Post a Comment
As far as I can tell there is very little to this story.
Automatic updates to WU only happens if one of the following are true:
* You have notifications turned on,
* You have automatic download turned on,
OR
* You have automatic installation turned on.
If Windows Update is turned off on your system Windows Update will not be automatically updated. It will however be updated (if necessary) when running Windows Update manually.
Microsoft should of course not install updates to WU automatically unless configured so. It should notify users of updates to Windows Update, and letting the user know that further notifications would not be received until WU had been updated.
EDIT: Forgot an "on". And fixed spelling error in "Forget" --> "Forgot".
Edited 2007-09-14 14:14 UTC
Do you have any evidence or at least sources for that claim?
The article itself makes that claim:
It only says "seem". It has not been confirmed to be true for all four settings, and it is a fact that the WU-client cannot connect to WU-servers when running without Administrator rights.
I know from my experiences with my firewall that WU does not attempt to do anything when turned off. And I know that because my firewall freaks out when I run WU and forget to un-restrict WU
Edited 2007-09-14 19:12 UTC
I don't have evidence for the thing he discribed, but I set it to "Notify, but don't download or install".
Yesterday evening I booted Windows and two updates were shown: The one security update and that "tool to remove baad software" (no, I don't know the correct english names for that
). I clicked on cancel since I wanted to do this later, after doing some work.
Half an hour later there was just ONE update left. Guess which one it was.
Okay, that's not really much "evidence" since you can't really prove that without making a video. Guess you just have to believe me here.
Wooot O_o ??? You have an IE7-uninstaller!!1?
Hmm.. I've had my own problems with the notification-thingy when manually launching Windows Update. The notification thing tend to conflict with the manual launch, so I'd need the output from Windows Update's list over installed updates and failed updates for your machine. HOWEVER: You MUST NOT give me that information. I'll haunt you in your nightmares if you do
Launch Windows Update manually and check the list over succesful updates and failed updates and see if it shows something. Until then I can only conclude you had a problem with the notifier.
I have had conflicts with manual launch of WU and the notifier running, but I haven't experienced what you describe. But that malicious software thingy is probably selfinstalling
No. That's the core of the problem. The core of the article was that Microsoft was invading your privacy and modifying the entire OS by installing all kind of updates without your consent.
Remove the sensationalism of the article and you would have had a good story. As it stands now it is just obnoxious second rate "journalism".
It was a stealth update to Windows Update itself. Now of course, they could have asked permission, but that would have looked something like this: http://kroc.deviantart.com/art/Marklark-Marklark-Marklark-32789616
Well, from a security-POV Microsoft could hardly do it much different.
MS of course has the following options:
1) Install updates to WU automatically if WU is set to "automatic updates"
2) Download updates to WU automatically if WU is set to "automatic downloads" and notify the user that the updates are ready to be installed. However! The user must be notified that WU will not work if the updates are not installed.
3) Nofify the user of updates to WU if WU is set to "Notify only". However! The user must be notified that WU will not to work if the updates are not installed.
The downside of doing 2) and 3) is that the user will not know about important updates until the user has updated the Updater (which is difficult to formulate properly, as Kroc proved with his earlier post.. heehee).
From a security-oriented POV Microsoft has not been unethical. From a privacy-oriented and technological POV Microsoft has handled in part unethical and in part incompetent.
But it doesn't mean that MS is trying to take over your computer (they may be doing that, but not through WU).
What Microsoft should have done is to inform of this in WU. No part of a operating system should be self-updating without written notice of what parts gets self-updated, and why.
Also, from a security standpoint, i think this is bad.
WU is propriary. We dont know how it validates new self-updates. What if Microsoft WU servers get hacked, and millions of Windows machines starts to automaticlly download compromised code?
Edited 2007-09-14 14:34
Well for me it ain't a risk, since my firewall blocks Windows Update unless I specifically allow WU to contact the servers
Remember, the Windows Update servers do not contact your machine. It is your machine that contact the Windows Update servers.
If the servers are compromised people using automatic download/installation/notification are in (a rather hypothetical) risk. It's however only a matter of having a properly configured firewall - and not the built-in btw.
