According to graduate student Ola Amayri, the concept behind the statistical framework developed by her and her thesis advisor Nizar Bouguila, to fight spam is a relatively simple one.
“The majority of previous research has focused on the textual content of spam emails, ignoring visual content found in multimedia content, such as images,” Amayri explained in a statement. “By considering patterns from text and images simultaneously, we’ve been able to propose a new method for filtering out spam.”
Over the years, spammers have become increasingly clever at disguising their junk payloads, according to the PhD candidate. They obscure text in their messages to foil spam filters looking for key words that tip a spammer’s hand, like Viagra. They also vary the look of an image in a message by tweaking its background.
What compounds the problem for administrators is that spam fighters have taken a bifurcated approach to the tricksters. Filters target text tricks or image tricks, but not both, according to Amayri.
The researchers developed their new approach to spam identification by studying existing text and image filters for patterns. Then, using pattern recognition and data mining techniques, they forged a strong method for spearing spam.
Although the researchers’ initial efforts have been directed at English-language spam, they’re confident the techniques can be applied to other tongues without much difficulty.
Now in the stage of polishing up their work, the researchers hope to use their junk mail fighting technology in a plug-in for SpamAssassin, the most widely used open source spam filter in the world. Not only will that move give the technology a wide distribution, but it will allow other researchers to review the Concordia couple’s work, with an eye toward improving it.
If technology developed by the Canadian researchers is as effective as they claim it will be, it can be a potent weapon in the arsenal of spam fighters. Such weapons are important because although spam volumes are decreasing, they’re getting increasingly malignant.
According to a recent report from Kaspersky Lab, spam volumes declined in the quarter ending September 30 by 2.8 points to 71.5 percent of all Internet mail. However, spam containing malicious content rose by almost a point, to 3.9 percent from three percent.
The report acknowledged that some of the reduction of spam on the Net could be attributed to seasonal factors. For example, the quarter includes the summer months when business activity slows down, especially in Europe. The decline in eyeballs checking email removes the incentive for spammers to ply their junk. In addition, even hard working spammers have to take a summer vacation.
Nevertheless, Kaspersky’s report noted, much of the spam pushed by advertisers — those good old fashion pitches for cheap proscription drugs and Rolex watches — have moved to other Internet vehicles, such as banner ads, social media, coupon services and contextual advertising.
As advertising spam, which is relatively benign, has slipped, criminal spam — junk mail containing malicious attachments, offers for illegal goods or malicious links — has grown. That means it’s more important than ever to prevent spam from reaching inboxes, which is exactly what the Canadian researches are promising.