Slashdot Mirror
The Dark Side Of DefCon's Wireless Network
An anonymous reader writes "While there's been a few postings on events happening at DefCon 12, one event seems to have been overlooked. A new wireless packet injection tool was quietly released (unleashed?) during DefCon: AirPwn. Here's a write-up of the tool as deployed by its author and crew at DefCon 12."
. . . but you need two wifi cards one on machine to use it.
everything in moderation
I was a victim of this at defcon, but since I was using lynx, I really didn't see any of the images mentioned. Actually, most of the surfing I did at defcon was using links or w3m over ssh (on a home box).
airpwn - bringing goatse (and friends) to Defcon 12!
Images from Dave's camera
Movies from Dave's camera
Images from my phone
At Defcon 12 this year my cow-orkers and I brought along a little piece of code called "airpwn." Airpwn is a platform for injection of application layer data on an 802.11b network. Although the potential for evil is very high with this tool, we decided to demonstrate it (and give it its first real field trial) on something nasty, but harmless (compared to say, wiping your hard-drive)
Over the course of defcon, we fielded 7 different airpwn configurations to see how well it worked, and of course to watch as 31337 h4x0rz got goatse up in their mug. The configurations were:
HTTP goatse, 100% of the screen
HTTP goatse replacing all images
HTTP goatse as the page background via CSS
HTTP tubgirl replacing all images
HTTP "owned" graphic, replacing all images (eventually I felt bad about all the ass pictures)
HTTP javascript alert boxes, letting people know just how pwned they were
FTP banners (while this worked, nobody pays attention to FTP banners so we abandoned this quickly)
How does it work?
airpwn requires two 802.11b interfaces, one for listening, and another for injecting. It uses a config file with multiple config sections to respond to specific data packets with arbitrary content. For example, in the HTML goatse example, we look for any TCP data packets starting with "GET" or "POST" and respond with a valid server response including a reference to the canonical goatse image. Here's the configuration file used for this mode:
begin goatse_html
match ^(GET|POST)
ignore ^GET [^ ?]+\.(jpg|jpeg|gif|png|tif|tiff)
response content/goatse_html
and here is the content that we return when the match is triggered:
HTTP/1.1 200 OK
Connection: close
Content-Type: text/html
pwnedOPEN YOUR MIND -- TO
THE ANUS!!
Each of the 7 modes mentioned previously varied in the configuration and content returned. In each case the poor user of the web browser was left feeling disgusted, afraid and/or confused. While I was busy operating airpwn at the laptop, my accomplices wandered the show-floor taking pictures and the occasional video of our victims. Links to our victims are at the top of the page.
In all honesty, the reaction to airpwn wasn't exactly what I had expected. When I was writing the code, I imagined that the second I turned airpwn on we'd hear immediate groans of disgust radiating out at the speed of light. In practice, airpwn's effect was simultaneously more private, and more full of personal drama. First off, the full-screen goatse seemed to be too powerful . The second it flashed on the screen, the savvy user would have the browser closed already. This made it incredibly difficult to actually catch the victims on film. Based on the logs generated by airpwn we would be hitting multiple people per second, but finding someone with goatse up on their screen was still a bit of a challenege.. Once we did find a victim, the results were pretty hillarious.. I had tears rolling down my cheeks on multiple occasions. The typical goatse reaction went something like this:
Open browser, see goatse, jump backwards a little
quickly close browser, take a breath
open browser, see goatse, close browser (faster this time)
scratch head, quit browser process, re-launch browser
see page indicating that goatse will load soon (page header, etc.) immediately close browser.
open up browser preferences, click all the tabs, look for the "no goatse" checkbox
clear the browser cache
open browser, see goatse, close browser
open network preferences, click on all the tabs, look for the "no goatse" checkbox.
disconnect from network, re-associate
open browser, see goatse, close browser
At this point, the less l33t people would generally give up and either 1) do something else or 2) look deep into goatse's anus with a 10-yard stare.. The m
It's a hacker conference. There is probably no more tolerant place to release such a piece of code, where your talents will be respected instead of persecuted. There were also no doubt many members of the computer security community present who would want to be aware of any new vulnerabilities immediately. I think it's a great thing it was tried and released at DefCon first.
