Spam Filtering

Please Note

Spam Filtering Is Only Enabled On Your Account By Request

It was found that our previous spam filtering system was deteriorating in quality because it was endeavouring to judge what is spam, or not, for all users. Whereas what is spam for one person may not be considered spam by someone else, so with many users training the system with conflicting results, it led to a less than optimal situation for everyone.

The new system is targeted towards individual users rather than all users so it can be specifically trained and tailored to recognise spam according to each user's own judgement. The downside of this approach is that each user has to be pro-active and mindful of training their own filter which takes some initial effort but, with due care, the end results can be an almost totally spam free mailbox after 2500 messages have been fed into the system.

If you would like spam filtering enabled please send an email to

Once you have requested spam filtering on your account, you will notice a new set of folders appears in RoundCube (see image below).

If you are currently in the habit of downloading all your email before checking RoundCube first, to take advantage of the spam filtering, you will need to log into RoundCube BEFORE opening your email program & downloading the email. Once you log out, open your email program & download the email in the usual way.

Alternatively, you can configure your email program to leave a copy of messages on the server for a certain period, say 5 days, which allows you to log into RoundCube later and train your spam filter.

To train the filter you simply need to drag & drop any emails that you consider to be spam into the Spam folder as shown in the image below. (IMAP users will also be able to subscribe to these folders and use them in the same way).

Roundcube Spam Folders

How Spam Filtering works

Fighting spam is probably the most complex task of a postmaster. Our new spam filtering is based on a statistical engine that analyses a text and produces a probability based on the content.

It employs an adaptive filter which means it is capable of learning and adapting to each user's email. Instead of working off of a list of "rules" to identify spam, the probabilistic engine examines the content of each message and learns what type of content the user deems as spam (or non-spam). This approach to machine-learning provides much higher levels of accuracy than other "hodge-podge" solutions.

Additional Training

As well as dragging & dropping your spam into the Junk.Spam folder, you should regularly check the Junk folder to ensure that no genuine email has been marked as spam. If you happen to find a genuine email in Junk you should drag it to the Junk.Not Spam folder.