The whole wi-fi security issue is starting to crop up with
increasing regularity in the mainstream media, but there is never
any discussion, just warnings of the dangers and warnings of the
consequences.
Most home wi-fi boxes offer three levels of "security"
under the appropriate tab in the internal web server config page,
none, WPA & WEP.
In various other tabs in the internal web server you can set
things like channel number and network name, but chances are most
of this is way over the head of the average user and they will
instead use a set-up wizard and accept whatever defaults it
suggests.
My home network consists of a cable modem feeding into a Belkin
F5D7231 wireless hub, one cat5 in from the cable modem, and one
cat5 out to the Planet DH1600 inside the Sun cabinet.
From the Planet there is cat5 to the streaming mp3 jukebox, to
the lan file server, to the RAQ, to the networked colour laser,
and one cat5 to the living room "multimedia" PC, and
one cable running upstairs.
The Belkin provides wireless access to this laptop, to the
girl's computer upstairs, and to the workshop (CNC
controller) computer.
Everything in the Sun cabinet I have set up with fixed IP in the
192.168 range, the wireless stuff is DHCP assigned in a different
part of the 192.168 range with permanent leases.
The Belkin also provides NAT and DynDNS to the RAQ, and I use
OpenDNS nameservers for it.
Security is "none", but the Firewall tab on the
internal web server I have MAC address filtering enabled, with
the MAC address of every one of the above machines enabled.
From the security perspective I'm not so much worried about
anyone stealing my bandwidth, I'd be more concerned about an
outsider accessing correspondence or worse still deleting files.
There are two other wireless networks in range of this laptop,
one is unsecured with a default name, one is secured with WPA and
is called s****snetwork, but Network Stumbler would make easy
meat out of either one. Both of them are often powered down.
My method is "picky", it is quite easy to create a set
up where (which is actually the situation) one computer (in my
case the workshop one) can access the LAN and play streaming
mp3's from the LAN and so on, but not access the internet as
a whole. In my case this is deliberate, but of course most people
want something that they can set up by clicking "OK" a
couple of times and everything just works.
I am not sure why, when using the word "security" with
wi-fi the default response is "encryption", even though
my RAQ is now all https, this is not a security response per se.
The default option when wishing to secure a server is not to
encrypt all traffic, the default option is to restrict access.
I am not someone who has vast experience of configuring wi-fi
networking, but one feature I have never ever seen on a wi-fi
internal web server is a feature that detects all stations in
range, and asks if you want to grant each one access, and if so,
what type and level of access.
Nor have I ever seen a wi-fi router that flatly refused to
function as a wi-fi router until you directly link to it via USB
and change settings from factory default, specifically setting an
admin password and a workgroup / network name.
This would be a LOT simpler than many of the wizards that
do ship.
I was particularly disturbed, a couple of years ago, when I
discovered that PCI and PCMCIA wi-fi cards weren't all hard
wired, but that most had a small amount of non volatile memory in
which settings were saved. I discovered this because I dual boot
my laptop between WinXP and Debian, and for the longest time ran
a Prism card for debian and a generic one for xp, then one day I
discovered that running the windows wireless network setup
wizards altered the data in this non volatile memory in such a
way that the card would no longer work reliably with linux, and
these cards do not come with buttons to reset to factory
defaults.
It appears to be the mentality that since THIS is the machine I
am using, then the device plugged into THIS machine is the one to
run the wizards on, and everything views the whole world of wi-fi
from that perspective, instead of the true perspective, which is
that all the clients revolve around the AP, the AP is the centre
of gravity, not whichever machine I happen to be sat at.
It is this mentality that gives us the broken security model we
have, where traffic is encrypted by default, but brute force
challenges can simply overcome them, whereas the method I use
requires you to brute force spoof every possible MAC address
before you can join the network.
I have watched competent coders fiddling with their devices for
many minutes trying to get net access in my house, their first
recourse is to see if there is a network broadcasting, there is,
and their second recourse is to see if they can log on to it,
they can't, and their third recourse is to just assume that
since their machine didn't log them on automagically, they
need a password from me.
This is despite the fact that even the default windows wizards
and tools actually give them enough information to indicate that
it isn't a password authentication problem, but a more
fundamental authentication problem.
I suppose if I was truly paranoid I could add encryption as a
SECOND level of defence after the first level of MAC address
filtering, but quite frankly it is overkill and creates more
problems that it is worth.
