Print - Researchers Crack Windows Passwords in Seconds

Researchers from Switzerland have developed a scheme that lets them crack most Windows passwords in about 13 seconds, reducing the time it takes to break these codes by more than a minute and a half. The scheme enforces a growing concern in the security community that the way in which Microsoft encodes passwords in Windows is inherently weak, opening the door for cracking programs to use brute-force methods to test and break passwords.

Philippe Oechslin, one of the Swiss researchers, recently published a paper online called "Making a Faster Cryptanalytic Time-Memory Trade-Off," which highlights the new password cracking scheme. According to the paper, the cryptanalytic time-memory trade-off is based on a technique first developed in 1980. "As an example we have implemented an attack on MS-Windows password hashes," the researchers write. "Using 1.4GB of data (two CD-ROMs) we can crack 99.9 percent of all alphanumerical passwords hashes ... in 13.6 seconds whereas it takes 101 seconds with the current approach using distinguished points. We show that the gain could be even much higher depending on the parameters used."

Oddly, the researchers weren't really interested in cracking Windows passwords, but were rather trying to demonstrate the previous theoretical cryptanalytic time-memory trade-off technique. They note that Microsoft's passwords are weak because, when encrypted, they do not include any random information. Thus, the same password on two different Windows machines will always be the same when encrypted as well. This makes it much easier to break the password encryption, they note.

While its possible to generate more secure passwords by using non-alphanumeric characters and other special characters, the researchers say that even this won't solve the inherent problem in Windows because all they'd need is more time or a larger data set (or both) to crack those passwords as well.

I read this article on CNET, it suggested that the weakness was in the LANManager hashing method, but did not say that this scheme was used in Windows 2000 or XP - does the weakness apply to these Operating Systems as well? Editor's note: Only in certain situations. However, the way passwords are encrypted under the new system is identical; that is, Microsoft still doesn't add random characters, so the same password on different systems results in the same encrypted password. --Paul

This is not your regular Windows password. This is the backwards-compatible, deprecated, horribly obsolete LAN Manager password format which Windows sends out to down-level servers (Windows 95 and Windows for Workgroups). It's really insecure anyway - the researcher demonstrated a way to improve cracking speed from about 11 seconds to 5. To prevent this ever happening, use a password that the LAN Manager scheme can't encode - non-alphanumerics, including characters entered with Alt+numeric pad, longer than 14 characters. When talking to other computers in an NT 4.0 SP4 or higher domain, or Windows 2000 or Server 2003 domain in backwards-compatible mode, or in a no-domain environment, Windows NT 4.0 SP4 and higher uses the NTLMv2 protocol. In this scheme, the password is never sent across the wire: instead, the hashed password is used as the key to encrypt a random number (the challenge) sent by the server, and the resulting number is sent to the server as the response. In Windows 2000 and Server 2003 native mode domains (with no NT 4.0 domain controllers), the Kerberos protocol is used for authentication. Administrators can set a group policy (with the Group Policy Editor console, GPEDIT.MSC) to refuse LAN Manager and NTLM version 1 connection requests, and to prevent sending of either type. For this to work, downlevel clients e.g. Windows 95 and 98 will need to be updated with the Active Directory Client Extensions, so that they can use the NTLMv2 protocol. This software is shipped on the Windows 2000 Server CD. More information in MS Knowledge Base articles http://support.microsoft.com/default.aspx?id=147706 and http://support.microsoft.com/default.aspx?id=299656.

To be fair to MS, the biggest cause of weak passwords is the user. I remember an admin whoose password was the name of his daughter - how predictable, and the number of people whose password is "password". Gahhh! Unix passwords do include random info, but I think the "root" user system is a bit stupid. At least you can disable and never use the Administrator account on Windows.

Root account stupid? On Windows there is a System user account, read GOD, that not even the Administrator can get access to. Many worms and exploits have been able to take advantage of this to devastating effect. You can resrict root on a Unix system. It's generally called SETUID. At least Unix-like systems only have one root account, as opposed to the stupidity of multiple Administrator accounts on Windows. If you have multiple Administrators you never know who is doing what.

Actually Paul in the editor’s note below you are only mostly correct. The way that password are encrypted COULD BE identical, if you are referring to the LM hash or NTLM hash. However you can remove the LM/NTLM hash from the system by requiring NTLMv2 and using group policy. NTLMv2 is salted and is therefore immune to precompiled database cracks (the kind used in the research project referenced in the article) in addition to many other benefits. As always, it seems you can make Windows as secure as any operating system, its just that most people don't know enough or don't bother. CK Editor's note: Right, you "could" do that. But the problem is that it should be done right by default. In the end, any Windows install is only as secure as the people who set it up. But that's true of Linux too, right? --Paul

Paul You need to make clear which kind of Windows passwords were cracked and under what conditions. I doubt the Swiss folks cracked Kerberos hashes. There is no good reason for a serious enterprise to still be using LanMan hashes! Also, I question whether the later LanMan2 hashing scheme does not, in fact, use a salt vector (a random vector that prevents identical passwords from generating identical hashes.) Larry Editor's note: The way Microsoft encrypts stored passwords is the problem, it has nothing to do with Kerberos authentication, which protects passwords over insecure network connections. --Paul

Odd how you delight in putting up articles that put Apple in a bad light. But here you did some clever editing and skipped out the part how OSX is much more secure to this attack. See http://maccentral.macworld.com/news/2003/07/24/password/ Shame on you. Editor's note: No, no: Shame on you. My real name is attached to everything I write. --Paul

Our organization requires password changes every 90 days. I believe this became a tradition back when everyone had access to /etc/passwd but it took a month or so to crack a hash. Obviously this is no longer the case since more secure encryption is available. Changing passwords causes endless confusion for users, and I can't find any good evidence that it increases security in the real world. Any ideas? Thanks, Dan Woodard Merritt Island, FL