DeepSecure: Novel Python Vulnerability Detection Datasets for Broken Authentication and Hard-coded Credentials

This repository provides novel vulnerability detection datasets for two Python vulnerability types not previously benchmarked in prior work: Broken Authentication and Hard-coded Credentials. The datasets were collected using a commit-based weak supervision methodology adapted from VUDENC, with newly designed keyword sets to identify security-related commits from GitHub. These datasets were developed for the DeepSecure framework to extend benchmarking coverage beyond the four vulnerability categories commonly used in existing Python vulnerability datasets. Files are named following the VUDENC convention (plain_*.json) for direct compatibility with the DeepSecure codebase. The dataset is released under the CC BY 4.0 license to support open research and reproducibility.