I’ve been meaning to work on this little project for quite some time, but life got in the way and I was always too busy. Now I finally got some time back for myself and I’m here talking about it.
Quite some time ago I checked out the Wireless LAN Security Megaprimer course by Vivek Ramachandran (very nice, highly recommended) and incidentally in the same period I was doing some traveling which meant I got to stay in different hotels all providing Wi-Fi. Needless to say, my brain started to go crazy and I’ve thus been thinking about “unconventional” ways to get Wi-Fi passwords.
Can’t turn my brain off, you know.
We go into some place,
and all I can do is see the angles.
– Danny Ocean (Ocean’s Twelve)
The idea that I’m about to describe is pretty simple and, probably, not that new either. Nevertheless it was a fun way for me to get my hands dirty and play around with my Raspberry Pi which has been sitting on the shelf for too long.
Pretty straightforward, right?
In order to achieve this result I had to figure out these three things:
- How to create a fake access point
- How to force people to connect to it
This is pretty simple, it was also covered in the Wireless LAN Security Megaprimer and there are a number of different github repositories and gists that one can use to get started and create a fake access point. I thus won’t go too much into details but for the sake of completeness let’s discuss the one I used. I used
hostapd to create the Wi-Fi access point,
dnsmasq as a DHCP server and DNS relay server and
iptables to create the NAT network. The bash script that follows will create a very simple Wi-Fi access point not protected by any password. I put some comments in the code that I hope will improve readability.
# the internet interface
# the wifi interface
# The ESSID
# bring interfaces up
ip link set dev $internet up
ip link set dev $phy up
# Set default gateway
# Set DNS servers to announce
# Upstream DNS server
" > tmp-dnsmasq.conf
# start dnsmasq which provides DNS relaying service
# Enable Internet connection sharing
# configuring ip forwarding
echo '1' > /proc/sys/net/ipv4/ip_forward
# configuring NAT
iptables -A FORWARD -i $internet -o $phy -m state --state ESTABLISHED,RELATED -j ACCEPT
iptables -A FORWARD -i $phy -o $internet -j ACCEPT
iptables -t nat -A POSTROUTING -o $internet -j MASQUERADE
# all mac addresses allowed
wmm_enabled=0" > tmp-hotspot.conf
# Start hostapd in screen hostapd
echo "Start hostapd in screen hostapd"
screen -dmS hostapd hostapd tmp-hotspot.conf
DISCLAIMER: I left this section intentionally at a high-level of description without any code for different reasons. However, if it gets enough attention and people are interested about it, I will probably extend it with a more in-depth discussion providing, perhaps, some code and practical guide.
Depending on your target there might be different ways to try and get someone to connect to a fake access point. Let’s discuss two scenarios:
In this case the target is connected to a password protected Wi-Fi, perhaps the same password protected Wi-Fi the attacker is trying to access. There are a couple of things to try in this case, but first let me discuss something interesting about how Wi-Fi works. The beloved 802.11 standard has many interesting features, one of which I’ve always found … amusing. The 802.11 defines a special packet that, regardless of the encryption, the password, the infrastructure or anything at all, if sent to a client will simply disconnect that client from the access point. If you send it once, the client will disconnect and immediately reconnect, and the end user won’t even notice that something happened. However, if you keep sending it, the client will eventually give up meaning you can actually jamming the Wi-Fi connection and the user will notice that his not connected to the access point anymore. By abusing this characteristic, you can simply force a client to disconnect from the legitimate access point it is connected to. At this point two things can be done:
- The attacker can create a fake access point with the same ESSID of the access point the target was connected to, but without a password. In this case the attacker should hope that once the user realizes his not connected to the access point, he would try to manually connect again. The target will thus find two networks with the same ESSID, one will have a locker while the other one won’t. The user might try to connect to the one with the locker first, which won’t work because the attacker it jamming it, and chances are that he might also try the one without the locker … after all it has the same name of the one he wants so badly to connect to … right??
- Another interesting behavior that can be exploited, is that whenever a client is not connected to an access point, it keeps sending beacon packets looking for previously known ESSIDs. If the target had connected to an unprotected access point, and chances are that he did, the attacker can simply create a fake access point with the same ESSID of the unprotected access point visited by the target. As a result, the client will happily connect to the fake access point.
In this case the target is not connected to any Wi-Fi access point, maybe because the target is a smartphone and the owner is walking on the street. However, chances are, that the Wi-Fi card is still turned on and the device is still looking for known Wi-Fi ESSIDs. Once again, as discussed previously, chances are that the target had connected to an unprotected Wi-Fi access point. Thus the attacker can just create a fake access point with the ESSID of an access point the target had connected to and just like before, the Wi-Fi client will happily connect to the fake access point.
