Kaspersky Lab, a leading developer of secure content management solutions, announces the successful patenting of advanced technology capable of identifying spam in raster images.
Spammers often send out messages containing graphics with the aim of avoiding detection by spam filters. In order to identify this type of message the text has first to be distinguished from the picture. In order to make detection even more problematic, spammers often add complex diversionary graphics to the background image, interfere with the geometry of the letters and break up messages using bogus frames and lines.
Optical Character Recognition (OCR) is the conventional method used to identify text in images. However, this is resource-intensive and does not offer the necessary level of accuracy. Unlike OCR, the newly patented technology guarantees rapid and precise detection of any spam contained within the images. It readily identifies any additional graphics placed in the image for the purposes of obfuscating the text and is not deterred by distorted text, which significantly increases the program's detection levels.
At the core of the new technology is an algorithm that uses statistical analysis based on probabilities to determine whether an image contains text or not. The program examines the characteristics of the image and uses the algorithm to decide if the image contains recognizable text.
The new technology, authored by the head of Kaspersky Lab's Anti-Spam Technology team Evgeny Smirnov, was issued with two patents – Nos. 7706613 and 7706614 – on 27 April, 2010 by the U.S. Patent and Trademark Office.
On 4 May, an enhanced variant of the technology was issued with patent No. 7711192. This variant incorporates optimized object isolation that makes objects more readily distinguishable, and includes improved filtering of spam.
"Mechanical methods of recognition require symbols to be of the same size and placed at regular intervals. Our new technology can work with a variety of warped or distorted letters or words, greatly enhancing the accuracy of detection. The patented method is also much faster at processing images," said Nadezhda Kashchenko, Chief Intellectual Property Counsel at Kaspersky Lab.
Kaspersky Lab currently has more than 50 patent applications pending in the U.S., Russia, China and Europe that relate to a range of unique IT security technologies developed by the Company's personnel.