credi-grain
收藏资源简介:
CrediGrain是一个大规模、细粒度的网络域名数据集,按照网络安全、来源可靠性和流行度三个维度进行分类。该数据集基于domain-pool构建,整合了来自50多个不同来源的数据,主要来源包括UT1(占34.6%)、Tranco List(占30.5%)、黑名单(占29.1%)和DNS阻止列表(占11.6%)。数据集总规模庞大,涵盖了数百万个域名条目,每个条目都附带了细粒度的分类标签。该数据集由CrediNet组织策划,该组织由来自Mila - Quebec AI Institute、牛津大学、麦吉尔大学、康考迪亚大学、加州大学伯克利分校、蒙特利尔大学和AITHYRA等机构的研究人员组成。资金支持来自英国工程和物理科学研究理事会(EPSRC)和人工智能安全研究所(AISI)等机构。该数据集适用于网络安全分析、来源可靠性评估、网络流行度研究以及相关机器学习任务,如域名分类和威胁检测。
CrediGrain is a large-scale, fine-grained web domain dataset categorized along three dimensions: cybersecurity, source reliability, and popularity. Built on a domain-pool, the dataset integrates data from over 50 distinct sources, with its primary sources being UT1 (34.6%), Tranco List (30.5%), blacklists (29.1%), and DNS blocklists (11.6%). It has a massive overall scale, encompassing millions of domain entries, each paired with fine-grained classification labels. Curated by the CrediNet organization, which comprises researchers from institutions including Mila - Quebec AI Institute, University of Oxford, McGill University, Concordia University, University of California, Berkeley, Université de Montréal, and AITHYRA, the dataset is funded by organizations such as the UK Engineering and Physical Sciences Research Council (EPSRC) and the AI Safety Institute (AISI). This dataset is applicable to cybersecurity analysis, source reliability assessment, web popularity research, and related machine learning tasks including domain classification and threat detection.
数据集卡片:CrediGrain
CrediGrain (credi-grain) 是一个大规模、细粒度的网络域名集合,从 domain-pool 中提取,并沿着三个主要维度进行标注。
核心维度
数据集对每个域名从以下三个方面进行评级:
- 网络安全 (Cybersecurity)
- 认知可靠性 (Epistemic Reliability)
- 流行度 (Popularity)
此外,每个数据点还包含其出现的年份、网站的功能类别以及每个标签的来源信息。
数据来源
数据集整合了多个来源的数据,主要贡献者包括:
| 来源 | 行数 | 占总数据百分比 |
|---|---|---|
| UT1 (图卢兹第一大学) | 5,617,370 | 34.6% |
| Tranco List | 4,944,553 | 30.5% |
| Blacklists | 4,718,888 | 29.1% |
| Dns Blocklists | 1,882,315 | 11.6% |
| Malicious And Benign Webpages Train | 755,500 | 4.7% |
| Phish DB | 496,270 | 3.1% |
| Radek Benign C | 436,807 | 2.7% |
| Radek Benign U | 335,031 | 2.1% |
| Malicious And Benign Webpages Test | 250,568 | 1.5% |
| Benign Malicious Urls | 210,492 | 1.3% |
| Malicious Urls | 192,961 | 1.2% |
| Radek Phishing | 163,499 | 1.0% |
| Url Phish | 102,745 | 0.63% |
| Radek Malware | 94,823 | 0.58% |
| Hosts Porn | 64,217 | 0.40% |
| Legitphish | 63,944 | 0.39% |
| Hosts Adware Malware | 58,834 | 0.36% |
| Phish Dataset | 44,102 | 0.27% |
| Urlhaus | 28,755 | 0.18% |
| Domains Quality Ratings | 11,480 | 0.0708% |
| Meta Threat Reports | 6,325 | 0.0390% |
| Redirection Domains | 5,725 | 0.0353% |
| MBFC | 5,447 | 0.0336% |
| Misinfodomains | 4,738 | 0.0292% |
| Hosts Gambling | 4,469 | 0.0276% |
| Wikipedia General | 3,372 | 0.0208% |
| Manual | 2,198 | 0.0136% |
| Hosts Fake News | 2,180 | 0.0134% |
| Iffy Index | 2,005 | 0.0124% |
| FakeNewsNet | 1,108 | 0.0068% |
| CheckThat! | 1,067 | 0.0066% |
| Wikipedia Miscellaneous | 822 | 0.0051% |
| Wikipedia Campaigns | 821 | 0.0051% |
| Providers (Manual) | 650 | 0.0040% |
| Providers Manual | 650 | 0.0040% |
| Wikipedia Fake News | 454 | 0.0028% |
| Zoznam | 337 | 0.0021% |
| Politifacts Almenac | 325 | 0.0020% |
| Dictionaries Manual | 297 | 0.0018% |
| Reliable Health Resources | 223 | 0.0014% |
| Sd22 Approved Software List | 193 | 0.0012% |
| Reliable Legal Resources | 183 | 0.0011% |
| Paperwall | 123 | 0.0008% |
| Ngo Report Uae Blacklist | 99 | 0.0006% |
| Ngo Report Israel Blacklist | 98 | 0.0006% |
| Ngo Report Saudi Blacklist | 85 | 0.0005% |
| Wikipedia II Actors | 62 | 0.0004% |
| Nelež | 51 | 0.0003% |
| Tools Manual | 48 | 0.0003% |
| C2 Domains | 20 | 0.0001% |
| Hasbara Tracker | 19 | 0.0001% |
| EDMO Hubs | 16 | 0.0001% |
| Ngo Report Russia Blacklist | 4 | 0.0000% |
数据集构建与资助
- 策划方: CrediNet 组织,合作者来自 Mila - 魁北克人工智能研究所、牛津大学、麦吉尔大学、康考迪亚大学、加州大学伯克利分校、蒙特利尔大学及 AITHYRA。
- 资助: 本研究由工程与物理科学研究理事会 (EPSRC) 和人工智能安全研究所 (AISI) 的资助支持。
许可
- 许可证: CC-BY-4.0
联系方式
如有问题、意见或错误报告,请联系:
- Emma Kondrup:emma.kondrup@mila.quebec




