DarkPT-BR: Labeled Posts from Brazilian Portuguese Dark Web Forums

DarkPT-BR is a labeled dataset containing posts collected from Brazilian Portuguese forums on the Dark Web. It supports research in Cyber Threat Intelligence (CTI) and malicious content detection. The dataset includes binary and probabilistic labels for relevance, based on the presence of Indicators of Compromise (IoCs), contextual keywords, and manual expert analysis. It is structured in three versions: Dataset I: Initial labeling via heuristics (IoCs + keywords). Contains 17,675 posts from the Dark Web (1,665 Relevant, 16,010 Not Relevant). Dataset II: Manually revised and expanded version. Contains 26,575 posts from the Dark Web (3,341 Relevant, 23,234 Not Relevant). Note: All posts from Dataset I are included in Dataset II. Dataset III: Contains 7,498 previously unlabeled posts from the Dark Web. The labels for these posts were assigned through the developed post identification model, assigning a probability between 0 and 1 for the relevance of the post. This dataset can support machine learning research in threat classification, text mining, and cybersecurity automation. The code associated with the data can be found in https://github.com/sebastiaoafilho/Malicious_Posts_IdentificationDarkPT-BR 是一款带标注的数据集，其数据采集自暗网（Dark Web）上的巴西葡萄牙语论坛帖文，可支撑网络威胁情报（Cyber Threat Intelligence, CTI）与恶意内容检测领域的研究。该数据集基于失陷指示器（Indicators of Compromise, IoCs）、上下文关键词以及专家手动分析结果，为帖文的相关性标注了二元标签与概率标签。 该数据集共包含三个版本： 数据集I：通过启发式方法完成初始标注（结合失陷指示器与关键词），涵盖17675条暗网帖文，其中相关帖文1665条、非相关帖文16010条。 数据集II：经手动修订与扩充的版本，包含26575条暗网帖文，其中相关帖文3341条、非相关帖文23234条。注：数据集I的全部帖文均已纳入数据集II。 数据集III：包含7498条此前未标注的暗网帖文，其相关性标签通过自研的帖文识别模型生成，为每条帖文赋予0至1区间内的相关性概率值。 该数据集可支撑威胁分类、文本挖掘以及网络安全自动化方向的机器学习研究。相关配套代码可通过https://github.com/sebastiaoafilho/Malicious_Posts_Identification 获取。