InBoxer Review Overall Rating:

Description

InBoxer is a Bayesian filter for Microsoft Outlook that is based on the famous SpamBayes project. It has Whitelist/Blacklist filtering and a unique plug-in architecture allowing additions to the base functionality.

Verdict

It is quite rare for us to get the opportunity to heartily enthuse about a product we are testing.
We have seen many filters - some better than others - and it has to be pretty special to bring a smile to these battle scarred and weary spam warrior's faces.
InBoxer is one of these rarities - we love it! Read on...

Installation

Installation went fine with no surprises.
Although InBoxer comes with a pre-configured database for the Bayesian engine (so that it can recognise good and bad messages "out of the box"), during the install process, it gives you the opportunity to learn from messages that you already have in your mail folders. This is a real bonus as it can get the filter working at optimum efficiency very quickly if you have a pre-existing store of good and bad messages.

Interface

Being a plug in for Outlook, the regular interface is very minimal indeed. At most it consists of "Block" and "Keep" buttons being added to a new toolbar.

There is also an icon added to the system tray, which changes as an new messages arrive. Holding your mouse over the icon will produce a summary of InBoxer's activity to date:

Just as you don't expect (or need) your favourite championship boxer to be pretty, InBoxer's interface is to the point, functional, and has the minimum of "glitz" and fuss.
Being a "died in the wool" old-school programmer, I do raise an eyebrow at such things as the lack of shortcut keys on the dialogs (these are the little underlined characters that allow you to select a control by holding the "alt" key and clicking the underlined character), but that does seem to be more and more prevalent nowadays.

On the whole, the interface lacks some of the polish of other offerings, but really that is of little importance when you consider that the dialogs are still clear and simple to use and especially when you consider the excellent filtering capabilities of this program.

Features and Operation

The nice thing about Outlook plug-in Bayesian filters, is that they are so easy to use in normal daily operations.
When a message is incorrectly identified (either a spam identified as good, or a good message identified as spam), then you simply click the relative "Keep" or "Block" button, to correct the mistake.

Unlike similar filters however, InBoxer sorts messages into three categories: "good" mail, "blocked" (bad) mail, and "Review".
Messages that it considers are certainly good go directly into your inbox. Messages that it considers are certainly bad go into the "blocked" folder. Messages that it cannot be sure about are put into the "Review" folder.

Now, here is where it gets good: our experience has been that after a little training, InBoxer's rate of false positives will become so very low that most people really could trust that the messages in the "Blocked" folder truly are just spam.
Taking the leap of faith and not bothering to check the "Blocked" folder really is the Nirvana of spam filtering but we feel that it is actually possible with this product.
In our tests, 11% of the messages ended up in the "Review" folder (we do not use the friends list facility when we test these types of filters). This means that once you learn to trust this filter, you will never even see 89% of spam messages. Now, doesn't that sound tempting? It sure does to us!
If, unlike our testers, you use the whitelist/blacklist features by importing your friends list, then you should achieve far fewer messages in the "Review" folder.

Like all Bayesian filters, after processing a message, it is awarded a "spamicity" rating which is expressed as a percentage. Most filters we have seen simply set a threshold above which a message is classed as spam. InBoxer merely takes it one stage further and provides two (user adjustable) thresholds:

Now, it could be said that what is happening here is that the onus is being shifted back to the user to check the "questionable" messages. This could be a valid point, but we feel that with training of the Bayesian engine (the natural process of reviewing and marking the messages), and with some adjustment of the settings, this really could be reduced to an absolute minimum. Furthermore, we feel that being freed from checking the "Blocked" folder "Just in case", is well worth it.

If you wish to see exactly how a particular message has been awarded the "spamicity" rating that it has, you can have InBoxer display a message analysis. This will show the "tokens" that have been analysed, the probability of it being spam and even the number of good and bad messages in the database that contain that token:

Should you with to re-train the Bayesian engine at any time, you can do so from the control panel, where you can also set preferences such as the folders that InBoxer watches and the folders that it "learns" from:

Filtered messages can be automatically deleted after a certain period of time, and can be automatically marked as read if you so wish:

Your friends can be automatically imported from your Outlook contacts list and can be manually added from individual messages.

Particular to InBoxer is the plug-in architecture. This (presumably) will allow later additions to the program and the program comes with three already added (two are only available as optional "Premium" features for more money):

In the order displayed, the currently available plugins are:

• MailCall simply plays a sound when a good message arrives in your inbox
• MailTones lets you select particular sounds to play when a message arrives according to rules you set about the content of the subject, body etc...
• Redirector (Premium plug-in) will watch incoming mail in Outlook and copy good mail into another folder which can be synchronized to a BlackBerry(R) device.
• PopBoxer (Premium plug-in) will forward all good messages (after filtering) to a separate Pop3 account. It Can also be used to delete spam messages from a POP account after being recognised by InBoxer. This could be a a useful feature if you retrieve your messages from a hand-held device.
• SendStats is not so much a feature for the user, as it is for the developers. It simply sends statistics about the number of messages InBoxer has handled to the InBoxer server to assist them in improving the product.

Accuracy

Due to the unique way in which InBoxer classifies messages, it leaves us with a dilemma as to how to quantify its effectiveness. Do we class "Review" messages as false positives/false negatives or not? In the end, we decided to hedge our bets and present both sets of figures

(Not classing messages needing review as misidentified)

(Classing messages needing review as misidentified)

Either way you look at it, InBoxer presented some very impressive numbers. We would have liked some more time to work with this particular filter as we feel that with careful adjustment of the thresholds and addition of the "trusted senders" facility, we could have achieved even better figures than these extremely impressive ones returned!

Conclusion

InBoxer is one of the most formidable spam filters we have seen. We particularly like the "Review" feature and that, combined with its stellar false positive rate, truly gives us the confidence that one need never look in the "Blocked Mail" folder again. This is no mean achievement!

We do not give our highest rating of five stars lightly. At the time of writing, this is only our second five star award. To earn this, the filter has to display exceptional filtering abilities combined with ease of use, innovative features and value for money. InBoxer satisfies these criteria and gives us new hope in the battle against spam - well done.

Return to Review Index