Well, since you cannot control what it does when it contacts the remote machine, and you cannot prevent it from changing software on the machine, it is not really your machine any more, is it?
http://www.microsoft.com/technet/archive/community/columns/security...
The really scary thing is that there is now at least one more known way to remotely install code in a Windows system without having to get permission.
All that is required is for the code one wishes to install on the target Windows system to successfully masquerade itself as being an update to WU.
True security would require that an OS insist that a local, authorised administrator of the system manually supplies credentials before any piece of new software can become executable.
It is plain that the Windows OS lacks this most basic security provision.
AFAIK, update notiifiers on Linux clients run with normal priveledges. The notifiers merely inspect what versions are locally installed, and what versions are in on-line repositories. When it is found that updates are available on-line, the notifiers show the user that updates are available.
It still requires the user to manually supply root credentials to the package manager software before the actual updates can be downloaded and installed.
This is the essential difference, I suppose, between an mere update notifier and Windows Update.
On Windows, there is apparently no absolute universal fundamental-to-the-OS requirement to manually supply administrator credentials before downloaded information can be made executable.
On Linux, there are execute permissions as part of local filesystems, that can only be set by a local user. After execute permissions are set, the newly-executable file can then only run with the maximum permission level of the user who provided credentials to set the execute permission bit. Therefore, in order to bestow universal and system-level execute permission, the authorising local user must be root.
Edited 2007-09-15 12:58
dylansmrjones and sappyvcv have already dealt with your arguments from a technical standpoint (successfully, IMO), but I want to respond to a statement of yours from a non-technical standpoint.
"The really scary thing is that there is now at least one more known way to remotely install code in a Windows system without having to get permission. "
It seems to me that given that the vast majority (or, at least a substantial percentage) of home Windows computers have WU set to download and automatically install security updates anyway (that's the default setting), that if someone wanted to infect systems by compromising WU in some way, they would have already done it, or at least already have been trying to do so. This "one more way" doesn't add much from a practical standpoint. It would expand the targeted users beyond the "download and auto-install updates" users to the "download but don't auto-install" and the "check for available updates" users, but the first target is so huge that baddies would have already been trying to compromise WU if so inclined. And any techniques that would exploit this "one more way", that first huge segment of users has already been subject to.
(I talk of home computers because corporate computers likely use WSUS or SMS rather than WU, so the IT staff controls how those machines are updated, regardless of this "one more way".)
Edited 2007-09-15 15:26
This is a typical Microsoft position.
The critical points are these:
(1) It is trivially easy to get data onto a client. Any website can do this. If you want to compromise a system, then getting your data onto a system is not a problem ... the problem is to get your payload data to be executed by the target system, hopefully with elevated privileges. WU has been shown to provide just such a mechanism.
(2) Microsoft's EULA reserves the right for Microsoft to update Windows systems. WU is one way (very likely not the only one) that Microsoft can update any Windows system. WU cannot be disabled as an entry point onto a Windows system, even apparently if the end user selects "never update". This effectively (if not literally) makes WU a "push technology".
(3) If Microsoft can push data onto a Windows system via WU and get it to execute (install) at system-level privileges without end-user authorisation, then that immediately presents an attractive doorway for other parties to try to do the same feat. Even if Microsoft honestly intend for this mechanism to only ever be used for WU to update itself, other parties are in no way constrained by Microsoft's original intentions.
(4) Finally, Windows makes no distinction between "binary data" and "locally authorised to execute program". WU illustrates this point beautifully. WU provides a mechanism (that cannot be disabled by the end user) for an external party to put new code onto the end user's system without permission of the end user and to have that new code enjoy system-level privileges. The fact that the intention is that Microsoft is the only party that can do it is moot ... the fact remains that it can be done. If Microsoft can do it, it remains only a matter of time before some other party finds a way to also do it using the same mechanisms.
This is fundamental security stuff. Either Microsoft just doesn't get security, or Microsoft does get it but doesn't care to provide it for end users. Either way is a very poor reflection on Microsoft.
Edited 2007-09-16 06:45
(1) It is trivially easy to get data onto a client. Any website can do this. If you want to compromise a system, then getting your data onto a system is not a problem ...
