How Spam Filters Work
When it comes to spam filters an email user has many choices. Each type works in a different way, but all share the same goal of keeping your inbox spamless, or relatively so. Each has advantages and disadvantages which the user should weigh based on the type of email address he or she is trying to protect e.g. work versus personal email addresses. The following list can serve as a basic guide to the types of spam filters out there as well as how they work.
User Defined Filter
This is a common type of filter that usually comes with a free email account. The filter sensitivity can usually be easily controlled by adjusting its "vigilance" to low, medium, or high. Some also allow you to flag specific keywords or addresses for automatic routing to your trash for all future emails bearing the same information. The major drawback to such filters is that they are often ham-handed if set too high and much too generous as bouncers if they are set too low. A sort of Goldilocks conundrum can quickly develop where no setting is "just right."
Language Filter
Automatically trashes incoming mail that does not match your preferred language. This type of filter is very efficient in what it targets and has a relatively easy target to look for. However, there are some serious drawbacks to language filters. As most of the spam produced is in English[1] and much of the world uses English as a lingua franca these filters might not be the best choice for the average person.
Header Filter
Many spammers forge their spam's header information, which can make tracing back the source of their junk very difficult. A header filter detects such a forged header and rejects its associated email. If only all spammers forged their headers this type of filter would be the Holy Grail of spam elimination.
Password Filter
Using this sort of filter a password is chosen that you can then give to people from whom you want to receive email. Any and all inbound emails that lack this password in their subject lines will be bounced back. This type of filter might be ideal for people who do not use the Internet that much or for parents trying to protect their children from spam that is often laden with pornography. On the other hand, such things as news letters and old friends trying to reach you will be a bit more difficult with this type of filter active.
Permission Filter
Much like password filters permission filters block all email not coming from an address that the user has approved. Adding new contacts might prove to be a very tedious process if you haven't vetted them first. Not to mention if your best friend decides to get a brand new email address without telling you first.
Community Filter
Filters out spam based on a group of users' criteria as to what qualifies as junk mail. This breed of filter can develop a skewed definition as to just what a piece of spam looks like. In addition, the use of a community filter may leave some members feeling overly restricted from or too exposed to what others classify as spam. Still, a point in this type's favor is that the definition as to what constitutes spam is constantly dynamic based on who's in the "community."
Content Filter
All incoming emails are scanned before the user ever sees them for certain telltale signs of spaminess. An email is apt to become a pariah quickly if it contains such common spammy traits as strange punctuation and miserable spelling. While this approach might be great for getting rid of a large amount of junk mail it can jettison emails from legitimate sources such as friends and family using its fixed ability to make a snap judgment. For example, emails sent by younger family members and those with a disability such as dyslexia may appear suspicious to a content filter. And as email has never been a format that encourages the fineries of grammar such filters may become a nuisance depending on who the nature of your senders.
Challenge-Response Filter
Automatically sends a return email to a new sender asking them to verify their identity. This usually takes the form of clicking on a link that will ask the sender to fill out some basic information such as the reason for the correspondence. Frequently there is also an authentication code that must be copied from a series of random numbers and letters appearing in a color obscured box. The purpose of these steps is to ensure that the sender is not a spammer relying on a zombie computer that is automatically spewing out loads and loads of junk emails. Using a challenge-response filter ensures that only actual living, breathing human beings who have taken the time to follow a few simple instructions will be able to reach you. One danger of this system, though, is that the sender's own filter may classify your challenge email as spam! This scenario would be like two people shouting at each other through a soundproof door.
Adaptive Inboxer/ Bayesian Based Filter
This type of filter "learns" to recognize spam as spam by drawing on its memory of what the user has classified as spam in the past. The logic behind an adaptive inboxer is based on the theories of Thomas Bayes, an eighteenth century English mathematician concerned with developing theories to demonstrate probabilities. The way an adaptive inboxer works is that the user must train it by giving it emails he or she knows are spam for it to scan. A Bayesian filter then assigns probabilities to certain words appearing in these emails as to how likely they are to show up in legitimate emails. With training and time this sort of spam filter can become very accurate. The main caveat to using this type of filtering software is that, unlike the choices above, it requires an investment of time and interaction by a human being, namely you.
References
http://www.news.com/Study-Five-countries-send-99-percent-of-spam/2100-1024_3-5253411.html