A massive hack of legitimate Web sites has been spreading malware to visitors' PCs, using a new tactic that has made detection "extraordinarily difficult," security experts said Monday.

According to the researcher who broke the news, the hack, which involves several hundred sites, may be related to a November 2007 break-in at Fasthosts Internet, a UK-based hosting service that in early December acknowledged that some clients' log-in credentials had been pinched.

"All of the affected domains either have or have had a relationship in the recent past with Fasthosts," said Mary Landesman, a senior security researcher at ScanSafe Inc.

Like the large-scale hack prompted by a series of SQL injection attacks last week, the sites identified by ScanSafe are legitimate URLs, many of them small mom-and-pop e-commerce sites. Among them are a bicycle shop and several tightly focused travel sites, said Landesman. Most, although not all, are businesses located in the U.K.

But in almost every other way, this was a different, and much more sophisticated, attack. "Usually, in an attack like this, code is injected into the [site] pages, and that code is static," Landesman explained. "In these, it's completely different. The JavaScript that was being created and the reference [to it] was being generated dynamically."

That has given researchers, including Landesman, fits. In typical attacks, such as the SQL injection mass compromise, researchers can easily capture the attack code, analyze it and even use Google to search for other infected sites. Not so in this case.

"The method made it extremely difficult to tell who else might be impacted," Landesman said. "Here, the JavaScript was being generated dynamically and the file name was random, so we didn't have a common denominator. From a forensics standpoint, that hampers things greatly."

Paul Ferguson, network architect at Trend Micro, seconded that. "It makes it extraordinarily difficult," he said.

The timing of those first attacks and the network break-in of Fasthosts was no coincidence, Landesman said. "The timing is what made me wise to the Fasthosts security breach," she said. "We've seen this attack vector only a couple of times since November 2007," the same month Fasthosts said it first noticed the theft of clients' passwords and usernames. "Timing is everything."

She speculated that, once armed with access to sites managed by Fasthosts -- the company's breach notification specifically mentioned that FTP passwords had been snatched -- criminals planted malware on the servers that they only recently activated. "Other malware is involved" that has aided and abetted the sites' hacks, said Landesman.

Visitors to the compromised domains have been assaulted with multiple exploits, notably one for a vulnerability in Apple's QuickTime media player that was patched only last month. Another exploit being served, said Landesman, is the "tried and true" attack against Windows 18-month-old MDAC flaw.

If successful, the client-side attack infects the PC with a variation of the Rbot Trojan, a backdoor also known as Zotob that has been active since the middle of 2005. There, too, however, users are at special risk. "Just three out of 33 antivirus vendors detected that [variation]," Landesman claimed.

To top it all off, this hack's impact is magnitudes larger than the SQL injection's. "That accounted for just 1% of the blocks [ScanSafe instituted for its customers]," Landesman said. "This [new] one accounts for 18% of all blocks. That's a pretty tremendous number."

Fasthosts did not respond to a request for comment on a possible connection between its 2007 breach and the recent site compromises.