$1,500 (50 users)
: The 'training' of the Bayesian filter may be expedited by distributing the workload among all employees.
: Users cannot inspect their own quarantined spam messages to share the workload, and to improve the timeliness, of releasing any false positives.
: An excellent and comprehensive content filter that is particularly easy to install and configure.
DynaComm i:mail acts as an SMTP gateway, running on a Windows NT/2000/XP/Server 2003 machine, which does not have to be dedicated to the task – it can also host the existing mail server, for example. It is a member of a family of products that together form DynaComm's i:series Enterprise Content Security Suite, which offers other optional components including general Internet access filtering (HTTP and FTP) and controls, and a component called i:scan that monitors file access and downloads across the enterprise in a very wide-ranging way (not just across the Internet) to control confidentiality and misuse of data.
Anti-virus is offered via the ability to use one or more third-party virus scanners, of which the most popular are directly supported. However, almost any scanner can be used provided that it can be invoked by a command line or implemented as a DLL.
Anti-spam uses a combination Bayesian filtering, Real-time Blackhole Lists (RBLs) and a general awareness of known spam techniques. White lists can also be set up to override any spam indications in messages from trusted sources.
We particularly like the way in which Bayesian filtering is implemented in this product because, as well as the usual 'training' of the filter on collections of known spam and legitimate messages, you can perform distributed 'training' from user input using its Bayesian Distributed Trainer. The way this works is that the server attaches 'tags' to each email categorising it as 'spam', 'indeterminate' or 'legitimate' and the user can click on a link to indicate to the server whether he agrees with this classification or not. This approach can speed up training by effectively allowing the workload of 'training' to be shared among all employees, but can be switched off later when classification accuracy rates reach an acceptable level.
Inappropriate content is dealt with by categories such as adult, games, shopping, etc. Such categorisation is based on message content analysis, and not on known blacklisted sources. Custom categories may be created to cover the protection of proprietary or 'company confidential' data, for example, by including product or project names etc.
Action taken is rules based, so messages may be blocked or delivered but logged, for example. Policy configuration of the rules is made very flexible by the use of Boolean operators. Comprehensive reporting is also provided and schedules allow reports and database maintenance to be automated.