Then WU is irrelevant at this point because the machine is already compromised.
the problem is to get your payload data to be executed by the target system, hopefully with elevated privileges. WU has been shown to provide just such a mechanism.
No, actually it hasn't. You've only claimed it has with no proof whatsoever.
WU cannot be disabled as an entry point onto a Windows system, even apparently if the end user selects "never update".
Ah, but it can. It's been stated by multiple people already that setting to "never update" didn't do the updates. Even further, it's quite trivial to disable the WU service.
(3) If Microsoft can push data onto a Windows system via WU and get it to execute (install) at system-level privileges without end-user authorisation,
A side effect of the poor security in versions prior to Vista. However, on other systems, most people would blindly authorize ANY updates done through the Operating Systems update client. Honestly, what percentage of people attempt to verify the update server is valid and the updates they are receiving are valid? Not many.
(that cannot be disabled by the end user)
I know it must be fun to state lies, but please stop already.
Well, there is the difference that Debian machines don't automatically download software from the Debian servers
But yes, any kind of updating service which is hacked poses a severe security threat. But then.. turning on your computer poses a threat in itself. And the greatest threat is sitting 40 centimeters from the monitor.
NOTHING ///
My machine uses RPM (yum or up2date) to install packages.
The packages have to be SIGNED by a private GPG key before they will install.
The key that is required to sign the pacakges is not stored on the update server at RedHat.
SO .. if someone replaces the real packages with fake ones, nothing at all happens on my machine, other than I am told that these packages are not signed by the proper key.
SEE ... don't ask retorical questions when the answer makes you look bad
True; this was raised from another point of view; what happens if someone can find how it works and attract the end users machine from that vector - using an apparent 'legitimate' open door where by the WU can be updated to point to an illegitimate source and thus, ability to deploy false updates which are actually anything ranging from adware to spyware to virus's.
The issue I think which people forget, and you have raised in your post is this; the issue isn't necessarily privacy per say but how this 'technology' can be exploited.
It is very easy to remedy. You simply have the updater update itself when you check for updates.
The updater informs the user about the need for the update every time a manual update check is attempted, they have to do it to use the current update system.
Simple. Very simple. No need to do anything automatically.
Of course, I think automatic updates ( auto-download-install ) is The Dumbest Idea Ever(R).
--The loon
But it can't check for updates without the updater update. In 3 of the 4 settings, the updater automatically checks for updates, which is when you say its ok for it to update itself.
The 4th setting, according to some people, doesn't automatically update the updater anyway.
The fourth approach will however result in Windows Update being updated automatically the moment you manually launch Windows Update. There will be a message with BIG LETTERS stating that it is checking to see you are running the newest version. I can't remember the exact wording now, but WU is updated the first time you run it manually after an update of WU is available.
If this was normal surely it would have happened before, perhaps it is my memory but I can't remember any other instance of it happening.
Asside: Has anybody with the updated files tried to use a 3rd party updates service such as Windiz Update yet?
http://windowsupdate.62nds.com/
"If this was normal surely it would have happened before, perhaps it is my memory but I can't remember any other instance of it happening. "
-------------------
According to the Microsoft blog, Windows Update has been like this since the introduction of XP, and has updated itself in the past many times in the past.
http://blogs.technet.com/mu/archive/2007/09/13/how-windows-update-k...
Just to reinforce what dylansmrjones is saying, according to the Microsoft blog, a user can set Windows update to one of four settings: "1) Install updates automatically, 2) Download updates but let me choose whether to install them, 3) Check for updates but let me choose whether to download and install them, and 4) Never check for updates", and that Windows Update components themselves auto-update for all settings except setting (4). The reason, as stated in the blog, is that if the user is using Windows Update at all, even just to check if updates are available, then the client-side Windows Update components must be kept in sync with the server-side components. According to the blog, the Windows Update components do NOT update themselves if the user completely turned off Windows Update, that is for setting (4).
The earth-shattering story that was reported yesterday is that someone found that Windows Update components were updated for settings (2) and/or (3). But according to the Microsoft blog, "This has been the case since we introduced the automatic update feature in Windows XP. In fact, WU has auto-updated itself many times in the past."
