New features of botnets created by the infamous Storm Worm allow denial of service attacks to be launched against security defenders that attempt to interrupt its operation.

Attempts to probe command-and-control servers can result in a withering counter-attack of malicious traffic that can swamp the internet connections of security activists for days, according to Josh Korman, host-protection architect the ISS security division of IBM.

"As you try to investigate [Storm], it knows, and it punishes," Korman told delegates at the Interop New York conference this week, Network World reports.

It's unclear whether the counter-attacks are launched automatically by the malign system or by botnet herders manually. What is clear is that the code behind the malware is evolving.

Instead of simply disabling anti-virus applications, the latest refinement to the worm means that such applications may appear to run but are unable to detect malware. "It's running, but it's not doing anything," Korman explained. "You can brain-dead anything."

The Storm Worm malware strain first surfaced in January, in emails attempting to trick users into visiting maliciously-constructed websites under the guise of messages offering recipients information about the storms ravaging Europe at the time.

Over recent months crackers have refined their tactics. Emails punting the malware now contain fake links to YouTube, for example. Hackers have also attempted to trick users into visiting maliciously-constructed websites via login confirmation spam or bogus electronic greeting card receipts. The attack methodology is much the same in each case.

The malicious sites are designed to load malware onto the PCs of Windows users, typically using well-known security vulnerabilities that a user has failed to patch. Compromised machines become clients in zombie networks under the control of hackers.

Estimates of the number of machines infected by the Storm Worm (which is actually more accurately described as a Trojan, although routinely described as a worm by security researchers) estimate from one to five million or more.

Last month it emerged that hackers had effectively segmented the Storm botnet into smaller networks. Individual compromised clients connect to other infected machines using Overnet, a peer-to-peer protocol.

Bot herders have begun using a 40-byte key to encrypt traffic sent through Overnet, since each node must know the password to unencrypt the Overnet traffic, providing a mechanism for hackers to segment the network into smaller components. ®