|
|
--- |
|
|
license: cc-by-4.0 |
|
|
--- |
|
|
|
|
|
# Dataset Card for *Domains by Reliability (DomainRel)* |
|
|
|
|
|
|
|
|
DomainRel is a large-scale dataset of web domains labelled binarily as per their overall reliability level. |
|
|
The domains span areas of general knowledge, (mis)information, phishing and malware. |
|
|
DomainRel is an aggregate and reconstruction of existing datasets originating from each of these specific areas, offering the first cross-domain dataset of this type and scale. |
|
|
|
|
|
|
|
|
## Dataset Details |
|
|
|
|
|
DomainRel can be downloaded in its simplest form (`labels.csv`), which contains 674,737 with a binary reliability score with the following class distribution: |
|
|
``` |
|
|
Unreliable (0): 361,537 |
|
|
Reliable (1): 313,199 |
|
|
``` |
|
|
|
|
|
An alternative `labels_annot.csv` is also available; it contains the same domain data, with disaggregated per-source scores tracing the provenance of each datapoint as per its original data source. |
|
|
|
|
|
- **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 Sources |
|
|
|
|
|
Refer to the below for a more detailed breakdown of the dataset's class distribution in terms of dataset of origin. |
|
|
|
|
|
|
|
|
**Total:** |
|
|
|
|
|
```bash |
|
|
0: 361,537 |
|
|
1: 313,199 |
|
|
``` |
|
|
|
|
|
which consists of the following datasets: |
|
|
|
|
|
### LegitPhish |
|
|
|
|
|
[LegitPhish](https://data.mendeley.com/datasets/hx4m73v2sf/2) is a URL-based dataset with binary labels (0 = Phishing, 1 = Legitimate). The data is processed by converting URLs to their domain name, averaging the scores of all URLs that resolve to the same domain, and then binarizing based on a threshold of 0.5. We also standardise the header (we always use `domain` and `label` in the processed csvs.) |
|
|
|
|
|
```bash |
|
|
0: 26,957 |
|
|
1: 37,113 |
|
|
``` |
|
|
|
|
|
### PhishDataset |
|
|
|
|
|
[PhishDataset](https://github.com/ESDAUNG/PhishDataset/blob/main/data_imbal%20-%2055000.xlsx) is a URL-based dataset with binary labels (0 = Legitimate, 1 = Phishing) |
|
|
|
|
|
```bash |
|
|
0: 3,730 |
|
|
1: 40,535 |
|
|
``` |
|
|
|
|
|
### Nelez |
|
|
|
|
|
[Nelež](https://www.nelez.cz) is a blacklist of misinformation websites from a Czech organisation of the same name. |
|
|
|
|
|
```bash |
|
|
0: 51 |
|
|
1: 0 |
|
|
``` |
|
|
|
|
|
### Wikipedia |
|
|
|
|
|
[Wikipedia](https://github.com/kynoptic/wikipedia-reliable-sources) is an aggregate set of a reliability ratings from multiple Wikipedia sources. The data is processed to keep only 'boosted' and 'discarded' domains (corresponding to 1 and 0 respectively -- they also have a neutral category we discard). |
|
|
|
|
|
```bash |
|
|
0: 1,029 |
|
|
1: 2,906 |
|
|
``` |
|
|
|
|
|
### URL-Phish |
|
|
|
|
|
[URL-Phish](https://data.mendeley.com/datasets/65z9twcx3r/1) is a feature-engineered dataset for phishing detection. URLs have label 0 if they are considered benign, 1 for phishing. |
|
|
|
|
|
```bash |
|
|
0: 10,551 |
|
|
1: 92,460 |
|
|
``` |
|
|
|
|
|
### Phish & Legit |
|
|
|
|
|
[Phish & Legit](https://www.kaggle.com/datasets/harisudhan411/phishing-and-legitimate-urls?resource=download) is a URL Classification dataset of suspicous (0) and genuine (1) web addresses. |
|
|
|
|
|
```bash |
|
|
0: 292,163 |
|
|
1: 146,316 |
|
|
``` |
|
|
|
|
|
### Misinformation domains |
|
|
|
|
|
[Misinfo-domains](https://github.com/JanaLasser/misinformation_domains/tree/main) is a collection of domains labelled unreliable if they are assessed as spreaders of unreliable information. |
|
|
|
|
|
```bash |
|
|
0: 2,170 |
|
|
1: 2,597 |
|
|
``` |
|
|
|
|
|
### Malware domains |
|
|
|
|
|
[URLHaus](https://urlhaus.abuse.ch) is a collection of malware domains. |
|
|
|
|
|
```bash |
|
|
0: 31,540 |
|
|
1: 0 |
|
|
``` |
|
|
|
|
|
## Dataset Card Contact |
|
|
|
|
|
[Emma Kondrup](mailto:emma.kondrup@mila.quebec) |
