Forgot your password?
typodupeerror
Bug Microsoft Security Windows Technology

Windows DLL Vulnerability Exploit In the Wild 178

Posted by Soulskill
from the is-it-tuesday-again-already dept.
WrongSizeGlass writes "Exploit code for the DLL loading issue that reportedly affects hundreds of Windows applications made its appearance on Monday. HD Moore, the creator of the Metasploit open-source hacking toolkit, released the exploit code along with an auditing tool that records which applications are vulnerable. 'Once it makes it into Metasploit, it doesn't take much more to execute an attack,' said Andrew Storms, director of security operations for nCircle Security. 'The hard part has already been done for [hackers].'"
This discussion has been archived. No new comments can be posted.

Windows DLL Vulnerability Exploit In the Wild

Comments Filter:
  • This is actually faulty programming in applications [computerworld.com], not Windows. Kind of like buffer overflows. It's what happens when you don't know what you're doing nor are you following secure coding standards.

    Because application developers, not Windows, are to blame, Microsoft can't patch the operating system without crippling an unknown number of programs that run on the platform.

    There are no reports of any Microsoft or default Windows applications containing the bug, so unless you have a specific third party app you're not vulnerable. Also, there is already a tool available from Microsoft [microsoft.com] you can use to block it from all applications, but some of the apps might obviously break.

    To protec

    • by eldavojohn (898314) * <eldavojohn@@@gmail...com> on Tuesday August 24, 2010 @08:16AM (#33353380) Journal

      There are no reports of any Microsoft or default Windows applications containing the bug

      Really? That's odd, from the original blog posting [rapid7.com]:

      At least four of Microsoft’s own applications have been confirmed as exploitable through this vector, two of which were already being addressed by the time I contacted them.

      • by Xacid (560407) on Tuesday August 24, 2010 @09:06AM (#33353814) Journal
        The link appears to be slashdotted, but my guess is that the emphasis from the parent is on "default". Microsoft offers more than just a standard vanilla OS install and applications.
        • by Zalbik (308903)

          The link appears to be slashdotted, but my guess is that the emphasis from the parent is on "default". Microsoft offers more than just a standard vanilla OS install and applications.

          Yes, but the original claim was:

          There are no reports of any Microsoft or default Windows applications containing the bug

          This claim is obviously false as there are Microsoft applications containing the bug. What we can say is that issue is not the fault of the Microsoft OS developers.

        • by jrumney (197329)
          Are you sure the emphasis is on the word default, and not the word or that comes before it?
      • by klui (457783)
        Just run regedit and search for all instances of rundll32.exe. You will find there are quite a few entries that do not fully qualify the dlls they want to access. Some entries even have rundll32.exe not fully-qualified.
    • Re: (Score:1, Interesting)

      by g0bshiTe (596213)

      Many Windows applications don't call code libraries -- dubbed "dynamic-link library," or "DLL" -- using the full pathname, but instead use only the filename, giving hackers wiggle room. Criminals can exploit that by tricking the application into loading a malicious file with the same name as the required DLL. The result: Hackers can hijack the PC and plant malware on the machine.

      This is far from a fence post bug, or a buffer overflow. This is more like either DLL injection, or like the paragraph alludes to

      • by Zocalo (252965) on Tuesday August 24, 2010 @08:38AM (#33353548) Homepage

        A simple fix would be for a programmer to have their app at initialization checksum any dll it uses.

        Bad idea. That would likely create more problems than it solves and bring back the worst of DLL hell, especially for frequently updated and used DLLs and also given how badly certain vendor's individual development teams seem to communicate with each other. Say App_A installs v1.0.1 of a DLL in a shared location, then later App_B then comes along and updates this to v1.0.2 - congratulations; you just broke App_A. OK, there's a fix for that, but only if you can call the awful kludge that is WinSxS a "fix".

