Security experts warn that a recently disclosed Windows security vulnerability may soon be used by a powerful Internet worm that could disrupt traffic on the Internet and affect millions of machines worldwide.

The vulnerability, a buffer overrun in a Windows interface that handles the RPC (Remote Procedure Call) protocol, was acknowledged by Microsoft in a security bulletin, MS03-026, posted on July 16.

On Thursday, the US Department of Homeland Security updated an earlier warning about the RPC vulnerability, noting increased network scanning and the widespread distribution of working exploits on the Internet.

The vulnerability affects almost all versions of Windows, and could enable a remote attacker to place and run malicious code on affected machines, giving them total control over the systems, Microsoft said.

No user interaction would be required for machines to be compromised, prompting security experts to liken the RPC vulnerability to the buffer overflow vulnerability in Microsoft's Internet Information Server (IIS) that was exploited by the Code Red worm in July, 2001.

"I would compare (RPC) to Code Red. It doesn't require user interaction and the number of infectable machines is on same order of magnitude," said Johannes Ullrich, chief technology officer at The SANS Institute's Internet Storm Center.

Net exploits However, where Code Red affected a component typically found on Windows servers, the RPC vulnerability affects a component found on both Windows servers and desktops, according to Tomasz Ostwald, a co-founder of Polish research group Last Stage of Delirium, which discovered the RPC flaw and reported it to Microsoft.

That increases the number of vulnerable machines from a few hundred thousand systems for Code Red to several million for RPC, he said.

Concern heightened last week when code designed to exploit the RPC vulnerability appeared on the Internet on July 25.

Soon after the release of the exploit, known DCOM RPC, after the flawed Windows Distributed Component Object Model (DCOM) interface, the Internet Storm Center noted an increase in scanning on ports used by the affected interface, Ullrich said.

Much of that activity is disorganized, however, and does not necessarily mean that a widespread attack or DCOM RPC worm is in the works, he said.

"Most of what we've seen is people using (the DCOM RPC exploit) as part of regular hacking activity: (Web site) defacements or people just compromising machines," he said.

Swap shop However, recent posts to security newsgroups suggest that hackers and computer security experts have been enthusiastically modifying and swapping the exploit code since it was released.

Whereas the original DCOM RPC exploit code worked only on machines running English language versions of Windows 2000, recent modifications show that the code has been modified to exploit the same vulnerability on French, Chinese, Polish, German and Japanese versions of Windows 2000, XP and NT.

RPC is at a stage similar to that of a widespread Microsoft SQL vulnerability after exploit code for that vulnerability was published by security researcher David Litchfield of Next Generation Security Software in August 2002, according to Ullrich.

That exploit code was later modified to create Slammer, one of the most widespread worms to exploit disclosed vulnerabilities .

In its present form, the DCOM RPC exploit code that has been made public is probably not ready for wide distribution as a worm, according to Ostwald.

The code is not fully developed and often relies on variables, such as the presence of particular flavours of Windows, to work, he said.

Proof spoof In contrast, Last Stage of Delirium developed so-called "proof of concept" code for use internally that works against a wide variety of Windows platforms and requires only the IP (Internet Protocol) address of the vulnerable machine to create a buffer overflow, Ostwald said.

Such code would be "very useful" to worm writers, making it easy for a worm to spread from machine to machine, he said.

Hackers are also working on shrinking the exploit code, narrowing the exploit to work on a small set of platforms that will net the most compromised machines, Ullrich said.

While it typically takes a couple months from the time of a published exploit to the development of a worm, the development of a worm that takes advantage of the RPC vulnerability may be influenced by other factors such as media attention or this week's DEFCON conference in Las Vegas, a popular gathering point for hackers and computer security experts, he said.

"These things are really random. It just takes one guy to put in the effort," Ullrich said.