Spam now accounts for over 65% of all e-mail volume, and worm/malware attacks are costing businesses billions in clean-up efforts. Sentinel Messaging™ helps organizations eliminate unwanted mail with unprecedented 99.9% accuracy and benefit from:

• Marked decreases in lost employee revenue.
• Increased protection from hostile workplace/sexual harassment consequences.
• Drastically improved return on messaging investment.
• Extended life for existing network and storage investments.
• The end of rip-and-replace security upgrades.

When developing Sentinel Messaging™, DigiTar went back to the drawing board to deliver an innovative messaging engine that is not restrained by outside components or outdated preconceptions about the nature of spam. The engine that resulted was iSTAT™: a fully modular, 21^st century messaging platform, based on a DNA approach.

iSTAT™ powers the 99.9% difference. Simply stated, the 99.9% difference means a 2,000% increase in accuracy over 98% accurate technologies, and a staggering 5,000% increase over 95% accurate technologies. Put differently, 95% accurate systems will mis-classify 50 messages for every 1 message mis-classification by iSTAT™. A 50x accuracy differential is the difference between spam overload and a usable inbox. Perhaps more important, it's the difference between receiving that critical legitimate message or losing “the big deal”.

What does it mean to be DNA-based in a world of heuristic, Bayesian and fingerprinting technologies? As spam volumes continue to skyrocket, 90%, 95% and 98% accuracies are no longer good enough. Traditional anti-spam technologies simply hit ceilings in accuracy they cannot break through, and their users suffer. DNA-based identification (eStrand™ technology) shatters the accuracy ceiling, providing technology that can keep pace with the nearly limitless volume and variations of spam. eStrands™ make deep contextual analysis possible in a way that eludes other technologies. Further, iSTAT™ understands that spam is different things to different people, and allows each user to tailor its performance to their needs. All this with nearly zero administrator involvement or maintenance.

Competing Technologies

Heuristic or rule-based systems, rely on human experts to deduce recurring patterns in spam messages, limiting them to 85-94% accuracy. When performing this duty, humans tend to isolate the wrong or incomplete patterns that identify spam. This is due to the fact that humans are pre-disposed to focusing on the emotional weight of words instead of their contextual meaning. Consider the word tokens "penis" and "#FF0000" (the formatting sequence for the color red). Humans will create patterns that isolate "penis" and ignore "#FF0000", when the reality is that "#FF0000" occurs in spam nearly 5x more often than "penis", and "penis" has more representation in legitimate mail. Since a human expert is involved examining spam and non-spam messages for patterns, only a relatively small number of patterns are ever identified, and rarely with even basic context. The root problem with heuristics is human bias.

Bayesian systems are subject to accuracy plateaus and literary attacks that limit them to a 95-98% accuracy range. Due to limitations in their designs, Bayesian systems are restricted to establishing a shallow level of context. While they are not subject to the same bias influence as heuristic systems, they too are subject to the changing tactics of spammers. This is particularly the case with literary attacks, which they have proven particularly susceptible to. Compounding this problem, is the weeks and months of training they require to reach the 95% accuracy level of high-end heuristic systems. Their primary advantage over heuristic systems is their higher accuracy once trained, and the capability of some to have different "training" for individual users. Once fully trained, their shallow context limitations cause their accuracy to plateau at 98%. Bayesian systems simply are not sophisticated enough to keep pace with the waves of mutating spam.

Fingerprinting technologies are progeny of Bayesian systems and offer a similar accuracy range of 95-97%. They use a database of spam signatures to determine the spam score of messages. These signatures are in fact typically generated using Bayesian engines. The primary difference between Bayesian engines and fingerprinting systems is that the fingerprinting vendor does the work of generating signatures for you. Fingerprinting also inherits Bayesian's accuracy limitations and adds a few of its own: primarily an inability to tailor results to individual users, coupled with a lag time between signature development and deployment by customers. This lag time can mean reams of new spam types in your mailboxes. Fingerprinting technologies trade accuracy for ease of use. A devil's bargain in today's turbulent spam environment.

A Word on Phishing

Phishing integrates the worlds of malware and spam to create a malicious message that is more than the sum of its parts. The iSTAT™ messaging engine has been enhanced to leverage its existing strengths and eStrand™ technology coupled with phishing-specific methodologies. True phishing protection requires more than just a spam engine, it requires a spam engine that speaks the phishing dialect. Our unique position as a purveyor not only of messaging security but also complete managed security, gives DigiTar an advantage in attacking all aspects of phishing. A multi-vectored threat requires a multi-vectored approach.

Sentinel Messaging ™ Features & Benefits