        • Or you place v1.0.1 of the DLL in the same folder as App_A.exe. App_A will find v1.0.1 of the DLL before going to the shared location *IF* it's written properly. In "MSI packaging Land" this is called "Application Isolation".

          Application Developers are rarely good packagers.

          • by John Betonschaar (178617) on Tuesday August 24, 2010 @10:07AM (#33354620)

            Or you place v1.0.1 of the DLL in the same folder as App_A.exe. App_A will find v1.0.1 of the DLL before going to the shared location *IF* it's written properly

            What's the point of having shared libraries if only the application itself can load them, and can only load single, checksummed version of it?

            In "MSI packaging Land" this is called "Application Isolation".

            In the rest of the world we call it 'static linking' :-P

          • by Zocalo (252965)
            That's why I specified "in a shared location". :)

            I don't actually think there is an easy, one-size-fits-all solution to this problem without a radical shakeup of how Windows handles DLLs. If you insist on applications each installing their own versions of each DLL then you end up with a potential nightmare when there is a flaw found in some versions of a given DLL like with atl.dll a while back. At least you'd know which apps are vulnerable, but that's not going to be much help when one of those is ess
        • by Ralish (775196)

          OK, there's a fix for that, but only if you can call the awful kludge that is WinSxS a "fix".

          I always thought that WinSxS was quite an elegant fix to a difficult problem. Put it this way, I still have nightmares about DLL Hell from the bad old days, but have yet to encounter a problem due to WinSxS. The closest I've come is one or two applications making assumptions about dependencies (i.e. not bundling the required installers and not failing gracefully). Have you had issues with WinSxS?

          • by 0123456 (636235)

            I always thought that WinSxS was quite an elegant fix to a difficult problem.

            Weird, I always thought it was a horrible kludge.

            Put it this way, I still have nightmares about DLL Hell from the bad old days, but have yet to encounter a problem due to WinSxS.

            Either you're lucky or I'm unlucky, because my old XP PC has a serious case of 'SxS Hell' that I've been totally unable to fix. It's a long time since I've booted it, but I remember spending hours poking around in the SxS directories trying to figure out what the hell Windows had screwed up there to prevent some applications from running. It also refuses to install the service pack for the .Net Framework, apparently for the same reason.

            I'm so glad I run Linux

        • by g0bshiTe (596213)
          I never even considered updated DLL's.
        • by master_p (608214)

          And all this trouble because Microsoft denied doing the sane thing, i.e. name its DLLs with a version number (as well as the DLL name).

        • by adisakp (705706)
          Also for what it's worth, the whole idea of a DLL is that it can be updated independently from an app and other DLL's. Keeping a DLL checksum in the app to bind to only a specific instance of that DLL defeats the purpose. At that point, you might as well be just statically link the code to your app.
      • A simple fix would be for a programmer to keep data separate from executable code.

        FTFY.

      • by pelrun (25021) on Tuesday August 24, 2010 @08:49AM (#33353652)

        Yeah, and what happens when that DLL gets updated due to a different vulnerability, but the app doesn't? You either get a broken app or one using an insecure library *anyway*.

      • Re: (Score:3, Insightful)

        by arth1 (260657)

        A simple fix would be for a programmer to have their app at initialization checksum any dll it uses.

        That would defeat much of the purpose of using DLLs.
        Not to mention what would happen when Microsoft updates a DLL, or the user runs a rebase.

        To top it off, it's more complex than the real fix too.

      • Many Windows applications don't call code libraries -- dubbed "dynamic-link library," or "DLL" -- using the full pathname, but instead use only the filename, giving hackers wiggle room. Criminals can exploit that by tricking the application into loading a malicious file with the same name as the required DLL. The result: Hackers can hijack the PC and plant malware on the machine.

        Another Result: you can load applications onto network shares and you won't have to load .DLLs on every single machine. This "fea

    • Re: (Score:3, Interesting)

      by UnknowingFool (672806)

      How much fault is really debatable. Yes if an application is coded in a such a way that the exploit exists that's partly on the fault of the programmer; however, why should that translate into an exploit on the OS? Also from what it appears, many applications have this problem including some from MS so it does not appear some obscure programming quirk.

