Classifier Categories


Categories form the core of the Redwood filter engine. Redwood classifies page requests into Categories, which are assigned Actions of Block / Allow, etc.

A Category is a group of Patterns that are matched in the Request or Response of HTTP requests. These patterns include Domains and URL Paths matching against URLs, and Phrases matching against the Response body.

After an HTTP request is processed, Redwood tallies up the total score of all patterns found, and attaches the total Category scores of all Categories found in the request. If the Category score exceeds the minimum threshold (200 by default) then that category is added to the list of ACLs that match the request.

Redwood then applies the action of the top-scoring Category to the page request. If the action is Allow, the user sees the page. If the action is block, the user sees the block page instead.


Classification Accuracy

Categories should consist of as many relevant patterns as possible, with lower scores rather than a few patterns with high scores.

Patterns can also include negative scores, when a pattern clearly does NOT apply. For example, the term stripper could be used in an Adult category, with subtracting patterns for wire stripper, sulfur stripper, and varnish stripper.


Builtin Categories

The Log Cabin Console includes nearly 200 regularly maintained Classifier Categories by default.S uch Categories are available under the Builtin Categories menu.


Revised Categories

If the user has adequate permissions to do so, he may add additional patterns to the Builtin Categories. Such Categories are available under the Revised Categories menu.


Console Categories

If the user has adequate permissions to do so, he may create entirely new Categories. Such categories are available under the Console Categories menu.


ACL Categories

ACL Categories differ from those mentioned above in one significant way. For an ACL Category to be added the list of ACLs that match an HTTP request, its score only needs to exceed the minimum threshold.

This makes ACL categories useful for creating Whitelists and Blacklists where the administrator intends a site to always Allowed or Blocked, regardless of the scores of Classifier Categories.