Kaspersky Lab patents a new method for combating image spam

30 Aug 2012
Product news

Kaspersky Lab’s experts have developed and patented a new, effective technology that detects spam by scanning images. It incorporates a signature-based method that is already used to detect unwanted correspondence, and adapts it considerably in order to work effectively with images.
Spam is one of the most prominent threats today — in the first quarter of 2012, approximately three out of every four emails was spam. According to the data in a survey conducted by O+K Research in May 2012, 69% of users around the world have encountered spam. Kaspersky Lab devotes a lot of attention to the problem of unsolicited emails, including image spam, and is constantly improving its anti-spam technologies.

Currently, there is no shortage of relatively effective tools for detecting text spam. But image-based spam presents a much more complicated problem. First, the text in the image attached to the email needs to be properly recognized. Only then is it possible to determine whether or not the email constitutes unwanted correspondence. Furthermore, the image might contain objects other than text, such as photographs of goods that are being advertised, and that makes detection more difficult.

Kaspersky Lab has registered in Russia Patent No. 2453919 which sets out a new method that can be used to detect image spam. First, the image is broken down into separate objects, and each object is assigned within a defined contour. A signature is then created for each object, based on the different elements within the shape (segments with straight lines, or arcs) and how these objects fit within the boundaries of the shapes. The resulting object signature is compared against the signatures of other objects that have been detected in image spam. A similar analysis is performed for all graphical objects detected in suspicious emails, and then a final verdict is issued as to whether or not the email should be categorized as spam. Furthermore, this technology takes into account not only exact signature matches, but also similarities among different objects, such as an offset object or differences in size.

Thanks to its talented inventors, Kaspersky Lab’s Russian patent portfolio is growing steadily. Currently, the company holds 60 patents in Russia. Kaspersky Labs complete patent portfolio includes over 110 patents in the US, Russia, China, and Europe.

© 2016 AO Kaspersky Lab.

All Rights Reserved.