Is my hard drive opaque on a public network?

July 19, 2009 4:09 PM

Wi-Fi security. How much of the contents of my a) iPhone or b) iBook can be "seen" if I'm on a public network? Particularly concerning the latter, I've never been clear whether, in logging on to network X--be it a free municipal net or my employer's "guest net"--I'm simply an inscrutable blip on a network or I'm making my hard drive an open book to anyone who cares to peek. Any insight would be appreciated.
posted by the sobsister to computers & internet (11 answers total)

One thing to check on your iBook are your settings in System Preferences -> Sharing, to make sure all the boxes are unchecked. Beyond that, I don't know, but that's one way you could have accidentally made it so someone can look at some of your files.

Your network traffic can be seen by anyone unless you're connected through https, I think.

Although, if your computer isn't up to date through Software Update, maybe someone could get in and look at your files through a security vulnerability.

I don't know anything about the iPhone.
posted by losvedir at 4:17 PM on July 19, 2009

First off, there's nothing special about WiFi--all this applies even if you were to plug in an ethernet cable instead.

On most networks, anybody else on the network with a minimum of skill can see your network traffic. So, if I'm sitting next to you at the library, I can eavesdrop on your "conversation" with metafilter.com. I used to do this all the time during boring college lectures: I'd put my network card in promiscuous mode, and giggle at all the people hitting porn sites in a lecture hall.

This is the precise reason that we have HTTPS: so that when you log into your banking site, it's encrypted from your CPU to the server's CPU, and you can't be eavesdropped upon.

Please note that HTTPS encryption, like when you go to your bank's site, is totally separate from WiFi encryption. WiFi encryption prevents people outside the WiFi network from gaining access to it or listening to communications. Once somebody is part of the network (i.e. the have the keys), they can listen to all traffic that reaches them. It's like a fence: put a fence around your property, and people can't look in your windows... unless you let them inside the fence.

One thing to note is that while people can watch your traffic, it's pretty hard to identify what actual human being is generating that traffic simply from watching it go by. So, if I'm watching the network, and I see somebody go to a porn site, I can't tell that's Brad Pitt or the girl in the next row over unless I later get access to their computers or unless I see something on their screen that matches. The exception to this is, of course, if the actual traffic itself carries personally-identifiable information. So, if you fill out a registration form that asks for your name, an eavesdropper could catch that, and then relate that name to that MAC address indefinitely.

As for your whether people can read your harddrive: that depends on how you've configured your machine. If you have drive sharing turned on with no password, then of course somebody else can read your drive. On the other hand, if you haven't turned on sharing, and your computer is properly updated, the chances of your drive being compromised are very low.

Don't let all this make you paranoid, though: security-sensitive sites should already be using HTTPS, which prevents eavesdropping. And, if the content isn't sensitive, it doesn't really matter if it's being eavesdropped. The one thing I would suggest you be careful about is using IM and email applications to discuss sensitive topics on public networks. By default, neither IM nor email is encrypted (with most clients not even having an option). Without encryption, those conversations are open books. (Although you can certainly add encryption as a separate layer: see GPG.)
posted by Netzapper at 4:35 PM on July 19, 2009 [11 favorites]

The one thing I would suggest you be careful about is using IM and email applications to discuss sensitive topics on public networks. By default, neither IM nor email is encrypted (with most clients not even having an option). Without encryption, those conversations are open books. (Although you can certainly add encryption as a separate layer: see GPG.)

Although most (all?) webmail applications can be accessed via HTTPS. So if you're using, for instance, Gmail, access it at https://mail.google.com and your traffic will be encrypted between your computer and Google's servers. Of course, Google's servers then send it out in plaintext to the email's recipient, but someone on your network won't be able to eavesdrop on that part of the process.
posted by mr_roboto at 5:16 PM on July 19, 2009

HTTPS only offers a degree of security, it's not impregnable like some of the comments have implied. If I pretend to be the wireless gateway, and replace the SSL certificate from your bank with my own, I can then eavesdrop on all the HTTPS traffic. This attack is trivial to perform, I first saw it done in an office setting with Cain & Able over four years ago.

An nmap -sS of your macbook should give a good idea of what services are running on your macbook and possible vectors of attack.
posted by zentrification at 6:24 PM on July 19, 2009

zentrification, that only works if either (a) you got some certificate authority to give you a certificate saying that you're the guy's bank, or (b) the user is an idiot who clicks past the "This doesn't look like your bank. Continue anyway?" dialog.
posted by hattifattener at 6:35 PM on July 19, 2009 [1 favorite]

HTTPS only offers a degree of security, it's not impregnable like some of the comments have implied. If I pretend to be the wireless gateway, and replace the SSL certificate from your bank with my own, I can then eavesdrop on all the HTTPS traffic. This attack is trivial to perform, I first saw it done in an office setting with Cain & Able over four years ago.

This is true. I didn't mean to imply that TLS is bomb-proof, only that it prevents trivial promiscuous-mode eavesdropping. The man-in-the-middle attack has a chance of working, especially if the attacker is masquerading as infrastructure.

However, wouldn't the newer anti-MITM tools in, say, Firefox at least pop up a giant warning that says "This certificate doesn't match the one from last week"? Of course, that doesn't help if you go to a never-before-visited encrypted site while the MITM attack is active; but, if you connect to your bank from a "trusted" network (i.e. your home network), and then go out and use a compromised Starbucks network, your browser should give you some warning.

Furthermore, your fake certificate is going to be self-signed ('cause I doubt you can get VeriSign to issue you a cert in Bank of America's name). This also pops up a big, ugly warning on modern browsers--on Firefox, it actually makes the site fucking difficult to access.
posted by Netzapper at 6:48 PM on July 19, 2009

Well, as of at least 8 months ago it is possible to forge and SSL certificate, more from Schneier. Firefox or whatever browser may warn you about the certificate, but you can still use it, a clever attacker may only switch out certificates on lesser known sites...
posted by zentrification at 10:11 PM on July 19, 2009

browser may warn you about the certificate, but you can still use it

Well, duh, if you click OK to the dialog that tells you the certificate may be from an impostor, you shouldn't be surprised to discover that you can end up using a certificate from an impostor. That's not a "man in the middle" attack, that's a "user knows they're being messed with, doesn't care" attack.

The MD5 collision you link to is more interesting, but the CAs I spot-checked just now are issuing SHA-1 signed certs these days. I mean, RSA was recommending aganst continued use of MD5 in 1996; even Microsoft may have gotten that memo by now. SHA-1's kinda shaky too these days but not as bad as MD5 yet, and even MD5 is still only yielding collisions, not preimages. Until something like that happens, SSL will protect you from a MITM attack— your worries are a compromised client, compromised server, or compromised CA.
posted by hattifattener at 12:47 AM on July 20, 2009

Netzapper and other respondents,

Thanks for your answers. I'm certainly clearer on the situation.
posted by the sobsister at 7:27 PM on July 20, 2009

more attack vectors
http://www.wired.com/threatlevel/2009/07/kaminsky/
posted by zentrification at 12:02 AM on July 30, 2009

sslsniff, scary
posted by zentrification at 8:50 PM on August 4, 2009