SpamBully Review Overall Rating:

Description

SpamBully is a Bayesian based filter that integrates with Microsoft Outlook and Outlook Express. In addition to the Bayesian engine, it has additional Allow/block lists for basic keyword filtering, country blocking, language blocking and RBL lookup facility. It can automatically send fake "Undeliverable" messages to spam and can be configured for full challenge/response operation.

Verdict

We were very impressed with SpamBully, in terms of spam capture rate, low false positive/negatives and pleasant user interface.

Installation

Installation went smoothly. During installation, SpamBully asks for the location of both good and bad e-mail messages so that it can start "learning" from them. We see this as a good feature as opposed to just starting with a "pre-trained" Bayesian engine as it learns on the kind of spam that you get and exactly what it is you consider spam.

You are kept informed of progress by a useful (although innacurate time-wise), progress form:

One thing we did notice was that during installation, it offered country-specific blocking and language-specific blocking, with no explanation as to what these things were exactly. This did not become clear until we had finished intallation and reviewed the options.

Interface

Like most filters of this kind, SpamBully operates as an "Outlook add-in" and adds a few buttons to the standard interface of the program:

The interface is very clean, polished and intuitive with a very useful statistical graph which monitors your e-mail activity and its own performance:

One gripe we did have was the help files - or lack thereof. No help was installed with the application and instead, clicking "help" on the Spam Bully toolbar opened up a browser and connected to Spam Bully's web site. People on dial-up connections will not appreciate having to connect to the internet just to be able to view help on the program. The help system on the internet at first seemed quite cluttered and unappealing. Perseverance reavealed some quite useful animated tutorials.

In it's defence, Spam Bully's ease of use and intuitiveness makes not having help files readily to hand much less of a problem than it could be!

Features and Operation

Like many filters that integrate into Outlook/Outlook Express, operation could not be simpler. Once the Bayesian engine has learnt from a sufficient number of known good and bad e-mails, then it simply works away classifying new messages and learning from new messages as they arrive.

Should a message be incorrectly identified, simply click the "Spam" or "Not Spam" button and Spam Bully will be forced to "learn by it's mistake".

We loved the "Message Details" feature where selecting a message and clicking "Message Details" on the toolbar will bring up the following screen showing why the message was classified as spam and showing the sender's IP address and location on trick little world map! - nice.

One feature we question is the "Challenge Email", although in principle we believe the concept of challenge/response to be of great value. Spam Bully's implementation of challenge/Response will send an e-mail to the sender of a message that passes the Bayesian filtering, but does not come from a contact in your friends list. (Sending challenges to messages it doesn't think is spam?)
From the outset, it would seem that the better idea would be to send a challenge to the sender of messages that don't pass the filtering. I would guess that this step was not taken in the name of good "netizenship" - flooding the internet with challenges.
However, one of the advantages to Bayesian filtering over basic content-based filters is that the result doesn't have to be "spam" or "not spam". There could be any number of categories. A "nice to have" would be to class messages as either "good", "Possible Spam" and "Spam". With this range, it would make sense to send challenges to "Possible Spam" messages only. This would take us one step nearer to the "Holy Grail" of spam filters - where we no longer have to check the "Junk Mail" folder.

SpamBully has implemented a feature that would seem to hold a lot of promise for future spam detection. It can optionally follow links in an e-mail message and perform Bayesian analysis on the web site that the message links to.
This will handle the cases where a spam email contains little more than a link to a web site and seems to be a good feature to have given that the spammers will openly "give themselves away" at the web site.

Accuracy

All the slick interface and nice features of a program such as Spam Bully would mean nothing if the accuracy of the filtering didn't stand up to the test.

Fortunately, SpamBully most decidedly delivered the goods in this department as well with an outstanding rate of spam detection with very minimal "falses".

We gave it the hardest test we could by forcing it to rely solely on its Bayesian analysis and minimal country blocking for detection. We removed all contacts from the friends list. Despite this, it returned the following results:

Conclusion

Spam Bully gets a thumbs up for us for impressive statistics, innovative features and going the "extra mile" with the user interface. A hearty recommendation.

Return to Review Index