OARAB: code and results for a benchmark and construction ablation of post-authentication resource abuse in OAuth 2.0

Code and result files for the paper "OARAB: A Benchmark and Construction Ablation for Post-Authentication Resource Abuse in OAuth 2.0" (Mohd Adnan and Aasim Zafar, Aligarh Muslim University). OARAB generates labelled streams of authenticated resource access and evaluates whether a detector can separate abusive clients from legitimate ones once a bearer token has already been issued. Principals are human, machine-to-machine, power user, or abusive, and the abusive ones come at three difficulty tiers. Only behaviour carries the label; identity signatures and OAuth attributes are assigned independently of it. The archive contains the generator, the detector panel and evaluation harness, the loaders for the external corpora, the result files every number in the paper is read from, and the test suite that runs the validity guards. It also contains the audit trail the paper's abstract promises: a record of eight claims withdrawn during the revision, four of them results the submitted version reported, each raced to the artefact that refutes it. The generated snapshot and the behavioural profile bank are not included. The bank is fit to the ITI web-bot corpus, which is CC BY-NC-SA and which we do not redistribute in derived form. Generation is deterministic from one seeded stream, so the released snapshot rebuilds identically after fetching the corpus and fitting the bank; README.md gives the two commands. Until you do, the tests that generate from the bank will fail, which is expected rather than a defect. Requires Python 3.10 or newer.本数据集对应论文《OARAB：OAuth 2.0认证后资源滥用基准测试与构建消融实验》（作者Mohd Adnan、Aasim Zafar，阿里格尔穆斯林大学）。 OARAB可生成带标注的已认证资源访问流，用于评估在承载令牌（bearer token）已签发的场景下，检测模块能否将滥用客户端与合法客户端区分开来。主体类型涵盖人类用户、机器对机器（M2M）场景、高级用户以及滥用主体，其中滥用主体分为三个难度层级。仅行为特征携带标注标签，身份签名与OAuth属性的分配与标注无关。本归档包包含生成器、检测模块面板与评估测试框架、外部语料加载器、论文中所有数值对应的结果文件，以及用于运行有效性校验的测试套件。 本归档同时包含论文摘要中提及的审计轨迹：修订过程中撤回的8项学术主张的记录，其中4项为提交版本中报告的实验结果，每项撤回的主张均配有可证伪其结论的对应实验制品。 生成快照与行为特征库未包含在本归档中。该行为特征库适配ITI网络机器人语料库（ITI web-bot corpus），该语料库采用知识共享署名-非商业性使用-相同方式共享许可协议（CC BY-NC-SA），我们不会以衍生形式重新分发该语料库。生成过程基于种子流确定执行，具有确定性，因此在获取对应语料库并适配特征库后，可复现与发布版本完全一致的快照；README.md文件中提供了这两条执行命令。在此之前，基于该特征库生成的测试将无法运行，此为预期行为，并非程序缺陷。 本数据集依赖Python 3.10及以上版本。