Slashdot Mirror
UAC Whitelist Hole In Windows 7
David Gerard writes "Microsoft tried to make Vista secure with User Access Control (UAC). They relaxed it a bit in Windows 7 because it was such a pain in the backside. Unfortunately, one way they did this (the third way so far found around UAC in Windows 7) was to give certain Microsoft files the power to just ... bypass UAC. Even more unfortunately, one of the DLLs they whitelisted was RUNDLL32.EXE. The exploit is simply to copy (or inject) part of its own code into the memory of another running process and then telling that target process to run the code, using standard, non-privileged APIs such as WriteProcessMemory and CreateRemoteThread. Ars Technica writes up the issue, proclaiming Windows 7 UAC 'a broken mess; mend it or end it.'"
Hey, if security was easy, everybody would do it.
http://www.geoffreylandis.com
I still think that Microsoft will have a very hard time prying customers away from the fiercer of its competitors: WIN XP.
In all the financial institutions I work with, or know, WIN XP is the validated standard, and as far as I know no one takes the XP "expiry date" seriously, so no plan B is in place.
This is still in Microsoft favour, since no one is actively pursuing things like ubuntu/open office or such, but it's anyone's guess how long this state of grace will go on; after all, many applications work in terminal emulation, which is an ancient technology by any standard; why use Vista of Windows 7 for that?
"If a boss demands loyalty, give him integrity. But if he demands integrity, give him loyalty." (John Boyd, 1927-1997)
Microsoft's approach to security is like putting too much air into a balloon! And when exploiters find a way around their measures, it's like.. a balloon, and... something bad happens!
Aren't you glad this was caught in testing? Yeah, I am too.
Hail Eris, full of mischief...
E pluribus sanguinem
I had my try with UAC and came to the conclusion that it's just a lose/lose situation for Microsoft.
Lose 1. They're basically advertising to users that "The feature you're about to use is buggy as hell and totally insecure, so you'll have to accept the responsibility for using it". Great way to sell a product.
Lose 2. It's so annoying, people just turn it off completely, thus negating any "security" it supposedly provides
The only upside is that they insulate themselves legally by having the user do the "not recommended" thing whenever they use the OS. Then again, they've never been much to accept responsibility for security problems anyways, it's kind of a moot point.
Let's see, how long did it take for M$ to realize many users weren't thrilled with IE and it's so called security? I'm betting UAC is here to stay for a loooooong time. They will just keep trying to patch it and in the process further irritate users.
Normally I ascribe all life to intelligent design, but in your case I'll make an exception.
First, let me say where I'm coming from. I've been using Linux for over twelve years; I have two full-time Linux servers at home, and a desktop and a laptop that both dual-boot Linux and Vista. I have an XP box and a Linux box at work, where I'm a Linux/Windows sysadmin and programmer, and I do most of my serious stuff there on the Linux box. At home, I stay in Linux most of the time, and I just boot into Vista when I want to run iTunes, or a game, or something else that only runs on Windows.
That said, I actually like Vista. As I see it, its main problem is that is needs a fairly hefty machine to run it. If you're trying to run it with less than 1G of memory, or a not-very-fast processor, forget it. It certainly works for me.
And I don't mind UAC at all. When it comes up, it's usually trying to tell me that I'm about to do something that may have serious consequences, and that I need to think about what I want Vista to do before I press OK. It just takes a moment, really.
So why is everybody complaining about it? Have I missed something?
It has great documentation and with NoScript I feel safe everywhere on the Internets.
You "no script" people are so funny with your need to Slashdot brag about using the internet without scripts. Yes, we get it, you're so amazing! The internet without scripts, wow that's so neat!
Is that whiny users want something that magically protects them, but doesn't bother them. That's a nice idea and all, but you can't have that. You can't have it both ways with something like this: Either it is a real separation of privileges like it is in Vista, or there's going to be holes.
Well, they gave people the real security that they'd been crying about with Vista. When UAC is on it is a no bullshit, you have to escalate to do things as admin. There aren't exceptions or the like, you escalate when you need admin. This does mean it asks in a lot of situations. Well, there's no avoiding that. Like I said, no exceptions. It is also very granular. It isn't one of these "Oh just click it once and we'll escalate everything for the next few minutes," things. That again would be insecure. No, it is per item. That thing and that thing only gets the elevated privilege.
