Phishing Website Features for Security Classification

PhishingWebsites is a dataset from OpenML containing features extracted from URLs and web pages for identifying phishing sites. It is used for training and benchmarking machine learning models in cybersecurity. The dataset's creator and specific temporal coverage are not provided.

Use Cases

• Train a binary classifier to predict the 'phishing' label using URL-based features like length and special character count.
• Analyze feature importance, such as 'ssl_final_state' or 'domain_age', to understand key indicators of malicious websites.
• Benchmark model performance on security tasks by using the dataset's standard train/test splits from OpenML.
• Develop anomaly detection models to identify phishing sites based on deviations in page content features.

Strengths

• Standardized dataset available on OpenML, facilitating direct model benchmarking.
• Contains curated features relevant to web security and phishing classification.

Limitations

• Unknown row count and feature completeness limit assessment of statistical power.
• Lack of temporal and geographic metadata may reduce applicability to current or region-specific threats.

Provenance

Source

OpenML