Kaspersky ups the ante in anti-spam battle

The technology can detect text embedded in images without the need for machine recognition, which requires uniformity of size, style and arrangement of symbols in order to detect junk email.

Spammers can avoid detection by distorting the images, explained Kaspersky Lab's senior technology consultant, David Emm.

The new technology, which has just been successfully patented in Russia, is based on probability and statistics.

Images are filed as spam after an analysis of the layout of the graphic patterns of words and lines, or the content of the letters and words in those patterns, the company said.

The technology is also able to counter the obfuscation techniques commonly used by spammers, such as warping and rotating, and can detect text in virtually every language, according to the firm.

"When the spammers distort the images and put a lot of noise in there, they make it more difficult for machine recognition to detect," said Emm.

"This [new technique] gives us a performance benefit too, which is important when dealing with large volumes; it’s about detecting the stuff and detecting it in a way that doesn't slam the machine doing the scanning."

In related news, McAfee's February 2009 Spam Report (PDF), released this week, highlighted the increasingly malicious nature of unwanted emails.

The report details a new attack in which an email appears to confirm the purchase of an e-ticket.

The attached .zip file, which anxious users may click on to check whether their credit card has been used fraudulently, contains the ZBot Trojan.