in smart contract vulnerability detection, including the semantic opacity of bytecode, coarse-grained graph embedding, lack of formalization and annotation of vulnerability patterns in bytecode, and graph-level alignment issues.
数据集概述
基本信息
- 数据集名称:Smart Contract Bytecode Graph Constructor and expert pattern extrctor
- 主要功能:从智能合约字节码(操作码)构建图结构,并使用专门设计的专家模式对节点进行标注
核心特性
- 字节码到图的转换:直接从以太坊字节码构建控制流图(CFGs)
- 专家模式标注:实现针对特定漏洞检测的专用模式,包括:
- 时间戳依赖性
- 重入漏洞
- 无限循环
技术实现
专家模式实现
将引用论文中的专家模式扩展到直接处理字节码:
时间戳依赖模式 检测包含TIMESTAMP或BLOCKHASH操作码的块及其数据依赖的后继块
重入模式 识别遵循检查-效果-交互模式的危险调用序列
无限循环模式 通过跳转分析检测潜在的无限循环
输出格式
生成的图遵循以下JSON结构:
{ "nodes": ["block_1", "block_2", ...], "edges": [[0, 1], [1, 2], ...], "target_nodes": [3, 5, ...] }
学术基础
基于以下源代码方法的研究成果:
- [IJCAI 2021] Smart Contract Vulnerability Detection: From Pure Neural Network to Interpretable Graph Feature and Expert Pattern Fusion
- [TKDE 2021] Combining Graph Neural Networks with Expert Knowledge for Smart Contract Vulnerability Detection




