This product analysis compares ufdbGuard for Squid with other products.

Performance

ufdbGuard is the fastest URL filter with an amazing 60,000 URL verifications/sec on a single processor core (Intel Haswell at 2.9 GHz) running Linux.  Competitor’s products on the same hardware platform competitors only perform 20,000 URL verifications/sec.

Price and ROI

The license to use the URL database is very reasonable, while the ufdbGuard software is free.  This makes the web filtering solution of URLfilterDB extremely cost-effective and with an exceptional ROI.

Extensive set of additional features

In addition to the regular URL filtering capabilities, ufdbGuard offers a set of useful extra features to protect your network even better:

• dynamic, automatic HTTPS proxy detection;
• HTTPS usage control;
• advertisement blocking;
• blocking adult images produced by search engines.

Features and Updates

ufdbGuard has on average 2 release per year and new features are added with every release.  The ufdbGuard API is developed to be incorporated in 3rd party applications.  ufdbGuard 1.31-9 supports the new redirector API of Squid 3.4.x.

Superior filtering results

There are three methods available to block unwanted web content:
A. Content scanning: this method blocks access to web pages based on the occurrence of “bad” words in the content.
B. Artificial intelligence: a variant on content scanning, intended to render more accurate results.
C. Blacklists: this method blocks access to web pages based on their being listed in a website category to be blocked.

To compare these methods to each other, the following criteria can be used:

• User experience: performance must be sufficient for the individual user.
• Overblocking: sites that users should be allowed access to should not be blocked (e.g. sites about breast cancer should not be classified as sites with adult content).
• Underblocking: a filter should not fail to block a site that users should not have access to.
• HTTPS blocking: a filter should be able to handle web site content encrypted with HTTPS.
• Infrastructure costs: a filter should minimize the costs incurred by the bandwidth usage and computing power required by the filtering software.

Method A, content scanning, works by scanning web page content for “bad words”.  As scanning requires downloading the pages first, this consumes extra time and bandwidth.  Moreover, the method is prone to overblocking because word combinations like “breast cancer” may easily lead to incorrect determination of the web page as unwanted content.  On the other hand, there’s a real chance of underblocking as well, since this method cannot detect unwanted content that takes the form of a picture.  Since the method is time and resource consuming, a very fast computer system is required for the proxy server.

Method B, artificial intelligence (AI), aims to improve content scanning as a method.  AI also downloads web pages first before scanning them, consuming extra time and bandwidth.  AI usually rates both individual words and some word combinations in a page to reduce the overblocking and underblocking found in content scanning methods.  Some products may even try to use color detection to find out if a picture contains nudity.  The more correct the blocking results of such filters are, the more CPU power is usually required.  For 250 users and more, this requires a very fast computer system for the proxy server.

Method C, blacklist blocking, does not require web page downloading to decide whether or not to block it.  Instead, the URL filter module on the proxy server makes a quick decision based on a list of URLs to be blocked.  Usually this list, known as a blacklist, is stored in a database.  This method is fast and efficient, and requires minimal computing power.  The table below clearly shows how the methods compare to each other, and why the blacklist method is the best choice.

Method	User experience	Correct blocking	HTTPS blocking	Scalability beyond 1000 users	Infrastructure costs
Content scanning	±	–	– –	–	±
Artificial intelligence	±	±	– –	– –	–
Blacklist (ufdbGuard)	+ +	+	+ +	+ +	+