Classifier Overview


At its core, the Redwood engine should be considered a Classifier of text content. The first and most important question that that classifier attempts to answer is "What Subject does this body of text discuss?" On this answer all the rest follows - Filter Actions and Reporting. When we have a solid idea what Subject a body of text discusses, we can Act upon that knowledge by Allowing or Blocking the pages, and by Reporting on what Subjects are used.

Filter Actions


Each Category has a filter action assigned to it, which tells the Redwood filter engine what to do if when a page's top score matches that category.

The filter action options are:

  1. Allow
  2. Block
  3. Ignore

The first two are self-explanatory. Ignored categories are ignored in the sense Action - such categories never specifically Allow or Block a page. They're useful to de-activate a category and also provide useful Subject information for Usage Reports.


Category Ratings

So Redwood classifies text into categories - but what do those categories represent? The Security Appliance uses a Category Rating system to represent the tone of a category and represent it visually.


Category Scoring

Each Category is calculated independently. No category knows about any other category in the system. This can seem counterintuitive to those from other filter projects. For the filter administrator, main thing to keep in mind is that to for the filter to Act on a category (Allow or Block), the category must always be the top-scoring category of the HTTP Request or Response.

Note! ACL Categories are an exception to this. An ACL Category matches any HTTP request where the score exceeds the minimum threshold (200 by default), and is useful for creating Whitelists & Blacklists.


Classifying by Subject

The default categories classify by the Subject of the text content - what is the text talking about. When craigslist.org is classified by Subject the category is Auctions / Markets / E-Commerce. Sites like hp.com or dell.com are in the Computer category.


Classifying by Action

But Classifying by Subject is not the only way to think about sites. Classifying by Action is more intuitive to many filter administrators, who typically think of categories as tools to Allow or Block web pages, rather than find out what Subject the pages discuss. ACL Categories are especially suitable for this purpose.