MAC addresses can be trivially forged. It's a one-line
command to do so.
There are multiple sorts of security, access restriction is what
you are talking about. Some form of shared secret is
necessary to restrict a network where anyone can forge a packet.
You can restrict access on an open network using a zero-knowledge
proof, and still leave the traffic unencrypted. But you end
up having to do the proof in every packet. It was easier
to just encrypt all traffic, and the user probably does want some
sort of eavesdropping protection.
The best way to do wireless, IMO, is to make it a separate local
network from your hard-wired machines, gatewayed to Internet but
not gatewayed into the wired LAN. Then, provide a VPN-enabled
router on the wired LAN, so that you can VPN in from the wireless
network. This requires three routers, but routers are cheap and
many wireless access points incorporate one.
One router handles your cable or DSL connection and creates net
1. The wired network is net 2, and gateways to net 1, the
wireless network is net 3 and gateways to net 1. Net 2 firewalls
itself as if it's connected directly to the Internet, but
accepts VPN connections. Net 3 blocks port 25 access so that
spammers don't camp outside your home.
GNOME has integrated VPN access, at least for OpenVPN, into its
network management applet. I don't know how this works on
Windows.
GNOME has integrated VPN access, at least for OpenVPN, into
its network management applet. I don't know how this works
on Windows.
One thing I noticed is that winders can't see a local
networked printer if connected over VPN. I was trying to figure
out why my Mom's printer wasn't working one day, all
sorts of printers were showing up but not hers, and as soon as
the VPN session was killed...there she be. Being her work laptop
I didn't even look at it so maybe just a checkbox or
something.
I would think that you could incorporate all that functionality
into one wireless router with extra little features like redirect
all wifi web traffic that doesn't have a valid ssh session
running to the St. Obama home page. Add in a virtual bridge
between the localnet and wifi net that drops all but VPN traffic
and you're peachy.
Learn a lot from looking over the openBSD FAQ every once in a
while.
You can fix that with some static routes to cover your local
subnet.
The VPN will assign you a second IP, but the first is still
there. It'll also assign new DNS and default
gateway. If you have a static route to cover your original
subnet, you can still get to local resources like the printer.
i.e. -- If your non-VPN IP is 192.168.1.5/24 with a gateway of
.1, then you need a route specifying 192.168.1.0/25 via .1
That way it'll persist when the VPN assigns new defaults and
you can print away.
Wait a minute here: I never tried that, but the MAC address _can_
be changed on a lot of wired ethernet card, so why not on
wireless card as well ?
If so, then hacking your setup is easy:
sniff for legitimate mac address
wait for one of the machine (typically a laptop) to go
offline
profit...
That's why my home network use WEP _and_ MAC address
filtering, as a first pest deterrent, but the security is
host-based. Because security should of course rely on layers,
one-armor fits-all is another design error.
Now, of course, if you are 100% sure that there is no way to
change a WIFI card MAC address, then you and I are safe (wanna
bet ? ;)
eh, of COURSE a mac address can be changed, the spec demands it!
with some protocols, such as DECnet, the mac address is changed
to encode the node's assigned address.
And WEP isn't secure at all, trivially cracked with any
number of free wares out there.
I'll put my network with an openly broadcast SSID, and WPA-2
encryption up against your non-encrypted but MAC address filtered
network and see which one gets pwned first.
For all intents and purposes, WPA-2 encryption on a wireless
network is more encryption than anyone really needs (so, by that
metric, it's just the right amount). If someone can crack the
encryption, simple MAC address filtering is merely an annoyance
to be overcome, not anything remotely resembling a serious
obstacle. As my AirPort Base Stations requires a reboot to change
the MAC address list (grr) I simply don't use it - likewise,
I don't bother hiding my network name.
Setting up a wlan on Apple equipment is as easy as it could be,
and by default the highest level of encryption is used.
Encryption is the answer to wireless security. If you
can't see what's going on over the wlan, you can't
participate on it. Restricting the access to particular devices,
where the identity of that device is trivial to impersonate is no
security at all, unless it's used in conjunction with
encryption.
Guy, nowadays there really is no security any longer in using
WEP. It's not exactly "wired
equivalent privacy" as it's much weaker than
WPA(2);
especially for stationary networks (like homes and offices)
snooping is trivial. You really should consider upgrading your
network (even at the price of a more capable router).