The Microsoft blog also says that this does not affect those that use WSUS or SMS rather than Windows Update, so IT departments are still in complete control of the OS updating process.
The only issue here is that Microsoft should have more clearly disclosed that Windows Update components do update themselves for settings (2) and (3) (that they would update themselves for setting (1) goes without saying). But I'm not sure how to do that in a user friendly manner, because there is such a thing as "too much information" for a normal user to absorb. Probably just add a link to the Windows Update control panel that says "Click here for more information" that refers to a web page explaining in detail what the process is. I doubt normal users care; IT staff would care, but they probably already know and/or are using WSUS or SMS rather than Windows Update, in which case they wouldn't be affected anyway.
I don't think that if the user has set Windows Update for settings (2) or (3) that they should be notified every time that Windows Update components themselves need to be updated, and given the chance to deny that operation. I think it's just too much info and would make setting (2) and (3) too cumbersome for users. But I know that many tech geeks like to know everything that is going on, so they would feel differently about it. But if Microsoft did change Windows Update to allow the user to reject updating Windows Update components for settings (2) and (3), such an option must be accompanied by a big loud warning saying that if the user does reject updating the Windows Update components, then the Windows Update setting will change to setting (4). (I doubt many that intentionally chose settings (2) or (3) would go for that, which is why I think such an option is a waste of time for both Microsoft and the user.)
If you want more details, be sure to read the MS blog.
BTW, the term "stealth" that I saw bandied about yesterday (and sadly, today on osnews) is sensationalistic, as the event logs show exactly what Windows Update component files were updated and exactly when that update took place. A "stealth" update wouldn't record any logs for the event.
Edited 2007-09-14 17:32
Whilst the log exists, the real issue is that on closed source OS, you still don't know actually what has been changed. All you know is some filenames, but that doesn't actually prove anything.
I personally see only a small difference between the stealth update, and accepting an update. Either proves nothing about what code and functionality was actually changed in the update.
Ergo, this article blows things well out of proportion.
"The only issue here is that Microsoft should have more clearly disclosed that Windows Update components do update themselves for settings (2) and (3) (that they would update themselves for setting (1) goes without saying). But I'm not sure how to do that in a user friendly manner, because there is such a thing as "too much information" for a normal user to absorb."
Yeah, if you're a clueless Windows user...
Oh, please, this is just ridiculous. What's so damn hard about saying that the Windows Update process will update itself at those 2 settings. It's not rocket science to explain that.
Better yet, it shouldn't do it. What's so hard about putting in a notification system that says you need to update the Windows updater. If you can set the system to notify you only, notify that WU needs to be updated.
Some people just want to bend over backwards to excuse any stupid thing Microsoft does.
There is one problem with your scenario.
The Windows Update servers don't send anything out. They don't contact clients. It is the clients that contacts the servers. As such Windows Update is only a little more insecure than gentoo Portage, FreeBSD ports or Redhat repositories. Unless of course you have turned automatic updates/automatic downloads/notification on. It is easy fixed though. Turn off WU and launch Windows Update the first tuesday every month 
Windows Update cannot run unless it has Administrator rights.
And it doesn't do anything in stealth mode unless you put it in stealth mode.
If you configure your gentoo box to automatically update (a cron-job) itself without notifying the running User you would have the same situation.
gentoo does not ask for the root password. It just fails to run because of missing rights (just like Windows Update)
- unless it is configured to allow the user to run it as normal user. I don't recommend that. Use sudo, please.
In pre-Vista Windows a Limited User Account (or Restricted User Account) cannot run Windows Update. A user cannot even receive notifications. The problem stems from most users running XP with Administrator rights. Imagine that!
WTF!? Anything to backup THAT claim?
My machine cannot connect to anything when no users are logged-in. And Windows Update does not run at all when the logged-in user do not have Administrator Rights. At least that is true for pre-Vista Windows.
I have several xp boxes behind a Firewall One firewall with content filtering turned on. And if the article is true then the versions are on there and no one has been loggin to this computer for quite some time and no admins for sure. We use SMS to do our up dates and have not installed this one. Through GP we have turned off AU because we use SMS.
