Apple Responds to Exploit
Dave Schroeder writes, "This isn't so much of a root vulnerability as a default configuration that trusts the integrity of the local network services. This functionality has been around since NeXTSTEP, and is designed to allow for auto-configuration of new servers/machines brought into the network. The quick 'fix' for the vast majority of users who choose to implement it is to uncheck LDAPv3 and NetInfo altogether in Directory Access. Or, if LDAP services are used, just uncheck 'Use DHCP-supplied LDAP Server' in LDAPv3. ... One could argue that these features should be off by default, but if they are, it kind of wrecks the whole auto-configuration scheme." This sounds related to a great new feature in Mac OS X Server 10.3/Xserve called "automatic setup" that -- for machines that come with it preinstalled -- will get their address and LDAP server via DHCP and look for configuration files, and automatically configure the entire server, without any interaction beyond plugging it into the network and turning it on.
but it's as valid today as it ever was. There is a dichotomy between security and ease-of-use. Hitherto it has been impossible to have the one and the other simultaneously. Choose one.
Apple choose ease-of-use, and get criticised for leaving an open security "hole". Microsoft choose the same, and get criticised for (well, just about everything except wonderful marketing), and Linux chooses the other, and is criticised for poor ease-of-use.
That's not to say it's impossible, but it needs more than the current level of effort that goes into multi-node design. Apple is taking the first steps, and they've been somewhat burnt. Let's hope that doesn't discourage them from carrying on down the path... Unix as a genre can only learn from a successful easy-to-use and secure implementation of multi-machine computing. The thing is that you only learn by trying....
Simon.
Physicists get Hadrons!
Realistically, an issue trusting the LDAP server that your DHCP server points you at?
What is the world coming to?
Do I need to manually verify every single setting supplied to me by my DHCP server because I don't trust it?
These days, the internet is not a safe place, we all need to be more than just a little paranoid - but are you paranoid enough?
Visit CryptoGnome in his home.
No matter what sort of spin Apple puts on it, it's still retarded of them to trust LDAP to the point that UID=0 is trusted to be root.
Still, I don't think that this exploit is really that easy to take advantage of... the circumstances which would lead to it are fairly limited for now (until WiFi is as pervasive as air, anyway).
In many discussions, people downplay the importance of exploits like these because the attacker has to be on your local network to take advantage of the security hole. What about all of the mis-configured (or deliberately) open wi-fi networks out there? I think that wireless networking has changed the importance of "local exploits" by allowing somebody passing by to become a local entity on an open wi-fi network.
This problem seems little worse than other problems related to DHCP. If someone had access to your subnet and was able to configure a rogue DHCP server (e.g. to exploit the OS X ldap bug) they could just as easily return a rogue proxy as the default gateway or a tainted DNS server. If you are not vigilant about SSH warning messages or best practices you could be connecting to a server which is just recording your password and passing it along to the real server.
There may be something I missing, but this does not seem to be a problem with Mac OS X as much as it is with DHCP. DHCP in its simplest form is not secure. Using DHCP on a subnet requires trust. As with any other kind of security you will have to trust something, whether it is your computer or your home network.
I hope people do not blow this bug out of proportion too much.
I don't mind this at all.
:-)
No professional I know connects a server to the network BEFORE they configure security and network settings.
Shame on you if you do
In this case, the software is actually more vulnerable in a work environment, because it requires a compromised DHCP server on the local subnet. Most home users would probably notice if you plugged in another computer in their house. It's less likely to be noticed in a corporate environment, at least for long enough to compromise a few servers.
Besides, if it's possible for someone to sneak a compromised DHCP server on your network, you're basically screwed anyway.
.... turns out, if someone had RTFM, nobody would be talking about this.
> you have to have a malicious dhcp server on your subnet.
Keep in mind "your subnet" could be the WLAN at the coffee house (I must have seen 6 macs down there today - near the Castro in SF, in case anyone's interested), or a cable modem connection. This also means that if you can own one box on the network, you automatically get root on the all the others.
from the dictionary --
One who is zealous; one who engages warmly in any cause, and pursues his object with earnestness and ardor.
Doesn't sound so bad to me. Are you a Linux zealot? A Windows zealot? Does having a strong opinion make you a zealot? Or are you opinionless.
You are the lemming. Spewing the same tired crap about 'Mac zealots'.
Shut up, Anti-Mac zealot.
I work tech support, and if I had a dollar for every Windows owner that didn't understand the difference between right and left-clicking I could buy Slashdot and every AC posting to it.
Ah ahem, several storage servers like Snap and etc also come with this 'feature'..
and those run Linux...
Don't Tread on OpenSource
Sure, just send fake DHCP requests until the basestation uses up it's IP address pool.
[o]_O
Since this is an autoconfiguration feature, why not have it on only for the first boot after installing the OS? This way the computer can autoconfigure and then when it is configured it turns the feature off again.
The interactive way to Go -- http://www.playgo.to/iwtg/en/
so why the hell are you running a mission critical server via dhcp? give it a static address to negate even the possibility of the exploit you are talking about here.
The messenger service is used by many orginazations for alerts. Where I work, our servers use it to send alerts to those that manage them. Works well since, unlike e-mail, it will get immediate attention. A web browser that is able to execute scripts is much more complex and therefore venurable than one that just doens't execute code at all.
Get off it, when you provide services to the world, you open yourself to the poiibility of getting hacked. Look at Linux. Consider the holes in OpenSSH. Is it essential? No. Is it useful? Hell yes. When you run services that the whole world can get at, you run the risk that there is a flaw in the coding that someone exploits.
Now, a valid solution to this is to have everything turned off and/or locked down by default. Ok, that works, but is a pain in the ass (read not easy to use) since you must then figure out how to enable everything and make it work. IF you have useful services enabled by default, it runs the risk they are venurable and can be exploited by default.
By the way, if you have to reinstall Windows continually, you need to get some skills with Windows. To fuck it up that often and that bad indicate poor skills of the user.
So, you don't care whose rules YOU break, you just care that others follow YOUR rules.
Typical liberal.