New Image-Based Spam: No Two Alike
Spammers now modify each image sent out to sneak through filters undetected, causing a higher percentage of messages to reach the inbox.
Spammers now modify each image sent out to sneak through filters undetected, causing a higher percentage of messages to reach the inbox.
Image-based spam has increased twelve-fold in the past year, and a higher percentage is making it past spam filters. According to research released by security gateway provider IronPort, images are varied each time a message is sent out.
Spam images are changed each time a message is sent. The difference may be a change in the border, or the variance of one pixel, but the change is enough to get past traditional content and signature-scanning filters. These spam messages are compared to snowflakes because each one is different.
“Every image is in fact unique, but from a distance it will look identical,” said Craig Sprosts, senior product manager at IronPort. “Think of it as a snowflake having different patterns. The image will have random dots inserted.”
Over the past year (June 2005 to June 2006), image-based spam increased from 1 percent to 12 percent of spam volume. Image-based spam accounts for over five billion messages per day. About 78 percent of this pervasive spam passes through first- and second-generation spam filters. Sprosts estimates about 30 percent of spam delivered to an individual’s inbox can be this type image-based messaging.
Image-Based Spam Activity | |||
---|---|---|---|
Parameter | June 2005 | June 2006 | Change |
Image spam prevelance | 1% | 12% | 11 percentage points |
Duration of spam URL | 48 hours | 4 hours | 12 times faster |
Volume of spam | 30 billion | 55 billion | 83% higher |
Source: IronPort, June 2006 |
“This technique is primarily built to get through spam filters,” said Sprosts. “The images are sent in the file and not pulled down, because these images are imbedded as a file in the message itself, there is more of a chance of seeing the offer, it could increase response rates.”
The imbedded image also increases the size of each message. Sprosts said image-based spam can be about eight times larger than a regular spam message. A typical message is around 8k in size, compared to 70k for messages sent with this new tactic. “It will tax the infrastructure [of a business” even more,” he said.
The program that creates these unique, image-based messages is distributed to and sent from botnet (define) or zombie (define) machines. The company estimates that over 80 percent of spam is sent from these hijacked machines. Spammers rotate through zombie networks every few hours and change IP addresses to evade blacklists.
A similar rotation takes place with the URLs used in spam. The average lifecycle of a domain used in a spam message was 48 hours in June of 2005. In that time the domain would be detected and added to blacklists. The average duration of a URL is now four hours or less. The quick changeover of Web addresses evades blacklists and also exploits domain registration. Domains are used and allowed to expire before registration is paid for. “In April there were over 35 million domains registered, 32 million of which were never paid for and expired after five days,” said the report. The practice brings the cost of registering a domain to zero and removes any barriers associated with the cost of switching.
“It makes it more difficult to detect spammers because of the speed at which they are changing,” said Sprosts.
The study was conducted using SenderBase data, which represents 25 percent of the world’s email traffic and data from more than 100,000 ISPs, universities and corporations worldwide.