For the record, I use a single wlan router between the Internet
and my lan, I plainly broadcast the SSID, and use WPA2-PSK
encryption. I do also filter on the MAC address level, but I am
aware that it's just a matter of form. I see a steady level
of port scanning going on, but have not seen any evidence of
break-ins (yet).
BTW, and OT: My router is running DD-WRT, but I've just found out
that Tomato does
some things nicer. Does anyone have experience with DD-WRT vs.
Tomato?
Yes, I have experience with both, make the jump, you wont regret
it. Tomato is better by every metric I can think of, uptime,
memory usage, graphs, asthetics, interface, and especialy quality
of service. which works infinitely better on tomato.
I did a little research about wifi security when i bought my
first wifi router.
It turned out that MAC adress filtering, SSID hiding and WEP are
totally useless as they could all be found in a matter of
seconds.
WPA encryption is already better... eventhough a friend of mine
told me he could crack most of them in about 10min, using the
tools provided on the Backtrack livecd and an Eeepc.
It seems the only more or less reliable security scheme with wifi
is to use a separate RADIUS server combined with the WPA
encryption, but that's a bit overkill for a home network.
WPA encryption is already better... eventhough a friend of
mine told me he could crack most of them in about 10min, using
the tools provided on the Backtrack livecd and an Eeepc.
Your friend is exaggerating. The only way known to crack
WPA at the moment is brute force. Here is a detailed
example. You can shortcut a few things with
pre-computing but the end result is, if you use a password of
moderate complexity and the recommended length (20-63
characters), you're secure.
The tool on Backtrack he is referring to is probably coWPAtty.
Where I work we have weekly technology presentations... mostly
it's very detailed work related stuff but occasionally we do
have something that is not work related. Mostly it's
students doing their diploma work. Last year we had a kid
teach us all about setting up wireless networks and about
wireless security. During this demonstration he broke every
security method available on some cheap router using a
laptop. The shortest times were virtually instantaneous,
the longest was finished in the time it took me to walk down the
hall, make a cup of coffee, and return. He also said
he could do it using my PDA but it would take longer.
If anyone over the age of about 5 told me they could break into
the wireless security available on consumer level routers in 10
minutes, I think I'd believe them.
You can download 4+ Gb of precomputed hashes for a few hundred
thousand passwords for coWPAtty to greatly speed things up.
I can crack WPA, depending on the password chosen. Poor
passwords are the bane of any system. But I haven't
seen any hints of anyone cracking a properly set-up WPA system
with "good" passwords. Nor have I heard whispers
of a flaw in the crypto algorithms or any vendor's
implementation.
WEP is trivial. And setting up the demonstration you talk
about is also trivial -- assuming you use either short or
dictionary-based passwords.
I have "WPA/AES" with a 63 character random password
generated by the "pwsafe" Linux port of Bruce
Schneier's "passwordsafe" program. I was
under the impression that all of the precomputed hashes were
based on dictionary words. I was also under the impression
that WPA was practically uncrackable for random passwords above
about 20 characters. There are some weaknesses to TKIP, but
I thought using WPA/AES uses CCMP instead of TKIP, closing that
hole.
When I say "WPA/AES" this is on a Linksys WRT54G
V3. (The Linux-capable one, though I'm still running
Linksys software.) I did some reading that indicated that
that setting got me away from the somewhat-weak TKIP.
The WRT54G (affectionately referred to as "Warthog")
has had its settings tweaked to act as a pure access point, dhcp
server turned off, etc, and my LAN is hooked into one of its 4
LAN-side ports.
It spends most of its time with ultimate security -
unplugged. When I bring the laptop home from work, and will
want to work wirelessly, I plug it in.
As other have said, MAC filtering and SSID hiding don't
really do much. But it's interesting as still others
have said, it adds a level of plausible deniability to the
security. There is no accidental connecting, in this case.
Pre-computed hashes are dependent on the SSID, so just about all
of the ones you can grab online are for things like
"linksys", "netgear" and
"dlink". If you have a non-default SSID, then
someone would have to precompute their own hashes.
For all intents and purposes, if you're using 20+ characters
as a WPA-2/PSK under AES passphrase, you're covered. If
anyone could crack that it would have to pretty much be an
implementation flaw in a specific router/firmware and it would be
shouted from the rooftops at BlackHat, DefCon or ShmooCon.