He who laughs last is stuck in a time dilation bubble.
Do people still do this? Packet injections of various and sundry sorts are old news.
;)
There's a worrisome pattern, in the IT security biz, of repetition. Hacks discovered a few years ago re-appear in new clothes as "new," technologies for protecting against them resurface every few years in the same way. Computing as a whole tends to re-invent things on something like a 15 year cycle, but security seems to be on a truly frenetic clock, cycling every 2 years or so (very very approximately
Is there some connection between this and that vulnerabilties re-surface in new clothes constantly as well?
Well, it looks like all you hax0rz got them back by slashdotting their site.
Mirror mirror on the wall?
Someone get to a local Starbucks with this, fast! Oh, and bring your camera!
Three scenarios to point this out.
You're at Joes Internet Cafe, munching on your slightly overpriced muffin and glad for the free Wi-Fi access since you're out of town, and don't get to check your email much on the road. You hit the link to a message you want to read on webmail, when all of a sudden, an ad comes up. Nothing too bad, but it seems that Joe has decided that instead of charging people directly for 'net access, he'll rig up an old desktop with wireless to transmit the ad source for every 100th HTTP request that comes through his system.
This is a potentially annoying way of using the technology, but it also sounds like it could be a good way for Joe to help recoup his costs on the internet. Not a place I'd mind going.
Scenario Two
You're at Joes Internet Cafe, munching on your slightly overpriced bagel, glad for the...well, you know. This time the 'net access isn't free, but Joe's giving it out for $1 an hour, more than reasonable. 58 minutes in, you make an HTTP request, and a small javascript window pops up informing you that you've just got a couple minutes left, more time can be bought at the counter. After 60 minutes, instead of locking you out, all your requests simply get a screen advising you that if you want to keep going, Joe's going to need a dollar at the counter.
Seems useful to me.
Scenario Three
You're in Joes Internet Cafe, sipping some slightly overpriced coffee and you try to get online. After you've payed your dollar to the friendly man at the counter.
You keep gettings ads. You click out, thinking that it's a popup window, and no, you really don't need to enlarge that, it's fine how it is.
All browser windows closed. You try again.
No, I don't really need those drugs...
Or those pieces of software
Or...
You get the idea. Turns out, that guy in the corner is making some quick cash by spamming everyone in the place. The only sites that are coming through are from those ads. He leaves after about 15 minutes, because it can't be long until someone figures it out, but you've just lost 15 minutes of your time.
I realize it's an extreme example, but you think someone won't try it?
Joe, if you're out there, we need to talk. I've got some ideas for you.
Once the shock wore off, I pointed out the issue to my friends sitting next to me. They spent some time analyzing ethereal output, while I downloaded and ran arpwatch. It's pretty sad to hear that some kiddies were checking browser settings....
The article claims there was no arp poisoning going on, but actually there was. I saw plenty of that. Which kinda confused us, since there doesn't seem to be much need for that in a wireless environment. You can sniff w/o arping, and you can inject traffic (as they were). But yes, it was definitely happening, though apparently by a different group. (Actually, I detected three different MAC addresses competing for the AP's IP.)
In hindsight I should have saved some of my packet captures. Might have been fun to look over later.
After reading a few posts on this thread, I find it peculiar that so many slashdotters don't know that IPSec or related vpn products can be used to secure wireless.
Hi.
I wrote the manual page for airpwn.
All I see in this discussion is either people joking, bitching or having no idea how airpwn works.
Let's just set things straight.
First of all, there is no arp posioning.
Do you disagree? Well it's a GPL app, go read the source, show me the arp posion part of the code. What's that you can't find it? Oh, well jesus, it's because it doesn't do that.
You can hijack any tcp connection with this, it cannot be blocked without blocking the legit traffic.
This is accomplished by using raw frame injection.
