SOFTWIN Total Security 2010 User's Guide download pdf (Page 238)

Image Filter

Since avoiding heuristic filter detection has become quite a challenge, nowadays'

inbox folders are full with more and more messages only containing an image with

unsolicited content. To cope with this growing problem, BitDefender introduced the

Image filter that compares the image signature from the e-mail with those from

the BitDefender database. In case of a match the e-mail will be tagged as SPAM.

URL Filter

Almost all spam messages include links to various web locations. These locations

usually contain more advertising and the possibility to buy things, and, sometimes,

they are used for phishing.

BitDefender maintains a database of such links. The URL filter checks every URL

link in a message against its database. If a match is made, the message is tagged

as SPAM.

NeuNet (Heuristic) Filter

The NeuNet (Heuristic) filter performs a set of tests on all the message

components, (i.e. not only the header but also the message body in either HTML or

text format), looking for words, phrases, links or other characteristics of SPAM. Based

on the results of the analysis, it adds a SPAM score to the message.

The filter also detects messages marked as SEXUALLY-EXPLICIT: in the subject

line and tags them as SPAM.

Note

Starting May 19, 2004, spam that contains sexually oriented material must include

the warning SEXUALLY-EXPLICIT: in the subject line or face fines for violations

of federal law.

Bayesian Filter

The Bayesian filter module classifies messages according to statistical information

regarding the rate at which specific words appear in messages classified SPAM as

compared to those declared NON-SPAM (by you or by the heuristic filter).

This means, for example, if a certain four-letter word is seen to appear more often

in SPAM, it is natural to assume there is an increased probability that the next

incoming message that includes it actually IS SPAM. All relevant words within a

message are taken into account. By synthesizing the statistical information, the

overall probability for the whole message to be SPAM is computed.

This module presents another interesting characteristic: it is trainable. It adapts

quickly to the type of messages received by a certain user, and stores information

about all. To function effectively, the filter must be trained, meaning, to be presented

with samples of SPAM and legitimate messages, much like a hound is primed to

Antispam

212

BitDefender Total Security 2010

1 2 ... 233 234 235 236 237 238 239 240 241 242 243 ... 443 444

No comments

SOFTWIN Total Security 2010 User's Guide Page 238