Worst Windows flaws of the past decade

June 25, 1998, and June 30, 2008, marked two important milestones in Microsoft's evolution of the Windows OS -- the passing of the torch from Windows 95 to Windows 98, and the less seemly transition from XP to Vista.

In the 3,659 days between, users of Windows have been forced to bear witness to another evolution of sorts: bugs that left Windows open to exploits that appeared almost as fast as you could say, "On the Origin of Species."

Uncovering -- and exploiting -- Windows vulnerabilities has made sport for many and careers for many more. Entire industries have sprung up to protect Windows users from previously unknown flaws, while malware authors have matured their practices from juvenile pranks to moneymaking criminal enterprises.

Caught in the middle of this never-ending onslaught is the innocent PC user and the besieged IT admin -- you. And though Microsoft and the entire software industry have labored tirelessly to handle zero-day exploits and to develop protocols for reporting potential security problems, we've seen and experienced several colossal security meltdowns thanks to the humble Windows bug.

These errors, buried in millions of lines of code, have steered great corporations and turned the tide of fortunes. It's high time they got the credit they deserve. Here are the worst Windows flaws we've endured since the introduction of Windows 98.

Password "password" would have been more secure

Bug identifier: VCE-2000-0979, MS00-072

Description: Share Level Password vulnerability

Alias: Windows 9x share password bypass

Date published: October 10, 2000

Windows 9x introduced a nifty little concept wherein users could host a password-protected mini file server, aka a share, on their PCs. The idea was simple: Allow users of networked computers to host and share files securely. Only the padlock Microsoft used to lock the door came equipped with a gaping hole that rendered it useless.

"When processing authentication requests for a NetBIOS share, Windows 95/98 would look at the length of the password sent by the attacker and then only compare that number of bytes to the real password," writes vulnerability expert H.D. Moore, who manages the Metasploit Framework project.

Oops. "This let the attack specify a password of zero bytes and gain access to the share," without actually knowing the password at all, Moore explains.

"The real damage," he continues, "was that by trying all characters of incrementing lengths, they could literally obtain the password for share from the server."

Upshot: Rather than functioning as a lock on a door, the password authentication scheme for Windows 95/98's File and Print Sharing acted more like a nail through a hasp -- to open the door you only needed to pull out the nail, with hardly any effort.

Page Break

Folder traversal: Total server control with a single URL

Bug identifier: MS00-078

Description: Web server folder traversal vulnerability

Alias: Directory traversal bug

Date published: October 17, 2000

If there's one thing we've learned from the past decade of Microsoft patches, it's that not everyone keeps on top of them. When Microsoft published this particular advisory, the patch that fixed the problem (MS00-057) had already been released two months prior.

With this bug, if you knew the layout of a Microsoft file system -- which folders appear where -- you could send a command to a Web server that essentially gave you total control.

As anyone who has spent any time using a Windows computer will tell you, it's not hard to find your way around the hard drive. Documents go in a particular folder path; most applications are put in another folder path; and so on.

By using dots and backslashes (or their respective unicode representations) in the URL, this bug allowed you to navigate up and down the file system and execute commands, just by knowing a few simple rules and how Windows organizes itself. While account permissions for IIS are somewhat limited, a related exploit helped escalate privileges, giving remote users the ability to do whatever they wanted to with Windows servers simply by sending a few URLs.

"Originally found as an anonymous post in the PacketStorm forums, this resulted in nearly two straight years of mass ownage against Windows web servers," Moore writes.

Upshot: Directory traversal opened up a new world for automated attacks that merely had to call a particular URL to do their dirty work.

Code Red: Deadly bug, disgusting soda

Bug identifier: MS01-033

Description: Unchecked buffer in index server ISAPI (Internet Server API) extension could enable Web server compromise

Alias: The Code Red bug

Date published: June 18, 2001

What happens when you send a ton of data at a Microsoft Web server? If it was the summer of 2001, well, you owned the network. At least that's what happened a little more than a month after Microsoft released this obscure-sounding patch for IIS Web servers.

The nature of the bug was simple: Take an IIS server, invoke a buffer overflow, and commands spill into other parts of system memory. Because the commands were issued in the context of the system itself, the bug opened up for exploitation virtually all aspects of the server's operation.

And exploitation happened, all right, on a scale that hadn't been seen before.

On the afternoon of Friday, July 13, 2001, security engineers at eEye Digital Security received reports of a worm that was spreading rapidly through its customers' networks. Fueled by a limited edition, crimson, caffeinated, high-fructose corn syrup-based beverage, Mark Maiffret and Ryan Permeh spent a weekend reverse-engineering the worm, and alerted the world to its presence.

