Last Year's CanSecWest Winner Repeats on Vista, Ubuntu Wins 337
DimitryGH followed up on the earlier news that the MacBook Air lost CanSecWest by noting that "Last year's winner of the CanSecWest hacking contest has won the Vista laptop in this year's competition. According to the sponsor TippingPoint's blog, Shane Macaulay used a new 0day exploit against Adobe Flash in order to secure his win. At the end of the day, the only laptop (of OS X, Vista, and Ubuntu) that remained unharmed was the one running Ubuntu. How's that for fueling religious platform wars?"
Re:Let me get this straight (Score:4, Informative)
Re:Let me get this straight (Score:4, Informative)
Something is Fishy (Score:5, Informative)
Flash, like all other plugins, run within the security context of the low-rights user used by protected mode. Even if the flash plugin had an obvious buffer overflow or other exploit, it would only be able to access the data accessible by that low rights user, NOT the user running IE. That's the point of protected mode.
For a flash plugin to allow for a hacker to access personal files of the user it would not only have to have a buffer overflow (or some other exploit) in flash itself, but also take advantage of a privledge elevation exploit in Windows simultaneously.
I didn't see them specify in the article what browser than were using. Since they said it was an issue with flash, and not Windows, they couldn't have been using IE. My guess is that it was Firefox, since they said they loaded "popular" 3rd party apps.
Futhermore, the file in question must have been accessible to the user running Firefox (or whatever non-IE browser) since that would also require a privledge elevation in Windows.
So I'm not really sure how you can blame this on Vista or even Microsoft. If they had been using IE, it wouldn't have happened, regardless of the flaws in Flash. This says absolutely nothing about Vista security. The exact same thing would happen on every other OS. If you have an app with an exploit, and that app is running as User A, the hacker using that exploit has the same rights as User A.
I suppose one could argue that various defensive techniques like ASLR [msdn.com] should have stopped this, but without knowing the details, that's impossible to say. A buffer overflow can just as easily be used to call APIs exposed by the exploited application as it can to call OS APIs, and since ASLR only applies to Windows APIs (indeed, many of these techniques only apply at the OS level), this wouldn't be a fair characterization either.
Indeed, I find it strange that they didn't mention mitigating factors. I realize they're trying to be responsible as far as reporting, but telling people that users running IE on Vista aren't affected isn't exactly giving anything away... aside from the fact that Vista did its job as best it could.
Not about the OS? (Score:0, Informative)
I don't know the details about the sploit so I don't know if it's OS specific even though it is Flash.
Re:Popcorn anyone? (Score:4, Informative)
The Monty Python joke goes along the lines of, "This lager is like making love in a canoe - it's fucking close to water"
Re:Hey! (Score:5, Informative)
Re:Newsworthy? (Score:4, Informative)
Re:Newsworthy? (Score:5, Informative)
First, this wasn't some script kiddie applying a known exploit. It was a new exploit that the winning team came up with. It isn't trivial to do.
Second, no, this "could have happened to any OS" is wrong. A well-crafted browser (in this case, the browser is part of the OS) can in theory prevent browser plugins from accessing anything of importance. However I don't think any existing browsers do that - but they should.
Second, and perhaps more important, the existence of 3rd party software on different OSes isn't the same. For example, most Windows users use Adobe Acrobat to view PDFs, whereas many Linux users use FOSS PDF viewers (Evince, KPDF). It might be the case - and I am guessing that it is - that Acrobat has far more exploits against it, both because it has far more code (what with all the functionality 99% of users don't need), and that it isn't open source. In general Windows users tend to have lots of 3rd party apps that are closed source and of dubious quality. That isn't the case on Linux.
Furthermore, even if two OSes run the same app - Flash, say - that doesn't mean they are equally vulnerable. Flash isn't identical between the platforms; if I am not mistaken on Linux Flash uses Alsa for sound (or some other Linux sound system). So if Alsa is more secure than Windows' sound system, that would be one difference.
I'm not saying this competition is a great test of OS security. It isn't; it's an anecdote. But it isn't worthless either. In fact the results are pretty much what I would have expected from the beginning: OS X is a great OS but security has never been a top priority (there wasn't as much of a need for it, so why bother). Windows has focused on security recently but is hobbled by having lots of closed-source 3rd party apps. Linux was always security-focused (starting as a server OS), and has the advantage of most of its software being FOSS and arriving from a repo under the control of the distro (in this case Ubuntu).
Re:Something is Fishy (Score:5, Informative)
Unless that file was specfically marked readable by the low rights user (which would be obvious cheating), or unless it was placed in a directory accessible by that user (temp directory, for instance), they could not have been using IE.
