5 ms·
False detection on file tcpip.sys
- vezzy-fnord 13y agoFor other similar gaffes, see: http://attrition.org/errata/sec-co/ http://attrition.org/errata/sec-co/ It's amazing what kind of things slip by.
- Scaevolus 13y agoIt's odd that they don't have a database of "known good" files taken from a clean install of each OS they support that they test against before releasing an update.
- jevinskie 13y agoYeah I'm wondering how something like this got through testing. Presumably they run a full suite of tests against every database update, right?
- xorgar831 13y agoProbably because the risk low, and it would require a good deal of work to maintain such a database. You'd have to keep it updated after every patch, a system may get a different version of the patch based on what is installed on the OS etc.
- acqq 13y agoNot much more work than any decent administrator has to do if he controls the updates to his own network. Note that even if you don't have in the database "all files ever" every entry lessens your chances of false positives. Actually I can't imagine that they can do their business while not having such databases.
- Scaevolus 13y agoHaving a version for every patch would be difficult, but the base level of {XP,Vista,7,8}{32-bit,64-bit}{RTM,SP1,SP2,...} should safeguard the majority of your users.
- acqq 13y agoAnd it's almost trivial to add new entries after each update. I can't imagine they don't have it, but I can imagine that they didn't use it this time ("check takes too much time, let's just ship it").
- greenyoda 13y agoI'd guess that the majority of users have automatic Windows Update turned on, since that's the default setting and what Microsoft recommends. The only people who have the base level of Microsoft operating systems (or don't apply patches between service packs) are the ones who are so clueless about security that they're not likely to have gone out of their way to install a third-party anti-virus tool. If they assume that most users apply all "critical" Windows updates in the order that they're pushed, the anti-virus vendor could snapshot their reference PCs before each update and record the hashes of all Windows files. It's possible to determine which updates have already been installed on a machine (there's an option for this in the control panel, and the information is probably stored in the registry).
- acdha 13y agoThis has happened multiple times to multiple vendors – so it's not clear that the risk is actually low and the resulting database is also useful for catching malware which attempts to replace system files. As for different versions, it doesn't seem like this should be a significant challenge to an organization which already has to maintain a test OS farm. The process of updating after a release is mostly automatic and if this acted solely as a guard against false positives, the consequences of missing a patch permutation are the current status quo.
- timsally 13y agoOuch. To make things even worse, the executable 'kaspersky_tcpip_fix.exe' to fix the problem is (1) not delivered over HTTPS and (2) not signed by Kaspersky (it is unsigned). So not only are you unable to receive an authenticated update to your antivirus database to fix the problem because your network is down, but you have no assurances that the offline tool provided to you to fix the problem actually came from Kaspersky. For those that are curious, you can use a tool distributed by Microsoft called Sigcheck to display signature information (http://technet.microsoft.com/en-us/sysinternals/bb897441.aspx http://technet.microsoft.com/en-us/sysinternals/bb897441.asp...).
- acqq 13y agoEven without any specific tool, you can observe the "Properties" of the binaries from Explorer and see the signature of the files that are signed. And when the fix isn't signed, of course, nothing can be checked.
- arnnr 13y agoActually windows does these checks automatically and replaces system files with signs of corruption/bad cert. You can force this check by using System File Checker tool in cmd: sfc /scannow
- pudquick 13y agoStill not quite as bad as when McAfee DAT 5958 misidentified svchost.exe (the parent process for all DLL-based services) on XP as malicious (and succesfully deleted it!) ... As you can imagine, this didn't go over well. I remember our shop being very glad we were on a delayed deployment for their DATs. http://slashdot.org/story/134550 http://slashdot.org/story/134550
- derleth 13y agoYou'd think that McAfee developers would be smart enough to hard-code a list of essential system files into their AV such that it will at least prompt you to get an install disk so it can replace them instead of just deleting them if it suspects they're infected. For bonus points, make the software realize that even if the file on the CD looks infected, it isn't actually infected, so make that file's SHA-256 sum a 'known good' profile. (If the file on the actual install CD really is infected, that falls into the category of "Problems The AV Can't Solve". At that point, the OS itself is controlled by Malign Forces Working To Destroy You and the game is over.) For extra special bonus points, code compressed copies of those essential files into the AV software itself, so they can be replaced on the fly without prompting anyone. They can be updated along with the malware profile data, if they ever need to be.
- greenyoda 13y ago1. Most consumer-grade machines don't come with install disks. 2. You can't put compressed copies of essential files into the AV software itself for a couple for a couple of reasons: - Microsoft will sue you for distributing their intellectual property without permission (and is not likely to grant such permission, since they can't control the quality of the third-party software). - These files are not static: they could have been modified by any Windows update, and might be dependent on other updated files. Keeping track of the hashes of "known good" files might work, but you'd have to account for files that were modified by Microsoft patches.
- nitrogen 13y agoWouldn't it be good enough to assume that any file with a valid Microsoft signature is safe? If Microsoft's signing system is compromised (which IIRC sort of happened once when a key issued for signing network credentials could also sign binaries, possibly used in nation-state malware?), you'll have bigger issues than antivirus software deleting essential system files.
- taspeotis 13y agoThis is why I don't run third party AV. I'm fairly good at distinguishing between what's a virus and what's not a virus. 99% of things are in the latter category care of fairly non-dubious web surfing habits. The last time I got a virus was sometime back when Vista was the new kid on the block. I was searching high and low for a keygen for a really obscure program and I thought I'd finally found one. Windows Defender was all up in my face like "don't run this, it's a virus!" and I was like "damnit Windows Defender, I want this key" so I ran it. Lesson learned. Viruses don't infect PCs. People infect PCs.
- derefr 13y agoLesson learned: wear a <s>condom</s> virtual machine.
- u12481632 13y agoOr as lighter solution use a sandbox program like sandboxie.
- bagels 13y agoThere's always good old fashioned drive by Java exploits served up by ad networks.
- ramidarigaz 13y agoPrecisely because of that I have "click to play" for all plugins in Chromium. I feel a million times better about it, and it's a very minor hassle.
- bcoates 13y agoI'm pretty sure click to play on Chrom* only keeps honest sites honest-- it doesn't prevent a site from activating its own plugin.
- ramidarigaz 13y ago
- grandpoobah 13y agoSucks for the grandma who runs Kaspersky and her internet is now broken and she has no clue why. That'll be a $150 callout from a local technician.
- qoo 13y agoMalware makers have gone offensive. One of the new tactics is to push fake data to antivirus vendors' servers. http://securitywatch.pcmag.com/hacking/317184-weaponized-antivirus-when-good-software-does-bad-things http://securitywatch.pcmag.com/hacking/317184-weaponized-ant...