But people whined and bitched, including many of the same people who whined and bitched in the first place, so now they are backing off. Well, as part of that, you open up some potential holes. Sorry, but that's just life. If there are exceptions to the rules, then something can make use of those exceptions.
You can't have a system that magically knows what the bad apps are, and only asks permission on those, well at least you can't without some sort of draconian trusted computing BS. That's what users want, but they can't have it, it isn't possible. Thus you've got three choices:
1) Allow everything for administrators. Assume the admin knows what they are doing, and let them do whatever they want. Don't ask for permission for any action. This is the Windows XP method. It's very convenient, but also means that you'd better be careful.
2) Have truly separate permissions, and require escalation. Everything has to go through the procedure, no exceptions. This is the Vista method. Means you get asked a lot (though personally I don't find it bad at all) but it is secure. Nothing gets to slide through because there aren't special cases.
3) Have separate permissions, but allow exceptions to make things easier. Ask only in certain situation, or only so often. Just let everything else go by. This is the Windows 7 method (and also several variants of Linux I've seen). Fairly convenient, and more secure than #1, but only superficially so. Because there are exceptions, there are back doors for things to sneak through.
So really, users have to come to terms with what they really want. The "I want it to protect me from bad things, but not bother me," doesn't work. That is akin to saying "I want security to make sure nobody sneaks a weapon on a plane but I don't want to go through a security checkpoint." No, sorry, doesn't work that way. If it is really going to work, then it has to be consistently applied to everyone or everything.
"Ending is better than mending. The more stitches, the less riches; the more stitches ..."
Squirrel!
At first glance I was wondering why Microsoft would supply and API function CreateRemoteThreat().
Even for Windows, that would be a little out there.
...is to re-configure the UAC to make it as strict as Vista.
Hell, UAC is good. It's better than sudo. With sudo I will be tempted to use "sudo -s".
The most common scenario to meet an UAC dialog for me is when installing new apps or drivers. Other than that, you shouldn't really see an UAC dialog...
Most of the apps I came across have adopted to require no admin privileges. After all, it's the App fault to requires UAC in the first place for those doesn't really need admin privileges.
BTW, I think in Win 7, AFAIK all Microsoft signed EXE are exempted for UAC prompt by default. There isn't a whitelist but simply all MS signed binaries are exempted.
I agree in spirit, but the implementation is bad.
I once tried to write a "sudo for Cygwin" that would bring up the UAC confirmation box and run a program with associated elevated permissions in Vista. (Other people have written programs that they call "sudo for Vista," but none of them do what I want. In particular, they don't run programs in the same console.) In the process of poking through the security APIs, I learned a little about what a mess UAC is uder the hood.
Windows NT/XP has a perfectly good security model, if only people would use it. In some ways it's more sophisticated than Linux's: For instance, file permissions are more fine-grained on NT. The problem really hasn't been with XP/NT; it's been "social:" it was the culture of software development on Windows to too often require, unnecessarily, that users have administrative rights.
Microsoft's solution in Vista was to restrict the rights of administrators and add GUI confirmation boxes. This was the wrong solution, I think. In my (admittedly armchair-quarterback's) judgment, the right one would have been to,
1 - Keep traditional XP-style administrator and user accounts, with roughly the same privileges as they'd always had.
2 - Require OEMs to ship computers with user, rather than admin accounts, enabled. Randomly-generated default admin passwords should be written on a sticker on the front of the PC's case.
3 - Add a "sudo" mechanism, perhaps with the following modifications from 'nix sudo to make it easier for novices:
... a - The sudo prompt pops up automatically when a program attempts to do certain classes of things for which it does not have privileges. This differs from Linux, in which a program will simply fail with an "Insufficient permissions" error; this would be pretty opaque to novice users I think.
... b - "sudo" could be configured (and perhaps should be by default) so that it is sufficient to click a "confirm" button in lieu of typing in a password.
This is almost what UAC is. But the devil is in the details. What Microsoft actually did was make "Administrator" accounts into something more like "user" accounts, and add a level of privilege yet higher than administrator. But it feels tacked-on, and not really "at home" in the NT security model, which in fact provides plenty of control on its own over what rights different users and groups have, if only it were used correctly.
