The files in this repo will be used for my "Machine Learning with Phishing" series on my blog, please refer to the original blog posts to have a better understanding of this project.
The dataset used contains the 10 'baseline features' that were selected in this study.
Here are the features used for this project:
| No. | Identifier | Value type | Description |
|---|---|---|---|
| 1 | NumDash | Discrete | Counts the number of "-" in webpage URL. |
| 2 | NumNumericChars | Discrete | Counts the number of numeric characters in the webpage URL. |
| 3 | NumSensitiveWords | Discrete | Counts the number of sensitive words (i.e., "secure", "account", "webscr", "login","ebayisapi", "signin", "banking", "confirm") in webpage URL. |
| 4 | PctExtHyperlinks | Continuous | Counts the percentage of external hyperlinks in webpage HTML source code. |
| 5 | PctNullSelfRedirectHyperlinks | Continuous | Counts the percentage of hyperlinks fields containing empty value, self-redirect value such as "#", the URL of current webpage, or some abnormal value such as "file://E:/". |
| 6 | FrequentDomainNameMismatch | Binary | Checks if the most frequent domain name in HTML source code does not match the webpage URL domain name. |
| 7 | SubmitInfoToEmail | Binary | Check if HTML source code contains the HTML "mailto" function. |
| 8 | PctExtResourceUrlsRT | Categorical | Counts the percentage of external resource URLs in webpage HTML source code. Apply rules and thresholds to generate value. |
| 9 | ExtMetaScriptLinkRT | Categorical | Counts percentage of meta, script and link tags containing external URL in the attributes. Apply rules and thresholds to generate value. |
| 10 | PctExtNullSelfRedirectHyperlinksRT | Categorical | Counts the percentage of hyperlinks in HTML source code that uses different domain names, starts with "#", or using "JavaScript ::void(0)". Apply rules and thresholds to generate value. |