\rvailc$WINDOWSsystem32SoftwareDistributionSetupServiceStartup wups2.dll7.0.6000.381
WU does not work when the logged in user does not have Administrator rights.
You can try for yourself. Log on Windows Update with a LUA (RUA in Win2K and Win2K3). It fails. Even with a power user does it fail.
I'll have to check it to be sure, but it'll have to wait. I'm not in the mood for rebooting (into Windows) 
Microsoft own your copy of Windows so they can do pretty much what they want to it. It's besides the point that the updates NEED to be done, but people dont want to be force fed by a spoon from them.
For the Windows users, it's like it or lump it and you should be used to it by now. Windows users just put up with it and moan and it's always the way.
Hmm... all my programs run fine as Restricted User.
Of course Installers need to run with Administrator rights. Use "Run as..."
Oh, and yes. Players utilizing DirectShow-filters need to run as Administrator to work (almost) flawlessly.
Apart from that I only have one application (from 1997) that doesn't work properly. That has been fixed by giving my normal user read+write permissions for that applications folder in "Program Files". I have another one which needs extra permissions to run, but it is not installed since changes in the XML-format on Hattrick.org has rendered the application useless.
Everything else works correctly. Only admin tools require Administrator rights, and rightly so.
On Tuesday AM, out of the blue, my copy of Vista self-destructed when MS "determined" I wasn't using a legitimate copy. This was with a pre-installed OEM version on an HP laptop I purchased a few months ago; hence, no activation required. Yet, I woke up one morning and logged in to a screen telling me my version of Windows Vista was had an invalid activation key and that I could just go and bugger off because it wasn't let me going to do anything.
In fairness, it was reasonably quickly resolved with a call to the MS activation hotline, but why the f*ck should I have to call MS for permission to use my computer when they randomly decide to shut it down? Particularly when it was purchased on a system to avoid activation headaches? It really irks me because the main reason I use Windows is for work, and it would have been really fricking embarrassing to show up at a customer site for a training preso with a locked laptop because MS had a hiccup.
As far as I'm concerned, MS needs a little more scrutiny on this from legal authorities. I'm still bitter. But it underscores the fact that users do need to realize that as long as they are connected to the internet, they have no true idea of what is occurring communication-wise between their system and Microsoft.
edit: typo
Edited 2007-09-15 02:31 UTC
In this scenario, only the update software is being updated without user notification and not the system itself. That really isn't a big deal.
To be honest, I wish the Linux distribution I use had some sort of automatic check for updates mechanism. I have to do it manually from command line everyday.
What distro are you on? If you mentioned that, you would have had a solution already. I use adept notifier, which adds some stuff to /etc/apt/apt.conf.d so I can just change the values to auto check and download. Combine that with anacron so the machine doesn't need to be on at update time, and life is very good.
It checks (apt-get update) and downloads (apt-get upgrade -d) without installing anything, then the adept-notifier pops up a little thing in the lower right of the screen, and I can install quickly when I want to without waiting for the download.
It seems Vista runs on a need to know basis...
Sometimes my hard drive is thrashing away and I don't have a clue why. I'm not doing anything, so why is windows. I have the knowledge to install process wathcers and HD activity monitors but I want an OS that keeps my informed at all times whats happing in the background. That does not mean it has to throw all that info in my face but I should be able to get a low down on what exactly is happing at any given time I choose to know.
"Sometimes my hard drive is thrashing away and I don't have a clue why. I'm not doing anything,"
Would you rather it be thrashing your drive while you're <em>using</em> the machine?
Vista will adapt to your usage patterns and try to do background stuff whenever it thinks you're not using the machine.
He didn't say he was not using the machine. He said he was not doing anything at the moment.
When the drive starts trashing and Winblows is doing it's "cloak&dagger" stuff, you can't use the machine because it's not responsive.
That's what Microsoft claims. You worked on Vista and have access to the source code that does that to back up that claim?
Microsoft makes a lot of claims, none of which are true though.
And in practice when using the OS it's very easy to see that. As the poster said about the disk trashing for no reason at all, and many can confirm that, myself included (yes, I work on Vista regularly too).
