How Do Email Spam Filters Work

If you are the one who works with emails on a daily basis, you are most likely to be using a SPAM FILTER to ease the job of sifting through a large number of spam emails every day.

Needless to say that spam filters do make our job a lot simpler by automatically filtering out the spam

without which, it is almost impossible to manually filter the junk emails that arrive in millions each day. However, it is often necessary to have a basic knowledge of how spam filters actually work and on what basis they flag an email as spam.

How do Spam Filters Work?

There are different kinds of spam filters:

Header Spam Filters:

Header spam filters work by examining the header information of a particular email message to check if it appears to have been forged. The header of every email contains information which tells the origin of the email. ie: The incoming email ID and usually the IP address (server address) of the sender. So, spammers often forge the header to input a false sender ID and IP address so as to make it difficult to trace them.

Thus, if an email is supposed to have a forged header or if the same message is found to have been sent to multiple recipients, it is most likely considered as a spam by many filters. This method of spam filtering is often quite effective, however, occasionally it may result in some of the requested newsletters from being misdirected into the spam folders.

Content Spam Filters:

Content spam filter is one of the most effective and widely used filter to combat spam emails. They use a sophisticated algorithm with a set of pre-defined rules to determine whether a given email is a spam. They work by scanning the entire text/body of the email to search for specific words and patterns that make them resemble a typical spam message. Most content spam filters work based on the following criteria and check to see:

1. If the message speaks a lot about money matter. Commonly suspected words include: lottery, discount, offer, bank account, money back guarantee etc.
2. If the message contains adult terms like: viagra, pills, bed, drugs, hot and so on.
3. If there is any sort of urgency. Most spam emails call for an urgency by using terms such as hurry, offer valid till etc.
4. If the message contains a single large image with little or no text, it is often considered as spam by many filters.

Each content spam filter may have its own set of additional rules using which it evaluates each incoming email. In most cases, content and header spam filters are combined together to achieve higher level of accuracy.

Language Spam Filters:

Language spam filter is designed to simply filter out any email that is not in the user’s native language. Since spammers come from all parts of the world with different languages, a language spam filter can help get rid of those annoying emails that come in the languages that you cannot read!

User Defined Spam Filters:

User defined spam filters can be very handy, however they need a considerable amount of time investment in configuring and setting up a set the rules using which the filter works.

For example, the user can configure to have all the emails from friends and company to reach the inbox, newsletters to reach a secondary inbox and all those remaining to the spam folder. Here the user must carefully examine the patterns of spam emails that he receives from time to time and needs to set up the rules accordingly. This filter when improperly configured can sometime lead to false positives or false negatives.

Other Types of Spam Filters:

Popular webmail services like Gmail, Yahoo and Hotmail combine both header and content spam filtering techniques. In addition to this they also use their own algorithms to combat spam.

For example, services like Gmail uses “optical text recognition” to identify spammy text inside an image. Also, the users are provided with an option to “Report Spam” whenever a spam email accidentally reaches the inbox. With the user feedback, the filter learns and becomes more powerful in carrying out the filtering process.