Version 3.3 (May 2008):

• Implemented selective greylisting for senders found on several blocklists. This only delays emails from known spam senders. Because most spammers do not retry delivery, approximately 70-80% of all spam does not have to be processed by Arska.
• Added possibility to define own word lists, which can give positive or negative spam points based on the word list and number of matches found.

Version 3.2 (July 2006):

• Implemented sender-based white/blacklisting that can be added globally, for specific recipient domains, or by end-users themselves that only affect them.
• Added CRM114 Markovian classifier, which is a trainable filter much like the Bayes classifier, but also takes into account the contexts words appear in (i.e. it is more specific than the Bayes classifier, but requires more training to achieve it).

Versio 3.1 (August 2005):

• Possibility to set user- or domain-specific filtering levels.
• User-specific vacation autoreplies can be defined for several time periods (activates automatically for specified dates).

Version 3.0 (April 2005):

• Ability to create multiple accounts for the web-based management interface, where each account can be fully configured what sections of the web interface is accessible, as well as which domains or email addresses should be included in reports and listings.
• End-users can easily create personal logins to the web interface, giving them limited access to only their emails.
• Quarantine notifications can be configured on a per account basis to send an email with a summary of stopped messages according to specific criteria (number of messages, time period, score range).
• Retroactive virus scanning that can rescan recent messages after the anti-virus database has been updated, and send warnings about infected emails that were delivered prior to the anti-virus update.
• Clustering support with several redundant mail filtering servers using a shared database for central management and summary statistics.

More detailed descriptions of all the features in Arska can be read in the Arska White Paper (PDF)