So no, Vista doesn't adapt to anything and it most certainly does not think. It just barely runs if you don't ask it to do too much.
GG, insert coint to play again Jason.
There you have a botnet that can take down COUNTRIES.
What freakin' difference does it make whether the update is automatic or not? Have you ever manually scanned each and every bit of every package update on your Linux machine? Whether it automatically updates or you manually tell it to update, either way, if the server is hacked you're going to get screwed.
Oh, wait, most Linux update services use signed packages? Unless you're telling me the Microsoft updates aren't signed, that's still irrelevant - Linux updates (manual or otherwise) would still be just as dangerous as any Windows automatic update - which would be "not very dangerous at all."
Unless of course the build servers for package updates are hacked, but then, I'm willing to bet that Microsoft's build server network has a more secured setup for that than any Linux distribution. Microsoft's build servers aren't community accessible, for one.
I've had Linux machines die after a bad manual update. Manual updates guarantee nothing at all, except that the updates are less likely to get applied in a timely manner, if at all. (Almost every Linux server I'v worked on as a contractor never had any updates applied to it until I came along. Wouldn't be a problem if they were automatic. Except, wait, binary compatibility is EVIL and only good for proprietary software, and has nothing to do at all with being able to safely update a live system and expect all of its existing self-compiled/installed software to keep working.)
The difference between automatic, and tell me the updates so I can say yes is that it allows me to determine when to install (in case of a kernel update that requires a reboot or glibc update that requires daemons to be restarted) and I can look at what is about to happen so that if the update borks the machine, I have a chance of knowing what actually broke things, making recovery easier.
Oh, wait, most Linux update services use signed packages? Unless you're telling me the Microsoft updates aren't signed, that's still irrelevant - Linux updates (manual or otherwise) would still be just as dangerous as any Windows automatic update - which would be "not very dangerous at all."
Ummm, Linux distro maintainers are TRUSTED by their users, Microsoft IS NOT trusted by anyone (can yo spell ANTITRUST ?)
You'd most likely lose that bet.
Was the source code to Win2k not stolen a few years back?
Besides, Microsoft treats its network as a bank, Linux distro sites are like community centres, yet they still managed to be VERY secure and trustworthy.
Well, that's a good testament to the quality and security by design of Linux. You can have a server running without any updates for a long time and it stays secure and running well without any updates.
Intall Windows without any updates and it's hacked inlike what, 15 minutes?
BTW, I took off a point off your post because you posted a rant without thinking. Hopefully you're not doing that when working as a "contractor" on Linux servers.
It's operational.
I had a client call me fifteen minutes ago. His workstation PC rebooted after this update. Then he can't remote login and his email didn't run in the morning, so his email isn't ready for him to read when he gets in.
If Microsoft didn't have this idiotic necessity to reboot every damn time it modifies the OS, it would be a lot easier to recommend automatic updates.
This is why you never do automatic updates on a server.
But it can be a problem even on a workstation.
First, Microsoft signs just about every binary they produce. If you've ever downloaded a Word doc from Microsoft's site, sometimes you used to find that it was wrapped in a self-extracting EXE. This was done solely so that they could put a digital signature on it. Obviously the private keys used for Windows are very tightly guarded and they are not likely to be on the Windows Update servers. Likely they are only available on the machines that do "Official" builds of Windows (a small, tightly-controlled set of machines).
WU updating itself automatically when enabled is an interesting decision. I bet there was a long discussion about this at some Windows group meeting before they came up with a consensus to accept this behavior. The choices are twofold:
1) Only inform the user of a single update (the WU update) and potentially require two update check/install cycles to fully update the machine. Pros: user gets full control of _every single_ update to his or her machine. Cons: A user has no idea how many updates they will actually get when a WU update comes out and may not notice further updates stuck behind the WU update.
2) Current behavior. Pro: Having a single update cycle makes it less likey that a user will fail to notice important updates after the WU update. Con: User loses control of the machine with regards to updates to WU itself. Mitigation: WU components should have pretty much no application compatibility or stability impact on the rest of the system.