Re:Popcorn anyone? (Score:5, Informative)
The question isn't "Is Flash vulnerable?", but rather does a vulnerability at the application layer allow you to hack into the OS. It is entirely besides the point if Flash is flawed in the same way, thought there is a reasonable likelihood that it is not in this case. There are significant differences in code compiled for the various platforms. We Software Engineers call that "conditional compilation."
Re:Popcorn anyone? (Score:2, Informative)
Re:Something is Fishy (Score:5, Informative)
You are right that plugins by default runs under the special low-rights "ieuser" account. Unless the plugin uses tricks to circumvent this security for some reason.
And that is exactly what flash does. It uses a special "broker process" which runs as a daemon/service. The restricted plugin then talks to this brokerprocess and thus breaks out of the sandbox.
The flash API indeed has methods for creating/deleting/reading files and even executing applications (Would you believe that?). Although Adobe/Macromedia have tried to ensure that flash actionscripts can only use these in a "safe" way; I believe it is probable that the exploit was somehove connected to a vuln in the broker process; quite possibly in some of these API functions. Using a broker process to break out of the sandbox can circumvent any security precautions taken by the browser.
Given that Flash vulns are often cross-platform I think it is quite likely that this also is a problem on Linux. Now, if the special file which the contestants had to retrieve required *admin rights* the yet another level of security had been broken (UAC). But at this time we can't really determine.
Re:Something is Fishy (Score:5, Informative)
A broker service is used when reading or writing to user files (such as when they save a file to their desktop, or upload a document to a web site). This isolates the potentially dangerous code into a very small (~10k lines) application that is far easier to audit. This application runs as the normal user, and essentially accepts requests from the low rights IE process when actions need to be performed on user files.
Re:Something is Fishy (Score:3, Informative)
Re:Know this: people use linux on desktop (Score:4, Informative)
Re:Popcorn anyone? (Score:5, Informative)
Re:Something is Fishy (Score:4, Informative)
Re:Different hardware, different incentive? (Score:2, Informative)
Re:Popcorn anyone? (Score:2, Informative)
Re:Popcorn anyone? (Score:2, Informative)
Re:Popcorn anyone? (Score:5, Informative)
Well on Windows, sandboxing of permissions is different. There might still be the exploit but the level of vulnerability would most likely be higher on a Windows system as a result of IE running at a SYSTEM level permission rather than a USER level like in Mac or Linux. Change to a different browser like Firefox on Windows and you will be safer.
IE does not, and never has, run as SYSTEM. Prior to Vista it runs as the user who starts it. In Vista it runs with privileges lower than a regular user.
I realise Slashdot is as anti-Microsoft as they come, but it's still surprising to see the same FUD about IE still being spewed 10+ years after it was shown to be false.
Re:Different hardware, different incentive? (Score:3, Informative)
Re:Software sucks. (Score:3, Informative)
Re:Something is Fishy (Score:5, Informative)
I don't even know why Microsoft bothers trying to secure stuff when morons like Adobe just go and fuck it up.
Re:What kind of exploit? (Score:5, Informative)
day one were pure network attacks nobody got in on day one. day 2 was email and url based attacks. only the mac got won on day 2. on day 3 you could add non default but popular software from a list (couldn't find the list anywhere on the net, sigh) and adobe flash was vulnerable, so the vista machine got taken.
Ubuntu held up for all 3 days, but because only popular and default software could be added, this could bring a false sense of security. there are many ways to 'design' a supposedly open source software package on say, sourceforge.net but to have a compromised binary that was made with slightly altered source code... to get a trojan on a linux system. repositories tend to be fairly well monitored, but there have been times where applications that are trojans have gotten into widely used repositories. as far as i can tell, sourceforge has no real method for testing if software contains trojans or not, so it's purely up to the community that uses sourceforge to report bad software, etc. i imagine that freshmeat is the same, and many many linux users use sourceforge or freshmeat to find specific linux applications they need or want...
maybe there aren't enough linux users yet to make this a huge issue, but with Microsoft's brand image going south (kinda the way IBMs did in the 90s) linux is sure to be finding more and more people who would rather deal with OSS than with bill gates.
Re:Popcorn anyone? (Score:1, Informative)
Re:Popcorn anyone? (Score:5, Informative)
So, prior to Vista, when it ran as the user who starts it, given that over 90% of the cases the default user has complete and unlimited access to the system files, how is running as user different from running as SYSTEM? (And, yes, I pull that "90%" figure out of my arse---but I'll bet it's higher.)
Firstly, because SYSTEM and Administrator have different privilege levels.
