Previously, four C&C servers controlled all Mahdi infections. "The interesting part is that one server is used mostly with Israeli targets, while the other three are for Iranian and Arab targets," said Raff. "The one used for Israel also targets other Middle Eastern countries, but there are no Israeli targets on the other three." All four C&C servers were also hosted by the same provider in Canada, although a whois lookup on the IP addresses claims that they're really based in Azerbaijan, and in one case on the premises of that country's Royal Bank. But according to Seculert, "we confirmed with several ISPs that the physical addresses of the C&C servers are indeed located in the headquarters of the Canadian hosting provider."
According to Seculert, about half of the 800 known systems infected by Mahdi--all via targeted attacks--have been in Iran, while roughly 7% of infections were in Israel. "Looking deeper into the Mahdi victims' IP addresses, we did find a few dozen IP addresses which seem to be from non-Middle-Eastern countries, such as the U.S and U.K.," according to Seculert, although it appeared that the infected machines were owned by people who were only visiting those countries. But those Seculert findings differed from research subsequently published by Symantec, which claimed that 72% of all Mahdi infections involved PCs in Israel.
What accounts for that discrepancy? "Symantec may have come up with 72% because they were only looking at variants which communicated with the C&C servers targeting entities from Israel," according to Seculert's latest analysis. "Or, maybe they are looking only at their customer's machines which they found to be infected with Mahdi. As an American company, Symantec is not allowed to sell their products to Iran, and therefore they can't see infections in Iran." But Raff noted that Seculert's analysis had come from identifying PCs that had connected to the Mahdi botnet itself.
One final piece of evidence that the botnet was built by Iranians involves its naming conventions. "Each bot node [infected PC] receives a unique identifier," Raff said, which is a text string combining reused prefixes with unique text strings, so that any individual machine can be quickly identified and controlled. Some of the words used to construct those prefixes include these names: Chabehar, Iranshahr, Khash, Nikshahr, Saravan, and Zabol. All of those names refer to cities or counties in Iran. Another prefix also used by developers, meanwhile, was "Flame."
Is that, finally, evidence of a connection between Mahdi and Flame? The answer is likely no. According to Seculert, "the first targeted victim with the 'Flame' prefix began communicating with the C&C server in early June, right after the Kaspersky Lab discovery of Flame went public." In other words, the inclusion of the word "Flame" in Mahdi appeared to have been made after Flame became public knowledge.