      According to reports that first appeared last week, developers, including Microsoft's, have misused a crucial function of Windows, leaving a large number of Windows programs vulnerable to attack because of the way they load components.

    • Re: (Score:3, Interesting)

      by dbIII (701233)

      To protect from stupid developers you would probably need something like selinux for Windows

      Considering the failure of antivirus to protect the first victims of any new virus it looks like that may have to be the way to keep the platform viable. A list of what is allowed provides far better protection than an ever changing list of what is not allowed.

      • by 0123456 (636235)

        A list of what is allowed provides far better protection than an ever changing list of what is not allowed.

        The problem is that vast numbers of Windows programs rely on being able to do things that shouldn't be allowed, and people only buy Windows because it will run all their old Windows programs. So Microsoft are screwed either way.

        Maybe it could be done in 20 years when most people have replaced most software with new versions, or by virtualising all old applications, but it would not be an easy job.

    • by MojoRilla (591502) on Tuesday August 24, 2010 @09:08AM (#33353836)
      Microsoft created a liberal dynamic library search path that allows [microsoft.com] (or even encourages) applications to not fully specify DLL locations. Now, after the fact, they publish this security [microsoft.com] statement saying not to use the dynamic library searching they documented previously. It is of course Microsoft's fault. They didn't consider security at all when loading DLLs, and now they are blaming applications that implemented the documented specification.

      The bottom line is that Windows was never designed to be secure, it was designed to have the most functionality, and trying to patch every hole now is almost impossible. Generally, when code reaches this level of complexity and brittleness, it is often the best course to start all over.
      • by mdwh2 (535323) on Tuesday August 24, 2010 @09:24AM (#33354018) Journal

        I agree - it's unclear to me what the "fault" of the developer is here, and which applications are at fault. I thought that loading a DLL by name without a specific path was standard practice? And how does it work with linking - in my experience, all applications I've written and used can either use a DLL in the standard path, or be overridden by a local DLL, so surely that's standard practice too? And wouldn't this affect almost every Windows program that uses DLL?

        But then, I'm not sure that this is a bad system anyway. Well, if it's possible to include a DLL loaded off a web page as being the standard path, that seems a gaping hole. However, if this flaw requires an attacker to already install a dodgy DLL in the user's path on their system, surely that would already be the security flaw? I mean, it's a bit like saying "It's a flaw that people can run exes by double clicking, there could be malicious code" - the flaw isn't in running exes, the flaw is how they got there in the first place.

        What is the proposed fix for applications that link to DLLs? And how do other operating systems work - again, I thought that having a path system allowing multiple possible locations for shared libraries wasn't uncommon?

      • by Anne Honime (828246) on Tuesday August 24, 2010 @09:38AM (#33354156)

        MOD PARENT UP !! Fact is, on any unix out there, no competent admin would leave '.' neither in executable path, nor in dynamic library search path. It's another of case of a security hole known at least theoretically since the 60's, and observed in real life in the 80's, that microsoft overlooked in the design stage when it was time to follow proper security assessments, and are now stuck with.

        They should be put on trial for dumb blunders like this one. When you hire top professionals who can't ignore the 'state of art' when doing an error like this, it should be considered a cause for limitless civil liability.

        • Fact is, on any unix out there, no competent admin would leave '.' neither in executable path, nor in dynamic library search path.

          I've yet to see any Unix which has . in either PATH or LD_PATH by default.

          microsoft overlooked in the design stage

          It's not overlooked. It's just one of those compatibility things that date back several decades.

          However, the corollary is that any Windows developer who is clueless of DLL search path in Win32 (and the implications) should really know better by now.

      • by mcgrew (92797) *

        Generally, when code reaches this level of complexity and brittleness, it is often the best course to start all over.

