Phishing Websites Dataset for Tabular ML IID Study, 2012

A dataset curated for the TabArena Tabular ML IID Study, intended for evaluating predictive models on independent and identically distributed tabular data. The data originates from a 2012 IEEE study on phishing website features and is licensed under CC BY 4.0. All features are categorical, ordinal-encoded variables with at most 3 values.

Use Cases

• Benchmarking classification models for IID tabular data based on the dataset's stated focus.
• Training phishing website detectors based on categorical features related to phishing characteristics.
• Studying the impact of duplicate data on model performance based on the noted 47% duplication rate.
• Analyzing feature engineering for ordinal-encoded categorical variables with limited value ranges.

Strengths

• All features are categorical, ordinal-encoded variables with at most 3 values, which may simplify preprocessing.
• Has a clear, open license (CC BY 4.0) and a cited academic reference for provenance.
• Was specifically curated for a systematic study on tabular ML, suggesting a defined evaluation purpose.

Limitations

• The dataset contains many duplicates (47%), which may affect model training and evaluation.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and sample data are unknown, which limits suitability assessment.

Provenance

Source

Mohammad, Rami M., Fadi Thabtah, and Lee McCluskey. 'An assessment of features related to phishing websites using an automated technique.' 2012 international conference for internet technology and secured transactions. IEEE, 2012.

Collection Method

Curated by the TabArena team from the original source for the TabArena Tabular ML IID Study.

Time Range

2012

Freshness

Dataset Year is 2012; last update date is unknown.

Geography

null