|
The Spam Scenario today
Spam menace is one of the biggest issues that cripple email traffic on the Internet today. Spammers try to sell you cheap drugs, mortgages, 419 scams, a range of tacky consumer products, easy college degrees, porn, dating services and everything else imaginable! Certainly you are driven to the end of your patience when you find 95% of your inbox filled with this garbage.
The immediate impact of spam can be measured by way of huge bandwidth loss, usage of network and computer resources and the wastage of productive time, both at system and human levels. Its aftereffects include financial theft, data and intellectual property theft, Phishing, Virus and other malware infections and online fraud. Spam has huge financial overtones. According to an estimate, the world lost 60 billion dollars as a direct and immediate impact of spam in 2006, while its second and third levels of impact on the economy could be much deeper and wider. And each one of us is bearing that cost!
The challenges faced by spam filters
One of the most difficult problems in countering spam is its nature of getting changed day-to-day, cycle-to-cycle. For example, a spammer regularly changes the contents of his email and also uses Spambots to send out the bulk messages. Spambots are programs dropped into the PCs of normal computer users by Worms and Trojans. They download spam content from the web, automatically create emails and send them out to tons of mail IDs on the Internet, all without the computer owner’s knowledge.
Thanks to Spambots, most bulk emails today have diverse, individual characteristics (including its country of origin, time, basic contents, originating server and more). This makes it virtually impossible for any anti-spam solution, even with the best of Bayesian or similar techniques, to pin-point a spam.
Traditional anti-spam solutions try to figure out specific patterns in emails and the ones matching this pattern is marked as SPAM and the rest are marked as GOOD. This was effective till a few years ago, but with the rising popularity of "spambots" and "spambot making kits", most spam now manages get through these filters.
Spammers show a high degree of adaptability and resourcefulness in challenging antispam technologies. To circumvent spam filters based on checks done for specific words and phrases, spammers inserted special characters like dots and hyphens between every letter of such probable words. Then came image spam which used images with printed text on it in place of regular content. The current trend is a slew of spam mails with attached PDF files, which again manage to slip through spam filters.
Non Intrusive Learning Patterns (NILP)
NILP carries with itself a continuously updating database containing DNA imprints of millions of SPAM mails detected world-wide. Once the DNA imprints are checked, NILP reverses its learning by classifying GOOD (Ham) mails, rather than using the traditional way of classifying BAD (Spam) mails.
How NILP works at an organizational level
Have a look at the following characteristics of any organization’s mail traffic:
1. Mail/proxy Gateway
2. Mail Databases
3. Mail Users
4. Contents
5. Policies
6. Network Structure
The above entities will continue to remain as the fundamental elements of a company’s communication for a long time. Users may be added or deleted, public IPs may be added or deleted, but overall these basic elements remain the same. To give an example, an engineering organization will MOST certainly not subscribe or ask for information about medical drugs. A software organization will rather have no interest in receiving mails from religious institutions.
Using the logic above, NILP tune s itself to understand the exact nature of the mail traffic of an organization, and then classifies all mails accordingly. It is extremely fast, scaleable and can regularly unlearn things to keep itself very efficient and adaptable in a real-time environment.
NILP at the mailbox level
At the user level, NILP employs advanced concepts of Artificial Intelligence to observe the behavioral patterns of the user, and starts thinking from the user point of view. This way, it very accurately understands what’s good mail for the user. The mechanism learns on its own, while it incorporates continuous feeds from MicroWorld Server as well. This makes NILP the most effective technology in Spam and Phishing filtering.
|