        Or in the immortal words of Montgomery Scott (Star Trek III), "The more you overengineer the plumbing, the easier it is to stop up the drain."

        • Or in the immortal words of Montgomery Scott (Star Trek III), "The more you overengineer the plumbing, the easier it is to stop up the drain."

          When he's in(or talking) engineering, he's "Scotty," you heathen.

      • by CKW (409971)

        > The bottom line is that Windows was never designed to be secure

        Oooh, that makes me wonder. Can Linux/Solaris/Unix be attacked *simply* using the PATH environment variable? Forget limiting the attack to shared objects, anything that is loaded or exec'd by any other binary/scripts.

        The complicated thing is many applications build their own PATHs, and you're looking for a directory on that path that is writable to whatever user you are, one that you can put a file that doesn't yet appear that high up on

        • You'd need to already own the computer to pull something like that, which pretty much defeats the purpose. But, yes, it's been used in the past, along with other techniques. The intrinsic problem here is not to have a search path for execs and libs, but to provide hardcoded at system level a default path with $PWD included, instead of a path exclusively containing system directories under admin control.
      • by Ralish (775196)

        Microsoft created a liberal dynamic library search path that allows (or even encourages) applications to not fully specify DLL locations. Now, after the fact, they publish this security statement saying not to use the dynamic library searching they documented previously.

        So basically, your suggestion is to design an OS that ensures that it is secure by taking away API calls that could be misused in a way that compromises security? By your own admission, it is a documented specification, and it is behaving exactly as it is intended to do so. It isn't a "bug" in the API, it's misuse by various developers. However, Microsoft is at fault for how developers (its own or 3rd-party) misuse an API call that is fully documented and behaving exactly as intended? This makes absolute, p

      • by master_p (608214)

        There is a way out for Microsoft: they can "virtualize" the O/S so as that apps think it's like Windows 95, where they can own all files, but in reality the apps would only do damage to their own version of the files.

        • That's actually by and large what Vista and 7 do when an application without a manifest that declares Vista+ support tries to write into some known system folder that's not writable (e.g. its own install folder under "Program Files"). It silently creates a copy (if file already exists globally) to a user-local, app-specific folder, and then redirects all file operations to that copy.

      • What do you expect from an OS that has a default color scheme that's very much like the livery of a clown car?
    • by NiteShaed (315799) on Tuesday August 24, 2010 @10:42AM (#33355146)

      There are no reports of any Microsoft or default Windows applications containing the bug, so unless you have a specific third party app you're not vulnerable.

      Ummmmmm; "According to Moore, at least one Microsoft executable -- "explorer.exe," the Windows shell -- includes the flaw. [computerworld.com]"
      I'm pretty sure your Windows machine has explorer.exe loaded by default.

    • Because application developers, not Windows, are to blame

      Because clean design of many Windows APIs inspires secure coding practices.

      And MS has totally nothing to do with the fact that some of the APIs are so developer-friendly.

      [/sarcasm]

  • >'The hard part has already been done for [script kiddies].'

    Here, I fixed it for ya. No self-respecting coder would use a library like that.

    • Re:Not quite (Score:5, Insightful)

      by Lazareth (1756336) on Tuesday August 24, 2010 @08:32AM (#33353504)

      Yeah, because all self-respecting coders has this driving urge to reinvent the wheel whenever they're met with an already functional and documented piece of code.

      • Microsoft issues a patch. This conversation follows:

        Customer: Hey. This app stopped working. What gives?
        Support: Yeah, Microsoft issued a patch which broke it. It's Microsoft's fault that they have a buggy OS.
        C: Hey Microsoft, what gives with this patch breaking my app?
        Microsoft: The code used by that app was vulnerable to exploitation. We fixed it. You developer used code which was exploitable and will need to fix it.
        C: Hey, Microsoft say they made my PC more secure with this patch. Apparently your rubbis
        • You knowingly sold me software which had a vulnerability which could result in serious security issues?

