During a few months of hard-working testing the applications we found (and fix) several bugs:

• eml_parser_plugins: in order to support as many emails as posible is necessary to have a parser as flexible as possible when analyzing the RFC2822 emails structure.
• Fixed memory leaak on Core in email report.

Wirebrush4SPAM now includes some prescheduler algorithms in order to improve the filtering efficiency. Prescheduler technique implements diferent planification types in order to choose the most appropriate according to the needs of each user.

Future versions of wb4spam will include LAF scheme (Learning After Report). LAF modifies the behavior of current spam filtering middleware (such as SpamAssassin). Using this scheme, learning jobs are made after sending a response to the network client reducing the filtering lag. The reduction of filtering lag is useful to address the classification of the message during the SMTP transaction.

We are now implementing this feature and it will be included in future releases of Wirebrush4SPAM.

We had released Wirebrush4SPAM developers guide. Please, refer this documentation to develop your own WB4SPAM plugins.

The spamc/spamd server protocol development has started today while Alvaro Pérez fixes some bugs on META rule evaluation. The protocol will ensure the compatibility with spamc tool and all software able to query SpamAssasssin (for instance sa-exim). This feature will be developed by David Ruano due to the workload of Fabio.

The AWL plugin of Wirebrush4SPAM has been commited last week. We are now updating the documentation in order to describe the sintaxis of functions added by this plugin.

We are pleased to announce a new plugin based on SpamHunting filtering technique.

In particular the implemented technique is based on the use of an IBR  (Instance-Based Reasoning) that incorporates an structure for indexing messages, a voting strategy, a mechanism for calculating the quality of the generated solutions and a technique for the identification and elimination of irrelevant knowledge, in order to obtain a high level of accuracy and allow quick adaptation to changes produced in the environment.

For more information:

[1] SpamHunting: Adaptative and Intelligent Spam Filtering Technique

[2] SpamHunting: An instance-based reasoning system for spam labelling and filtering 

RXL plugin has been improved by including a cache for each realtime list. The effects of cache will be achieved when we complete the server protocol. Fabio!! we need you, go back from your erasmus activities. 😉

Concept drift is the main problem on spam filtering. Filters are not able to handle information about nuclear problems in Japan and consequently, are not able to accurately filter newest e-mails anouncing potasium iode pills for sale.

Concept drift references a lot of issues that can not be modeled (sometimes are unknown) and can be relevant for solving the target problem (spam filtering). A new spam trend based on advertising a new illegal product is a fact that can not be handle by filters and makes the filtering more difficult. What about new techniques and/or ideas used by spammers to avoid spam filters. Each of then involves changes on target problem and requires a filter update.

We believe that spam filtering can not be solved by using a software product. A spam filtering service is needed to manually handle the effects of concept drift.

References:

[1] El Mundo: El ‘spam’ ya vende yoduro de potasio

[2] SpamWarnings: If Japan nucl? Radiation reaches the surface? United States, you need potassium iodide pills?