The research, led by a team at the École Polytechnique de Montréal, involved collaborators at Carlton University, Canada. It also enlisted the help of researchers at Nancy University, France, along with anti-malware company ESET.
Scientists used a $1 million, 98-machine server cluster as a platform to create 3,000 virtual machines, each of them simulated with a different IP and email address. They then infected the machines with Waladec to measure statistics including how quickly it spread.
The project, described in MIT's Technology Review, was carried out earlier last year and discussed in December in a paper entitled "The case for in-the-lab botnet experimentation: Creating and taking down a 3,000-node botnet.
One significant finding from the experiment was that the Waledac botnet's weak cryptographic protection in the wild was a necessity. The botnet's command-and-control infrastructure used the same Advanced Encryption Standard (AES) session key for all bots for 10 months.
"We initially thought that this was a design error made by the bad guys, but when implementing the Waledac C&C server it turns out that it was impossible to generate a session key for each bot, because it overloads the server with cryptographic computation," said the paper.This approach is more ethical than attempting to disrupt botnets in the wild, the researchers said, adding that in some jursidictions it is considered illegal to create elements that join a botnet. It is also possible for sophisticated botnets to launch denial-of-service attacks on domains that they recognise as sources of botnet disruption.