          S: A patch is on the way. Please don't sue.

          Meanwhile, in the real world...

          • Re: (Score:3, Insightful)

            by L4t3r4lu5 (1216702)
            Idiots continue to put up with the "release now, fix later" model of software development? Continue to have sub-standard produce foisted upon them because an application isn't a product you can hold, and therefore has gotten around merchantability laws on a technicality?

            Yup, the general population repeatedly bending over to big corp budgets sounds like real life to me.
            • by StikyPad (445176)

              If all software had to hold up to merchantability rules, it would cost orders of magnitude more than top-of-the-line hardware and flaws would *still* exist. Hospitals, banks, governments, etc. pay a minimum of hundreds of thousands of dollars for this sort of software, usually for very specific hardware that's been explicitly and thoroughly tested, and the payback is marginal at best. In other words, be careful what the fuck you wish for.

      • I'd like to introduce you to Dan Bernstein [cr.yp.to], creator of QMail and DJBDNS.

        Certainly, he could've accepted that BIND, Sendmail and Postfix were functional, but what fun would that be?

        Even if Dan isn't self-respecting, I'm sure there are many (myself included) that respect him.

      • Re: (Score:3, Insightful)

        by Ecuador (740021)

        Yeah, because all self-respecting coders has this driving urge to reinvent the wheel whenever they're met with an already functional and documented piece of code.

        Really? Phewww! I thought I was among few paranoid enough to not trust anything I haven't implemented myself from scratch. Good thing to hear all self-respecting coders work like that and I am not a minority! I mean, it was kind of obvious, how can you trust a library or sample code some unknown guy wrote when you yourself, the master developer, can do the optimal implementation yourself.

        • by pclminion (145572)

          I mean, it was kind of obvious, how can you trust a library or sample code some unknown guy wrote when you yourself, the master developer, can do the optimal implementation yourself.

          Translation: "I am a better program than any other programmer on Earth." Yeah, right.

    • Re:Not quite (Score:4, Insightful)

      by jeffasselin (566598) <cormacolinde@gma ... Nom minus distro> on Tuesday August 24, 2010 @08:46AM (#33353614) Journal

      This strange belief that doing things "the hard way" is in some unfathomable way "better" has always been interesting to me.

      A self-respecting coder is a strange beast indeed if it acts the way you describe it doing. A competent coder would just use any tool it has access to (including metasploit) in order to achieve its goals. A nice, competent coder would use this tool or any other to check the applications he uses or writes for security holes. Ignoring this tool because it's "too easy" is stupid.

      And regarding those "self-respecting coders", I don't think they intersect much with the kind of malicious hackers who would be willing to use this exploit for nefarious reasons.

      • Re:Not quite (Score:4, Insightful)

        by sp3d2orbit (81173) on Tuesday August 24, 2010 @10:36AM (#33355026)

        Its not the rewriting that is important. It is the deconstruction and reconstruction from scratch that is important. If you pour over a piece of someone else's code, document it, talk about it, sleep with it, whatever, you still will not be as intimately familiar with the code as someone who wrote it.

        There is absolutely nothing wrong with starting with someone else's well documented piece of code, reverse engineering it, and implementing a "inspired" version of your own. It goes a long way to understand problems you would never get a chance to understand.

        Think about it this way, all sorting algorithms have been written a million times. Still, students have to struggle through implementations every semester so they learn. This is the same thing. Its not wasting time, it is improving oneself.

        • by StikyPad (445176)

          If you pour over a piece of someone else's code, document it, talk about it, sleep with it, whatever, you still will not be as intimately familiar with the code as someone who wrote it.

          The fact that I can't remember what I wrote last year, let alone how I wrote it, is incontrovertible evidence to the contrary.

    • by TheLink (130905)
      > Here, I fixed it for ya. No self-respecting coder would use a library like that.

