DARPA Cyber Grand Challenge Binaries

该数据集包含了200个独立的程序，这些程序是作为2016年发起的一项挑战的一部分发布的，旨在创建用于查找、验证和修补漏洞的工具。每个程序都包含一个或多个由人编写的错误，以模拟常见的开发者错误。该数据集已被用于对类似AFL的工具进行基准测试，并包含了不同复杂度的程序，每个程序都设计有特定的错误。规模上，该数据集共有200个程序（其中使用了24个程序的子集）。其任务是对模糊测试工具以及程序行为预测模型进行基准测试。This dataset contains 200 standalone programs, which were released as part of a challenge launched in 2016 aimed at developing tools for detecting, verifying, and patching software vulnerabilities. Each program includes one or more human-written bugs to simulate common developer mistakes. This dataset has been utilized for benchmarking AFL-like fuzz testing tools, and encompasses programs of varying complexity, with each program designed to feature specific bugs. In terms of scale, the dataset totals 200 programs, with a commonly utilized subset of 24 of these programs. The core objective of this dataset is to benchmark both fuzz testing tools and program behavior prediction models.

搜集汇总

背景与挑战

背景概述

该数据集是DARPA Cyber Grand Challenge Binaries，包含专门设计的自定义程序，这些程序模拟真实软件复杂性，嵌入多种崩溃软件漏洞，用于测试漏洞发现和补丁工具。数据集已移植到Linux、OS X和Windows平台，提供完整的构建和测试脚本，是评估程序分析工具性能的基准数据集。