As a target router to attack, I used my home Wi-Fi router, specifically a D-Link DVA-5592 which was (not so) freely provided by my ISP. Unfortunately, for the time being, I don’t have other devices that I can test, so I have to make due with it.
Let me borrow the definition from MDN web docs:
The same-origin policy is a critical security mechanism that restricts how a document or script loaded from one origin can interact with a resource from another origin. It helps to isolate potentially malicious documents, reducing possible attack vectors. https://developer.mozilla.org/en-US/docs/Web/Security/Same-origin_policy
Let me again borrow the definition from MDN web docs:
The X-Frame-Options HTTP response header can be used to indicate whether or not a browser should be allowed to render a page in a
<object>. Sites can use this to avoid clickjacking attacks, by ensuring that their content is not embedded into other sites. https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/X-Frame-Options
That’s pretty straightforward: the
X-Frame-Options is used to prevent pages from being loaded in an
So let’s see the response that we get from requesting the login page of the D-Link:
HTTP/1.1 200 OK
The response contains
X-Frame-Options set to
DENY (thanks god) meaning that if I was hoping to load it in an
Same-Origin-Policy will prevent any interaction with the router itself. The simple solution I came up with, and beware it might not be the only one, is the following:
iframe inside the infected page. The
src parameter of the
iframe will point to the router’s IP address. As already said, the router’s IP address has the
X-Frame-Options set to
DENY so the
iframe executes, the victim is still connected to the fake access point (remember the advantage point I mentioned earlier?). Which means that the request to the router’s IP address will be handled by the fake access point … how convenient. The fake access point can thus intercepts any request performed towards the router’s IP address and respond with a web page that:
- doesn’t have the
- includes headers to cache the page.
Since the fake access point poses as the legitimate router, the browser will cache a page for which the domain is the router’s IP address thus bypassing both the
Same-Origin-Policy and the
X-Frame-Options. Finally, once the infected client connects back to its home router:
iframepointing the router’s IP address,
In the case of the D-Link, the login request looks like this:
POST /ui/login HTTP/1.1
The important parameters here are:
userPwdwhich looks encrypted;
noncewhich is definitely related to the encrypted password.
Diving in the source code of the login page I almost immediately noticed this:
document.form.userPwd.value = CryptoJS.HmacSHA256(document.form.origUserPwd.value, document.form.nonce.value);
Which means that the login requires the CryptoJS library and takes the nonce from
Once inside the router, I need to look around for the location where I can retrieve the Wi-Fi password. The current firmware in the D-Link DVA-5592 shows the Wi-Fi password IN PLAINTEXT (oh boy) right after the login in the dashboard page.
// this is CryptoJS, a bit annoying to have it here
This page should be cached in client’s browser and sent as coming from the router’s IP address. Let’s now put this code aside for a moment and let’s discuss the network configuration that will allow to cache it.
// list of common router's IP .. which definitely requires improvement
iframe for every IP in the array
routers. This is because the IP of the home router might have been configured differently. This means that the Raspberry has to answer to different IPs on different subnets. To do so, I simply add more IPs to the wireless interface of the Raspberry. In this way, whenever the code that loads the
iframe is executed, a request is performed towards common routers’ IP addresses and the wireless interface of the Raspberry can pretend to be the router, answer to those requests and cache whatever I want.
var http = require("http");
At this point I had already done some tests on my own but I wanted to try it in a more realistic environment. However, I can’t just try it on anyone without their consent, so first I needed a victim that would willingly be part of this little experiment … so I asked my girlfriend. The conversation went something like this:
Great, now that I had a victim that willingly decided to participate in this little experiment, it was time to start the hack and lure her iPhone 6 to connect to my fake access point. I thus created a fake access point with an ESSID I knew she has visited before (yes, also intel on your victim can be helpful) and soon enough her iPhone connected to my fake access point.
I let her browse the web while still connected to the fake access point and patiently waited for her to end up on an HTTP only web site.
He waded out into the shallows,
and he waited there three days
and three nights,
till all manner of sea creatures
came acclimated to his presence.
And on the fourth morning, …
Mr. Gibbs (Pirates of the Caribbean: The Curse of the Black Pearl)
Finally, bettercap flickered and printed that something had been injected, which meant I didn’t need to have her connected to my access point anymore.
The whole thing is available in my github, it can definetly be improved and, hopefully, by the time you read this it already has. Support for new routers should also be added as well. If I have the change to put my hands on other devices I will definetly add them to the repository. In the meantime, have fun and …