The research lab is becoming increasingly important to companies that provide spam filtering tools. Two such firms have been plugging the increased importance of such facilities.
Proofpoint Inc, one of the highest-profile venture capital-funded anti-spam startups in Silicon Valley, is trying to differentiate itself in the fast-moving anti-spam market with what it says are uniquely sophisticated statistical models, branded MLX machine learning technology, that its software uses to root out spam.
Its Technical Advisory Council offers guidance on specific areas related to MLX, the name given to the set of equations and algorithms Proofpoint’s Protection Server uses to score emails’ spamminess. MLX uses variants of statistics called logistical regression and support vector machines.
Essentially, these methods allow a more accurate scoring system and let the software more easily distinguish spam from legitimate email in gray area instances where a message’s spamminess is not certain.
Proofpoint has also recognized that there is a need to add other functions, such as virus protection, content filtering and regulatory compliance, as modules for its platform.
The firm has licensed anti-virus engines from Network Associates [NET] and Sophos for the virus part, and intends to start releasing regulation-specific content compliance modules, the first of which will deal with the HIPAA US healthcare privacy law.
Tumbleweed Communications Corp [TMWD] is currently talking up the latest version of its MMS email security appliance, which for the first time includes a feedback mechanism.
The firm said MMS 5.6 automates the process of forwarding false positives to its Message Protections Lab, in order that its engineers can work on refinishing its algorithms to avoid the problem in future.
The latest appliance also adds to the number of reports available to email administrators, and has features for recognizing so-called phishing attacks, where spammers try to con bank account details out of recipients.
This article was based on material originally published by ComputerWire.