One network card listens on a given channel (or in the case of a cisco card, all channels) and the other card simply injects custom frames with perfect replies. If your reply (it's up to you how big it is) is the right size, it's injected so perfectly that the connection not only still works, all of your webpage stuff still works, images just load as whatever the attacker wants.
It works with ftp, http, aim or whatever.
You can just have a ball.
It would be entirely possible to write regex that replied over aim or icq or any of that crap with a raw frame telling the other people in the conversation that they were coming out, it's up to you.
The software uses a very customizable framework to allow for use of regular expressions for matching. It's really useful for things other than goatse, but at defcon, they deserve the best.
Anyway, the totally clueless people here that claim to know how it works haven't even compiled it, so don't listen to them.
If you have any questions, feel free to ask.
"Not my manner of thinking but the manner of thinking of others has been the source of my unhappiness." - M
To Actually answer your questions.
1. SSL would effectively block this attack IF the user pays attention to invalid certs. Your browser contains certain CAs it trusts and, unless they had control of your PC which is certainly possible but was not done in this case, the CA they would use would be invalid and generate that pop-up box telling you so. If you ignore that box and click yes you do so at your own peril.
2. What about it? Once the data is on wifi than it is fair game for any type of manipulation. That is why they have 2 nics. The first nic "hears" your request for content "GET" and then responded much more quickly than the remote web server can with the corrupted "POST". When the correct information finally gets to your PC it is simply ignored as invalid TCP traffic and a RST packet is generated.
3. WEP would have stopped it in this instant. WEP is breakable but requires a good amount of data to be sent over the wire. Since your average user is not going to send GBs of data over HTTP and the processing power needed to break 100s of connection would be more than a couple of laptops could handle this attack would have been alot less fun. Still possible but would need to be much more dedicated. I run WEP at home, I know it will not stop the determined hacker but the casual war-drive will ignore me in favor of my many neighbors with open APs.
4. You are correct AS LONG AS you pay attention to the cert's trail. SSL really is two seperate pieces in my mind. 1 - encryption - End - To - End data encryption and 2 - Trust - I know the data I am receiving comes from the correct website. This is done with certificates. Since there is no God of the internet and we have to trust someone initially companies like verisign, etc have working with Microsoft, Mozilla, etc to get their root certs pre-installed in your browser. Anybody can generate a certificate but only companies that have passed the "Idenitifcation Test" with Verisgin or whoever can issue certs that will have the proper path back to a valid root cert. Please note Verisign has been duped before and even given out valid MS certs to non-microsoft organizations.
You may think it is lame but it is actually a harmless example of things to come. Why is wardriving so popular? Because 90% of the APs do NOT use WEP. If everyone used WEP that would stop casual attacks. Consider two fences. One a 3-ft high fence. This fence is only going to stop people who don't want to go in. The 2nd fence is 10' high with barbed wire. This can still be overcome but will require some dedication. That is the difference between open and WEP. The problem is nobody uses WEP so this attack will work most of the time with ease.
Regards
You're at Joe's internet cafe, or in an airport, etc. Suddenly, your internet explorer gets a web page redirect to some random porno movie of 3 guys raping a rather unattractive asian girl, complete with audio... in full screen mode. Since your laptop's audio is on, everyone in the area, including your girlfriend hear, "No don't put it in my pussy. [scream]"... And you're joe blow who doesn't know how to use the keyboard to close the window to save your life.
Yes, it could happen, particularly, if the geek in the corner is sniffing your WiFi traffic, and singles you out.
More serious would be something which noted when you wanted a secure site, such as a bank, and proxied to a full-screen web page image complete with security icons that tricked the user into sending you their password in the clear.
There are malicious 14 year olds with laptops out there that would find this awfully amusing.
When using WIFI, I generally always use an SSH port forward to encrypt and tunnel my traffic back to a 'safe' host.
At home, my AP is connected to a dedicated interface that only allows SSH. You could add port knocking for additional security.
Sure, SSH port forwards can still be disrupted or messed with. But not like plain HTTP.
BTW, nice hack!