Dataset Card for WebFacets 0.1.0

web-facets is a large-scale, fine-grained collection of ~12 million web domains labelled along two axes: cybersecurity and source reliability. It is extracted from domain-pool.

Dataset Details

web-facets contains 2 main categories along which domains are rated: cybersecurity and epistemic reliability. Datapoints are all also attributed the year(s) of the dataset(s) in which they appear, the website's functional category, and each label's source.

The yearly distribution is illustrated below. Multiple large original data sources are `live' datasets, explaining the 2026 skew.

Dataset Description

Label Attributes

Functional Website Category

Website categories are mapped to one of two categories when applicable:

Cybersecurity	Epistemic Reliability
malware, adware, adware/malware, malicious, phishing, benign, blacklisted, security_threats, abuse_tlds, defacement, command and control, redirection, redirector, dynamic-dns, dynamic_dns, shortener, cryptojacking, provider	fake news, coordinated campaign, impersonation, spoofing, questionable, unreliable, misleading, conspiracy, propaganda, false satire, pseudoscience, parody site, imposter site, fact-checker, reliable

Cybersecurity

Epistemic (Source) Reliability

Reliability labels exist under three modalities:

For continuous reliability scores,

For binary reliability flags, which are sourced either from original 'reliable'/'unreliable' labels or by mapping domains with extreme bias or factuality scores to their respective category,

For categorical labels pertaining to reliability,

Data sources

Some of the primary contributors to the dataset are:

• UT1 by the University of Toulouse Capitole (47.7%),
• Blacklists (40.1%);
• DNS Blocklists (16.0%)
• 50+ others with <10% each.

The full list:

Source	Rows	% of Total
UT1	5,616,544	47.7%
Blacklists	4,718,888	40.1%
Dns Blocklists	1,882,315	16.0%
Malicious And Benign Webpages Train	746,082	6.3%
Phish DB	496,270	4.2%
Radek Benign C	436,807	3.7%
Radek Benign U	335,031	2.8%
Malicious And Benign Webpages Test	247,537	2.1%
Benign Malicious Urls	210,492	1.8%
Malicious Urls	192,961	1.6%
Radek Phishing	163,499	1.4%
Url Phish	102,745	0.87%
Radek Malware	94,823	0.81%
HOSTS (Porn)	64,217	0.55%
Legitphish	63,944	0.54%
Hosts Adware Malware	58,834	0.50%
Phish Dataset	44,102	0.37%
Urlhaus	28,755	0.24%
Domains Quality Ratings	11,480	0.0976%
Meta Threat Reports	6,325	0.0538%
Redirection Domains	5,725	0.0487%
Misinfodomains	4,738	0.0403%
HOSTS (Gambling)	4,469	0.0380%
MBFC Ratings	4,435	0.0377%
Wikipedia General	3,372	0.0287%
Manual	2,198	0.0187%
Hosts Fake News	2,180	0.0185%
Iffy Index	2,005	0.0170%
MBFC S Questionable List	1,881	0.0160%
FakeNewsNet	1,108	0.0094%
CheckThat!	1,067	0.0091%
Wikipedia Miscellaneous	822	0.0070%
Wikipedia Campaigns	821	0.0070%
Providers	650	0.0055%
Wikipedia Fake News	454	0.0039%
Zoznam	337	0.0029%
Politifact's Almenac	325	0.0028%
Dictionaries Manual	297	0.0025%
Reliable Health Resources	223	0.0019%
Reliable Legal Resources	183	0.0016%
Paperwall	123	0.0010%
Sd22 Approved Software List	117	0.0010%
Ngo Report Uae Blacklist	99	0.0008%
Ngo Report Israel Blacklist	98	0.0008%
Providers (Manual)	92	0.0008%
Ngo Report Saudi Blacklist	85	0.0007%
Wikipedia II Actors	62	0.0005%
Nelež	51	0.0004%
Tools Manual	30	0.0003%
C2 Domains	20	0.0002%
Hasbara Tracker	19	0.0002%
EDMO Hubs	16	0.0001%
Ngo Report Russia Blacklist	4	0.0000%

• Curated by the CrediNet organisation, which consists of a team of collaborators from the Complex Data Lab @ Mila - Quebec AI Institute, the University of Oxford, McGill University, Concordia University, UC Berkeley, University of Montreal, and AITHYRA.
• Funding: This research was supported by the Engineering and Physical Sciences Research Council (EPSRC) and the AI Security Institute (AISI) grant: Towards Trustworthy AI Agents for Information Veracity and the EPSRC Turing AI World-Leading Research Fellowship No. EP/X040062/1 and EPSRC AI Hub No. EP/Y028872/1. This research was also enabled in part by compute resources provided by Mila (mila.quebec) and Compute Canada.

Dataset Card Contact

For questions, comments, or bug reports, please contact:

• Emma Kondrup