In other words, Microsoft shouldn't have restricted Admin accounts in this poorly-documented way; it should have instead added a sudo mechanism to make it more feasible to run as a User, and kept the nicely-documented and well-designed security model that NT has always had but people have simply never used.
In the original Vista release, this activity would cause an annoying back-to-back double elevation: once to create the folder, and again to rename it to its intended name. Service Pack 1 streamlined this a little, reducing it to only a single elevation, but Microsoft clearly wanted to get this down to zero.
NO! Bad monkey, no cookie! There is NO reason to allow ANYTHING to write to my /Program Files (or /Applications if you prefer) folder without my permission. None. Zero. I want a prompt. Yes, just one, but I want a prompt!
And that passes right into the hands of an almost unbelievable standard method in windows:
Unfortunately, the "Microsoft-signed application" restriction is easily bypassed using a standard Windows trick that allows one process to insert code into a second process, as long as both processes are being run by the same user. The limitations of the file management component are probably unavoidable (it can only do the things it has been programmed to do, after all), but it turns out it doesn't really matter. The file management component can place files into various locations on the system that an unelevated user cannot; an auto-elevate program can then be tricked into loading those files and executing code from them.
The result is, just as with the rundll32 problem, silent and automatic elevation, able to do anything.
WHY ON EARTH would you arbitrarily allow any random program a user is running to pass commands to a signed application that by its signature can walk right through locked doors?? I'll admit there probably are instances where you would like to pass commands (requests) to another app to handle something, you either (1) have to severely restrict the scope of the requests it will process, or don't sign it to give it rights to do whatever it pleases. This is like a mall security guard being given the keys to the maintenance halls, and the guard letting any joe public in that asks him. Either give him some common sense or take away his keys. A filemanager that has the power to do anything you ask it to, and will do so blindly and willingly, is just a jaw-dropper.
Sometimes the scope of Windows security stupidity astounds me. And yet they consistently keep finding ways to top themselves.
I work for the Department of Redundancy Department.
As best I can tell from what this guy is saying, there are some places (like, for example, deleting a file in the /System or /Library directory) where the Finder would prompt you for a password. As OS X matures, there are still some times where the Finder simply doesn't do it right - and simply refuses permission, when it should prompt you for permission. This happens less frequently in Leopard than it did in Tiger. There is nothing separate from the POSIX permissions in OS X, there is nothing like UAC that can be turned on and off. If you have permissions, you can do something, if you don't, you can't, or you are prompted for a password (the gui equivalent of 'sudo').
The secret to creativity is knowing how to hide your sources. - Albert Einstein
In the windows side, people rarely ask the question "Why do you need admin privilege?" Till the app developers learn to write code that lives comfortably in user space with user privilege, you will have problems.
The problem is not users blindly klicking UAC dialogs or MS's auto privilege elevation is not perfect. The problem is users not asking the question, "why the hell you want to be root?".
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
They should be doing this:
https://bugs.launchpad.net/ubuntu/+bug/156693
http://slashdot.org/comments.pl?sid=1152645&cid=27105713
Summary:
UAC is like getting users to solve the "halting problem", e.g. figure out whether the program will halt or not (aka screw up your PC or not) without having the program's source code, or knowing all the inputs. Google the "halting problem" to see how hard it is.
My suggestion is analogous to:
Program: "Hi, I'm a flash demo, I want 30 seconds of real time"
User: "Sounds reasonable. OK",
The O/S then runs the program, and if the program is still running 30 seconds later, the O/S kills it.
So no need to figure out whether it will halt or not. The program will halt - the O/S ensures it.
If the program says "Hi, I'm a flash demo, I want infinite time", it should be far easier to train the user to go: "No" or "Too bad, you only get two minutes to do your stuff, that's all I'm willing to give you".
AFAIK, Microsoft has lots of very very smart people working for them. I'm sure they have already figured out something far better than my idea, after spending 6 billion dollars and thousands of man-years on Vista.
So UAC is either institution incompetence, or malice (they just want to shift blame to the users, or they don't actually want increased security).