Keyword filters are the simplest of spam filters.
A keyword filter simply looks for specific words--called keywords--in an incoming
message. If the keyword is found the incoming message is assumed to be spam
and is blocked. For example, the words "LIMITED TIME OFFER" may be a keyword.
When an email arrives the system looks through the email for these words. If the
words are found the system assumes the message is spam and blocks it.
Of course, keyword filters work by having someone look at spam and recognizing
keywords that often appear in spam but don't appear in good email. This is
time-consuming and the spam filter is always one step behind the spammer. If
the spammer comes up with a new way to spell Viagra (i.e., "V!agra") then that
must be added to the keyword filter--but it probably won't be added until at
least one spam gets through. Also, spammers can just mangle or misspell words
to get them past keyword filters.
Another problem with keyword filters is the high risk of false positives. While
"LIMITED TIME OFFER" seldom appears in real email, it could. If the system
automatically discards a message just because these words were found it could
easily discard good email. Additionally, what might be spammy words for one
user might not be spammy for someone else. A "one size fits all" type of
approach is necessary and isn't very effective.
One area where keyword filters are useful--and where PrismEmail implements
them--is in blacklisting known spam websites. For example, if we detect a
spam that references the website iamspamsite.com then we will add that to our
keyword filters. Since there is virtually no chance that a real message between
real users will be discussing any given spam website there is very little risk
in rejecting such messages.
Keyword filters, then, are not sufficient as the only line of defense against
spam. However, limited keyword filters--especially those that detect known
spam websites--can be useful at catching additional spam that might otherwise