What the worm did was probe vulnerable IIS servers, infect them, and create 100 threads of itself, which then spread to other computers. If the date was between the 20th of the month and the end of the month, it would attempt to spew data at www.whitehouse.gov. Permeh and Maiffret estimated that the worm could infect approximately 500,000 unique IP addresses per day.

Upshot: Code Red really drove home the importance of patching bugs soon after Microsoft released the patch, because the patches themselves give malware authors clues to exactly where they should look for new vulnerabilities.

Page Break

Fastest infection. Ever.

Bug identifier: MS02-039

Description: Buffer overruns in SQL Server 2000 Resolution Service could enable remote code execution

Alias: The SQL Slammer bug

Date published: July 24, 2002

While technically not an OS bug, the SQL Slammer bug deserves honorary mention due to the sheer velocity with which vulnerable systems were infected. The bug targeted Microsoft's database server. Vulnerable computers were subject to buffer overflows that, if properly crafted, could place commands into memory to cause the targeted system to execute those commands with the permissions of the database service.

Patching was complicated by the fact that admins needed to run an earlier patch before they could run the MS02-039 fix. The bug affected primarily corporate server systems, but also affected home users who had MSDE (Microsoft SQL Server Desktop Engine) installed. That made a number of home users, some of whom didn't even know they had MSDE on their machines, unwitting participants in the carnage to come.

Because the Slammer worm primarily targeted servers running databases, it didn't infect millions of machines. It did, however, spread rapidly -- so rapidly, in fact, that it had infected roughly 9 out of 10 vulnerable machines within 10 minutes of being released on January 25, 2003. The entire worm was only 376 bytes, and fit into a single packet of data.

The MS02-039 bug was "one of the biggest oversights of all time," says Steve Manzuik, senior manager of security research at Juniper Networks, "not because it was an easy or obvious bug to find -- it wasn't."

"At the time of the patch, no one realized that every vulnerable SQL installation was also listening on a UDP (User Datagram Protocol) port that they could be exploited over," Manzuik explains. "Many administrators simply locked down access to the SQL TCP ports while forgetting about UDP."

A postmortem by the Cooperative Association for Internet Data Analysis revealed that the worm was a model of efficiency, doubling the number of infected systems every 8.5 seconds, and flooding the Internet with so many infection attempts that routers shut down. When restarted, so many routers attempted to update their routing tables simultaneously that normal Internet traffic simply couldn't get through the gridlock.

Upshot: SQL Slammer demonstrated the power of a vulnerability that could fit within a single data packet, and brought home the lesson that a single application weakness could cause the entire Internet to grind to a standstill. And it's still out there, drifting around on a few old systems, looking for new hosts to infect.

Page Break

Billy Gates, stop making money! Make malware instead.

Bug identifier: MS03-026

Description: Buffer overrun in RPC interface could allow code execution

Alias: The Blaster Worm bug

Date published: July 16, 2003

The DCOM RPC interface is a common component of NT-based Windows OSes, including NT, 2000, XP, and Server 2003. In the summer of 2003, it became the subject of intense scrutiny.

As Microsoft described in the bulletin that accompanied the patch, a successful exploit only required the attacker to send a "specially formed request" to a vulnerable PC -- a bit like dangling candy in front of a ravenously hungry baby.

By August 11, the Blaster worm arrived, and though it spread rapidly, it was fairly easy to block with a firewall.

Unfortunately, protecting home systems with firewalls wasn't common practice at the time. Home users' PCs -- connected directly to the Internet -- got whomped by the worm. When the worm's code crashed the infected computer's RPC service, the computer would display a message warning of imminent shutdown, and unceremoniously reboot itself.

The worm had another message, this one to Microsoft's founder, and embedded within its code: "billy gates why do you make this possible? Stop making money and fix your software!!"

But it was fixed. Or at least it would have been if people had patched their systems.

At the end of the summer, Microsoft released a second set of updates in MS03-039 that blocked additional ports that attackers could use to mess with the RPC service.

Upshot: We're all in better shape thanks to the wide adoption of firewalls in the home. Thanks in part to Blaster and its ilk, most broadband modems have one built in.

That sassy bug has a lot of spunk

Bug identifier: CVE-2003-0533, MS04-011

Description: Stack-based overflow in certain Active Directory service functions in LSASRV.DLL

Alias: The Sasser bug

Date published: April 13, 2004

In yet another example of ironic buffer-overflow goodness, this bug made the security subsystem of Windows the agent of evil itself. And, once again, malicious coders used Microsoft's own patch to figure out exactly where to target the OS.