      Why? Are there bad bugs or security exploits in that library?
  • Releasing these exploits mostly helps who is best equipped to use them. Malware-biz and Microsoft. Should just write some stuff to share all data files in p2p networks and let it run.
    • by mcgrew (92797) *

      TFS says there's a detection tool; run the tool and uninstall any apps that have the vuln and you're safe(er).

  • The exploit requires user action. So the exploit isnt going to be as bad as it could have been.
    • by 0123456 (636235)

      The exploit requires user action. So the exploit isnt going to be as bad as it could have been.

      Until some virus starts writing compromised .DLL files to every network share in the company.

  • Huh? (Score:4, Interesting)

    by FranTaylor (164577) on Tuesday August 24, 2010 @09:02AM (#33353768)

    "but considering how much pain in the ass it is on Linux too, it wouldn't really work for all the casual people."

    I have Fedora 12 on my desktop with SELinux enabled. I didn't have to do ANYTHING AT ALL. I haven't seen an un-intentional alert in months. I was worried so I set one off myself just to make sure SELinux is still working, and yes it caught it.

    • Have you updated your kernel lately?
    • by Stray7Xi (698337)

      I have Fedora 12 on my desktop with SELinux enabled. I didn't have to do ANYTHING AT ALL. I haven't seen an un-intentional alert in months.

      Fedora uses targeted policy by default, meaning there's confined and unconfined processes. Services/daemons are confined, User processes are unconfined (full permissions). Why? Because it's too freaking hard to set selinux contexts for user processes.

      So your web browser, email, and bittorrent all are unaffected by selinux in most distros. I wouldn't recommend trying to setup contexts for them either, selinux is a nightmare. Apparmor at least you can get sane profiles for user applications.

  • Here's what Microsoft recommends:
    "Wherever possible, specify a fully qualified path when using the LoadLibrary, LoadLibraryEx, CreateProcess, or ShellExecute functions. "
    "Consider removing the current directory from the DLL search path by calling"

    In other words, they want programmers to use LoadLibrary("C:\program Files\my software\somedll.dll") instead of LoadLibrary("somedll.dll"). This is very counter-intuitive, as if you were app developer, you would want all of your DLLs be distributed with binary, and

    • In other words, they want programmers to use LoadLibrary("C:\program Files\my software\somedll.dll") instead of LoadLibrary("somedll.dll"). This is very counter-intuitive, as if you were app developer, you would want all of your DLLs be distributed with binary, and reside in same directory. Take a look in your program files directory, and almost every app does it that way...

      Um, that’s why they have the %programfiles% environment variable, and it’s why you install applications there, and it’s why the current directory when you launch a file (%userprofile%\Default\Documents\) should never be where you’re getting executable content (such as a .dll file).

      When you launch a file via its shell extension, the “current” directory and the directory where the executable is located which opens that file type are not the same... and you shouldn’t b

    • by 0ld_d0g (923931) on Tuesday August 24, 2010 @09:57AM (#33354480)

      Well, fully qualified doesn't mean static. You could compute the fully qualified name at runtime to pass to the LoadLibrary call. Or you could just stick a SetDllDirectory call somewhere in your app startup and keep the rest of the code the same.

  • in an "open" LAN environment: an exploited machine sets up a share, emails links to others in the contact list, remote exploit ensues. But who's allowing egress (outbound) SMB, WebDAV (at least not to a whitelist of remote hosts) on their network? Putting Windows Firewall up on all the workstations to drop ingress SMB traffic (with a few host exceptions for those pushing out updates via SMB) would be a smart thing to do as there's really no reason for workstations to be sharing files in a network with fil
  • by js3 (319268) on Tuesday August 24, 2010 @09:58AM (#33354486)

    If the user's machine is compromised to the point where unauthorized dlls are replacing valid dlls that's not my problem as a software developer. The only validity to this bug is that windows allows dlls to be loaded from remote network locations (isn't this sort of stupid in the first place?).

    I think the severity of this bug is blown out of proportion. The only idiots to blame is the idiot who did not secure his computer.

