Spam clogs up inboxes and networks globally

You're sitting at your desk, and you've received another 5 e-mails. Excitedly you open your inbox, only to find 2 messages are offering you a new loan, and 3 of them are adverts to help improve a part of your life for really cheap! Unsolicited e-mail, or SPAM as it is more commonly known, has become an increasingly large problem over the past few years. Organisations and home users, on a global scale, are assaulted by retailers from every field trying to peddle their products, as e-mail is so fast, easy and cheap to use. The impact of which costs companies hours of lost productivity, and Internet Service Providers millions of dollars in bandwidth fees.

Someone who isn't a spammer shouldn't be blocked from sending e-mail!

The difficulty in stopping spam is trying to figure out who's allowed to send e-mail and who isn't. Someone who isn't a spammer shouldn't be blocked from sending e-mail! Computers are used to analyse messages and determine if they're spam or not. The problem with doing this naively is that computers are mathematical machines and can't understand what sentences mean very easily - so instead they try to match basic patterns between spam and non-spam messages. This problem becomes even harder when some people see a message as spam, but others don't - pattern matching becomes extremely difficult at that point!


A.N.N.A. is an Antispam Neural Net Agent that combines the knowledge of multiple aspects of Artificial Intelligence when trying to analyse e-mail. She reads e-mails in several different ways, using various text analysis methods, and in the process of doing so - tries to evaluate how important, and how pertinent the result of each method is.

...she managed to identify over 94% of all spam, and got less than 0.03% false positives!

She then combines the results of the tests and judges the message. A.N.N.A.'s power lies not in protecting just one person, but protecting an entire organisation. A.N.N.A.'s proficiency was tested against thousands of messages sent and received by Enron, and she managed to identify over 94% of all spam, and got less than 0.03% false positives! For an average person, that means less than one message lost a year! A.N.N.A. has demonstrated tremendous power and there has been commercial interest in deploying her, and even fusing her skills with SpamAssassin to form the ultimate enterprise wide anti-spam guardian. To sample a taste of her skill you can even challenge her.

A project by Kaiesh Vohra