Secondly, because there is a vast gulf of difference between the statements "IE runs as SYSTEM" and "IE runs as the user, which is sometimes Administrator, and I think that Administrator and SYSTEM are the same". One is a (serious) architectural problem, the other is an end-user configuration problem. Trying to say they are equivalent is at best ignorance and at worst lying.
Finally, while most home systems would certainly be running users as Administrator, most managed corporate systems would not. 90% is a ridiculous over-estimate of how many XP systems only have "Administrator" users.
Re:Something is Fishy (Score:5, Informative)
Not at all. What I'm suggesting is that when someone says that X is not possible because it isn't supposed to happen, it doesn't mean that it can't happen or won't happen. The Titanic was supposed to be unsinkable. AACS was supposed to be unbreakable. The four-minute mile was supposed to be unachievable.
I'm not foolish enough to claim that *nix cannot be rooted or cracked. Just that because of its design it is inherently more secure and more difficult to crack than a system that still allows apps to run in rootspace.
The baggage of supporting legacy apps that require(d) administrator access. Because Windows had been designed for so long to be run by a single user-administrator, there are plenty of apps that simply won't run without admin-level privileges.
Not exactly. When an OS is designed from the ground up as a multiuser system (such as *nix), it is very easy to restrict access to system resources. If I want to install a piece of software on Linux, for example, I cannot make the installation system-wide (by writing to /usr/bin, for example) without admin privileges. I cannot install libraries to /lib, /usr/lib, etc. I cannot write settings to /etc. Even when installed and executed, that program will only have a restricted set of rights based on the user/group that executes it. I can, however, compile and run executables as a user without needing admin access and without write access to system files and/or directories. I can put whatever libraries, modules, settings etc are required in my home directory without needing access to restricted areas.
Yes, I do run the risk of hosing my /home/user directory and everything inside of it, but I cannot touch any other user's files, and cannot touch system files.
Windows, on the other hand, has a hybrid model where a multi user model is tacked onto a single user-admin model, or rather support for a single user-admin model is bolted onto a basic multiuser model. Basic, because a true multi-user system would never have a single repository for all settings, like the Windows registry.
Please explain.
No. What I'm saying is that the my sysadmin's argument is very similar to the OP's argument. The OP said that because IE7 isn't supposed to allow a system level exploit via something like Flash, then therefore it isn't possible. My sysadmin said that because she configured Exchange to block autoforwarding to public webmail then it isn't possible. It is clearly possible to to autoforward my mail to gmail, and I did it and showed her to prove a point. She seems to think I manually forwarded the messages and somehow spoofed the reply-to field, and that autoforwarding is impossible because it shouldn't happen.
It's the same point I'm making now, and am running out of ways to say: Just because something shouldn't happen doesn't mean it won't or can't.
More on topic, if an app has elevated rights, then exploiting a vulnerability in that app will give the exploit/exploiter elevated rights. There are very few apps on *nix (none that I can think of) that run or need to run with elevated rights. There are a lot of apps on Windows that expect to have admin rights, regardless of whether or not such access is needed. This is why the problem is structural, and why I used the example of the incomplete wall.
Re:Popcorn anyone? (Score:4, Informative)
And this only works if the server in question is in your Trusted Sites or Intranet Sites in IE.
Non-admin users installing printer drivers is something that is controllable via AD and Group Policy. If you set it, it loosens up acls and privs in a very specific and limited part of the system that lets non-admin users install printers.
This isnt rocket science or magic.
1. The server who hosts this printer driver is in your IE's Trusted Locations or Intranet Locations.
2. The configuration to let non-admin users install printer drivers is set on your machine.
3. There is a hole big enough within the security loosening from #2 to do anything interesting with to own the OS.
Re:What did you expect? (Score:5, Informative)
1. The contest did not require someone to 'own' the box to win. They just had to read the contents of some specific file somewhere in the OS. Unfortunately, they didnt publish where that file was, or what the file-system ACLs on it were.
2. The guy who took down the Vista box claimed in the article that it would only take them a few more hours of work to make the Flash vuln effective on OSX and Linux as well.
Re:Popcorn anyone? (Score:4, Informative)
I'll even be nice and give you some of the information.
There are special cache locations in the registry and user profile called 'Low' that are the only places readable/writeable by IE7 in protected mode.
I did mis-speak in one sense in my post
Re:The ONLY reason ubuntu won (Score:3, Informative)
Re:Popcorn anyone? (Score:3, Informative)
Permitting a program to connect to the X server is a pretty big statement of trust, since it has to have at least the same level of permissions the window manager does. So it's fairly carefully controlled. There are ways of making su work, which hail from back when you used telnet to do remote login and your GUI apps connected directly back to a public TCP/IP port on your terminal to get at the X server, but they're obsolete. These days, the fastest way to do what you want is to substitute ssh for su.