As Windows XP's gatekeeper, LSASS (Local Security Authority Subsystem) manages the permissions of a PC's user accounts. So when eEye -- the same company that discovered the Code Red bug -- quietly disclosed the details of this flaw to Microsoft in October 2003, it touched off six months of furious coding in Redmond that culminated in a patch that fixed 13 other Windows 98, NT, 2000, XP, and Server 2003 flaws, as well as the LSASS bug.

And, within 18 days, the Sasser worm was cruising the Internet, hopping from one unpatched machine to another. The poorly coded worm wreaked havoc, shutting down networks around the world. Even though a fix was already available, many users -- in particular, corporate IT managers -- still had not applied MS04-011. By May 1, 2004, work on fixing the unintended damage caused by Sasser had become a round-the-clock operation, says then director of the Microsoft Security Response Center, Kevin Kean, with "a number of war rooms and rotating shifts" for MSRC staffers.

Upshot: What was that about patching as soon as the updates are available? Lessons that should have been learned three years earlier didn't really sink in until Sasser publicly pummeled patchless PCs to pulp.

Page Break

WMF: Wherein malware is foisted

Bug identifier: CVE-2005-4560, MS06-001

Description: Vulnerability in graphics-rendering engine could allow remote code execution

Alias: Windows Metafile vulnerability, aka drive-by downloads

Date published: January 5, 2006

Over the holidays in 2005, security researchers began discussing a newly discovered vulnerability in a Windows library used by the OS to display various kinds of graphics in apps and the OS itself.

The problem stemmed from a particular image file format, native to Windows since the days of Windows 3.0, called WMF (Windows Metafile). Used as the native format for storing graphics within Microsoft Office documents, support for WMF was by that point thoroughly embedded into Microsoft products.

WMF files contain function calls that a program sends to the GDI (Graphics Driver Interface). Someone discovered that WMF files can contain executable code as well. This would allow you to, say, create a WMF file that, merely by being viewing, invokes Internet Explorer to visit a particular URL, download a file, and execute that file. Special.

The aftermath of the discovery followed a familiar pattern. Microsoft issued a patch on Jan. 5, 2006, in record time. But for a long while, unpatched computers running vulnerable versions of gdi32.dll roamed the Internet, slurping up mountains of malware.

The bug had far-reaching effects, enabling malicious code to be foisted on unsuspecting users and executed in a variety of ways: previewing an e-mail containing the malicious WMF file in Outlook; viewing an image preview in Explorer; viewing a malicious WMF in certain third-party graphics programs; indexing a hard disk that contained a malicious file; following a URL link in an e-mail, IM, or on another Web page to a site where the malicious file was embedded in the Web page.

Upshot: We learned that nothing is sacred, that any file format could be considered hostile. And we also got a cool new name for an exploit method: drive-by downloads.

MDAC: The component that keeps on giving (headaches)

Bug identifier: CVE-2006-0003, MS06-014

Description: Vulnerability in MDAC (Microsoft Data Access Components) could allow code execution

Alias: MDAC RDS.Dataspace ActiveX bug

Date published: April 11, 2006

Way back in 1998, Microsoft issued a security bulletin about a component of IIS that ran under Windows NT Server called Microsoft Data Access Components. In the bulletin, MS98-004, Microsoft warned that a part of MDAC called the RDS (Remote Data Service) had a vulnerability that allowed unauthorized people to browse databases.

Flash-forward eight years to the spring of 2006. Microsoft released a security bulletin about a component of MDAC called RDS, which has a vulnerability that permits malicious Web servers to perform drive-by downloads against the unpatched PCs of unsuspecting victims. Eerily familiar.

In the later case, it was an ActiveX control that allowed users to connect to RDS through IE and wreak havoc. The ActiveX control doesn't behave as intended, and can be loaded and exploited if you visit the wrong Web site.

Of course, by 2006, MDAC isn't just loaded on servers; you may have it on your PC. Moreover, the bad guys have changed tactics. No longer content to wait patiently for you to happen upon their malicious Web site, they spam you with links, buy ads based on Google searches, and load their pages with SEO (search engine optimization)-rich keywords. The result, however, is the same: Visit and be exploited.

In fact, the bad guys are now using off-the-shelf exploit software to put malware onto your machine. A tool called MPack that's loaded on malicious Web sites can check to see what browser version you're using and what patches you have installed. Based on this analysis, it delivers the exploits that will do the most damage. More galling is that they don't even bother to hide what they're doing, naming the Web page that performs the exploit "mdac4.php."

Upshot: The MDAC RDS is a complex system, with a multitude of patches available depending on which version you have installed. Manually choosing the right patch can be a complicated task. But with such a serious flaw, you can't afford to make a mistake. Patches like these have helped push advancements in Windows Update, which scan your system and pick the right patch automatically, so you don't have to.