devign-ast图

devign数据集没找到其ast图，用clang生成了它的ast结构，包含27000多个（全部）devign文件，及其对应的标签与描述，可用于进行漏洞检测任务的评估数据集。该数据集上目前最佳的效果在66%左右（包括acc、pre、recall、f1等），本人目前尝试探索更高的精度及更好的检测方案及可解释模型。

No AST graphs were found for the Devign dataset. Therefore, we utilized Clang to generate the AST structures for all 27,000+ Devign source files, along with their corresponding labels and descriptions. This dataset can serve as an evaluation dataset for vulnerability detection tasks. The current state-of-the-art performance on this dataset stands at approximately 66%, covering metrics including accuracy, precision, recall, F1 score and others. Currently, I am exploring methods to achieve higher accuracy, better detection solutions and interpretable models.

搜集汇总

背景与挑战

背景概述

该数据集名为'devign-ast图'，包含27000多个由clang生成的devign文件的AST结构及其对应的标签与描述，用于漏洞检测任务的评估。目前在该数据集上的最佳检测效果约为66%（包括acc、pre、recall、f1等），旨在探索更高的精度和更好的检测方案。