Posted by Virus Bulletin on Aug 9, 2007
Study investigates sites and servers linked to from spam.
A detailed investigation into the distribution and usage of webservers to host the various 'scams' pushed by spam campaigns has found that the vast majority of such scams use only a single server per scam, with 57.4% of these servers based in the US.
The study, carried out by researchers at the University of California, San Diego, gathered over 1 million individual spams over a week-long period, and tracked any URLs found in the mails back to the sites they intended to lead victims to. Using a technique dubbed 'image shingling', involving comparing screenshots of rendered web pages to determine which spams and sites belonged to an individual scam, the team were able to produce detailed information on the infrastructure used by the scams, a generic term for any site intended to make money from spamming, be it selling goods or services or phishing for banking details.
The analysis found that spam sources were fairly evenly distributed around the world, with the US-based systems relaying 14%, 28% coming from Western Europe and 16% from Asia. Hosting of the scam sites was much more heavily concentrated in the US however, with over 57% of all scams identified hosted on servers there, with 14% in Europe and 16% in Asia.
The results also showed that while spam campaigns were generally short lived, with more than half lasting less than 12 hours and less than 1% lasting over three days, 50% of the scam sites they led to remained active for a week or more. While many spams are sent from large distributed botnets, there is little evidence of distribution in the hosting of scam sites - 94% of the 2,334 separate scams identified only used one IP address hosting the endpoint of the scam, and 84% used a single domain name. While the use of aliasing and virtual domains is common to avoid DNS-based blacklisting, the researchers suggest that the prevalence of single-IP based scams could provide a single point of failure in the spammers' money-making systems, providing a possible vector to block the source of funds and thus reduce the spam problem.
A small number of scams, around 6%, did make use of distributed hosting however, although only a handful used more than ten and a single scam was seen to be using 45 separate IP addresses, suggesting that more sophisticated scammers have already addressed the vulnerability of their business to IP blacklisting.
The study is being presented today at the USENIX Security 2007 in Boston. A summary of the findings can be found at the UCSD Jacobs School of Engineering, here.
Posted on 09 August 2007 by Virus Bulletin