    • by ledow (319597)

      You map a network drive on your LAN, you execute a program on it that requires a DLL. That DLL can't be found unless Windows looks on the remote drive too, and chances are that if a program is bundled with a DLL, that DLL is actually a better match for the program than anything lurking in the Windows folder or other paths. So there is a use-case here - I'm pretty sure there's at least one school system I've worked on that would just break in a second if you couldn't load DLL's from a remote network locati

    • by arth1 (260657) on Tuesday August 24, 2010 @10:25AM (#33354880) Homepage Journal

      If the user's machine is compromised to the point where unauthorized dlls are replacing valid dlls that's not my problem as a software developer. The only validity to this bug is that windows allows dlls to be loaded from remote network locations (isn't this sort of stupid in the first place?).

      It's not replacing. It's you, as a developer, saying "try to load foo.dll", without specifying where "foo.dll" is to be found, but relies on the OS to find it for you. It traverses a list of possible locations, and as a last resort tries the current working directory.
      The problem is all yours if you don't specify where the OS is to look for the DLL. If you want to load DLLs from %installdir%\dlls, then all you have to do is specify that path. It's not rocket science.

      And no, allowing DLLs to be loaded from remote locations isn't stupid, as it allows for shared installations, which both saves boatloads of disk space, and allows for updating a single place instead of on each machine.
      Not considering that possibility is what's stupid.
      And not understanding how the DLL loading and file systems work on the OS you program for is even worse.

      Finally, the exploit doesn't depend on a remote share either -- that's Smallsquishy's damage control PR department working overtime. If you download a zip with hundreds of MP3 or picture files, extract it, and double-click one of them, the DLL that was in the archive will get loaded if your default player/viewer queries for that DLL without specifying a path.

      It's not a new exploit either -- it's been around for a long time on Windows. And before that, there were vulnerable Linux systems with "." in LD_LIBRARY_PATH, which basically amounts to the same thing.

  • How this works (Score:5, Informative)

    by Elbows (208758) on Tuesday August 24, 2010 @10:47AM (#33355212)

    I took me a while to figure out how this exploit works, but I think it goes like this:

    I have an application, foo.exe, that can make use of an optional system component (or 3rd-party DLL), bar.dll. I don't ship that DLL, and I can't guarantee that it will be present on every user's system. So to ensure that my program degrades gracefully, I open it with LoadLibrary("bar.dll"), and if it's not found I disable the features that depend on it. Since it's not my DLL, I can't speculate on where it's installed, so I use an unqualified path and let the loader do the searching (this is, after all, the job of the loader). The ensures that, as long as bar.dll is correctly installed on the system, my application will find and use it.

    From an application developer's point of view, this the right way to do things. If I did this on Linux or MacOS, it wouldn't be a problem. Unfortunately, Microsoft decided that the current directory (".") should be in the default search path (see http://msdn.microsoft.com/en-us/library/ms682586(VS.85).aspx [microsoft.com] ). It's even searched before $PATH!

    Now the exploit goes like this:
    1. On \\evilserver\evilsmbshare, I place a file foofile.foo, an extension which is associated with foo.exe. Right next to it, I create an evil version of bar.dll.
    2. I convince the user to double-click on foofile.foo, causing windows to open foo.exe, with a current directory of \\evilserver\evilsmbshare.
    3. If the user's system doesn't have bar.dll installed, Windows will eventually find my evil version of it at .\bar.dll and load it into the unsuspecting foo.exe.
    4. My evil code runs and does whatever evil deeds I want it to.

    If this is correct, then the decision my Microsoft to put the current directory in the library search path seems pretty braindead, and it's hard to blame application developers for assuming that LoadLibrary() will load a library in a sane and secure way. But I'm having a hard time imagining an application that would break if the current directory were just removed from the search path. Shipping DLLs in the application directory is common practice, but expecting them in the current directory? Why would you do that?

    It seems that this exploit requires you to trick the user into opening a file from a filesystem you have access to, at which point you could probably just as easily get them to open a trojan directly. I think local privilege-escalation attacks are more probable (e.g. tricking a system service into opening your evil DLL).

    • by master_p (608214)

      Thank you, please mod parent up.

    • by 0123456 (636235)

      But I'm having a hard time imagining an application that would break if the current directory were just removed from the search path. Shipping DLLs in the application directory is common practice, but expecting them in the current directory? Why would you do that?

      I've used a number of programs which would fail to run if you didn't start them from their install directory; I don't know whether they're looking for DLLs or data files, but I can be pretty sure that at least some programmers have relied on this behaviour without even realising... 'yeah, but we always run from our install directory, right?'

    • Re: (Score:3, Interesting)

      by cdrguru (88047)

      You are assuming the "current directory" is set to the location where an associated file is located. That isn't true. When an application is invoked by an association the current directory is usually the location of the program executable itself, which is passed the fully qualified name of the file that was clicked and caused the invocation of the program.

      So placing a DLL in the same folder as the associated file doesn't do anything. You have to put it in the same folder as the executable, which is (as o

      • Re:How this works (Score:5, Informative)

        by Elbows (208758) on Tuesday August 24, 2010 @12:52PM (#33357242)

        I just tried this out. When I launch a test program by double-clicking an associated file, the current directory (as returned by GetCurrentDirectory()) is set to the directory where the file was located. It ignores the location of the .exe, and it also ignores the "Start In" directory from the shortcut file (if the association was to a shortcut and not directly to an exe). This is on Win7 64-bit. So I think my evilsmbshare example from above would work as described. Of course it's possible that other Windows systems exhibit completely different behavior. :)

        I agree that it's still hard to exploit, but not quite as hard as requiring access to the user's local filesystem.

      • Re: (Score:3, Informative)

        So placing a DLL in the same folder as the associated file doesn't do anything. You have to put it in the same folder as the executable, which is (as of Vista and Windows 7) write-protected.

        With XP it is a lot easier because the Program Files directory structure is not protected and it was common to have applications writing stuff there, so you couldn't protect it. As of Vista the rules changed and you can't write there anymore.

        Actually it's write-protected in XP as well, as long as you don't use the FAT32 filesystem (so per-user file permissions actually exist) and as long as you're not running as admin (which nearly everyone did on their home PCs, but in organisations most people aren't). Vista basically changed the system from "local admin user = root" to "local admin user = sudoer".

    • by Lost Race (681080)

      It seems that this exploit requires you to trick the user into opening a file from a filesystem you have access to, at which point you could probably just as easily get them to open a trojan directly.

      Infected removable media would be the most likely vector. You plug in a USB flash drive or load a CDROM, browse to a file you want to open but don't notice the malicious DLL in the same directory, perhaps because it's hidden among hundreds of other files. We're used to being cautious about opening files on remo

  • Has anybody written a test to verify that Microsoft's fix has been properly applied? It would be a simple DLL with that pops up a message, and simple EXE that loads the DLL (which has new unique name). Or even two versions of the DLL, one with a good message and one with a bad message. One goes in the system path, one goes in the same path as the EXE, a temp folder.

    The MS kb patch has one typo, you add a new DWORD value to the registry, not a new key.
    http://support.microsoft.com/kb/2264107 [microsoft.com]

    MS fixed the other

    • Re: (Score:3, Informative)

      by colfer (619105)

      Or, correction, the good DLL would have to go into a folder that is in the PATH and not in any of the higher priority system folders. And you would have to register a file handler and a new type... since the directory of the EXE has first priority. Oh well.

      The priority list goes:

      1. The directory from which the application loaded
      2. The system directory
      3. The 16-bit system directory
      4. The Windows directory
      5. The current working directory (CWD)
      6. The directories that are listed in the PATH environment variable

The meat is rotten, but the booze is holding out. Computer translation of "The spirit is willing, but the flesh is weak."

Working...