Is WEP down to only a few minutes now? Last time I cracked
a WEP I had to sniff about 10 million packets to get enough weak
IVs to crack it. Took about 12 hours of unrealistically
high traffic (simulated) to get that amount of packets.
Security is "none", but the Firewall tab on the
internal web server I have MAC address filtering enabled, with
the MAC address of every one of the above machines enabled.
I assume you know it is quite easy to sniff and then spoof a MAC
address. MAC filtering will stop casual leeching, but not anyone
who is actually trying to crack your network.
I assume you know it is quite easy to sniff and then spoof a
MAC address. MAC filtering will stop casual leeching, but not
anyone who is actually trying to crack your network.
Oh, agreed, the Belkin itself has a "feature" built in
that allows it to spoof the MAC address of the PC originally
connected to the cable modem.
But you are the first one to make the salient point and
differentiate between the casual leecher (wireless zero config
machines) and the determined cracker.
This is the important point, good security is easy security, and
easy security doesn't consist of a hammer that sees
everything as a nail.
Hacking together a webcam to whack onto the wi-fi network as a
low power device doesn't require any encryption, it
doesn't benefit the functionality in any way, this is just
one trivial example.
I also take Bruce's point that splitting networks into in
effect a DMZ and a prive area works, but again I don't like
it, why not just work with the one network, it is simpler and
more elegant.
Others claim it will be easy to hack into my network, but again,
the way it is configured you won't get any DNS queries
answered unless you know what you're doing, and if you know
what you're doing you won't be wasting your time trying
to get into my virtual internet.
Plus, accessing a network doesn't give heirarchical access to
every device on that network, so the mp3 jukebox is deliberately
easy to access, but the RAQ is a lot harder.
President4242 gets it, without giving away precisely how my
network is set up, the definition of a real
honeypot is something that gives no data and no indication that
there are one or more devices that you don't know about, in
addition to the ones you do.
Wi-fi cards do not need to be any more "intelligent"
than an cat5 ethernet adapter, cat5 cables are just as much an
antennae as a wifi antenna after all, and these devices are sold
as HOME wi-fi APs, not ISP ones.
The whole none / WEP / WPA thing is a solution to a problem
created by the way home AP's are designed, MAC spoofing might
be straightforwards, MAC spoofing won't help with Radius
sessions, etc, and all these lego bricks were already out there,
there was no need to create new ways of doing old jobs.
WPA renders perfectly functional older hardware as obsolete, and
yet the very same people who lambast winders and praise the fact
that their obsolete computer will quite happily run linux, will
quite happily accept throwing away networking hardware.
WPA is also a pain in the ass if, for whatever reason, I decide
to change the key, perhaps I detect a failed intrusion and
realise my key is not complex enough.
WPA does not degrade, you are either in or out, and quite frankly
it is stupid to force a webcam on the network to behave as though
it is the machine with all my financial records on it.
WPA is not immune to any physical attack, unattended access to a
box, particularly something low tech like a webcam, gives me the
keys to the entire network, so it makes every single device a
keystone.
Windows has been (quite rightly) lambasted for its crap file
permissions system, the lack of owner / group / public limits
people to "just enable file sharing" in windows
wizards, or leaving it off.
This is the same binary response that wi-fi "security"
currently offers, if windows gets lamnasted for it being crap
then unless we are all hypocrites so should wi-fi, because it is
equally crap.
The actual justification of this original article is the
responses so far, most of them amounted to WPA is teh roxxor, and
nobody analysed the implications that usage of WPA forces upon
the users.
WPA changes, utterly, the fundamental ways in which networks were
designed, par for the course for anything with Microsoft and
other IP and DRM advocates as sponsors and members....
http://www.wi-fi.org/our_members.php
I have 4 wireless routers in the same house. 2 of these are
simple WPA encrypted, one for each laptop, both on channels other
than 6. The other two are wide open, non-encrypted, but
everything other than port 80 shut down, filtered out both ways,
as in not open even to https: browsing.
One of those is a T-Mobile hotspot that includes port 80 phone
calls- so my wife's phone is free when she's in the
house. The other is a www.personaltelco.net node, which I
hope eventually to attach a dish to, in order to provide some
small level of browsing capability at the park across the street
to the neighborhood.
Since the T-Mobile and Personal Telco hotspots are so easy to get
into, I've never even recorded a bad attempt to get into the
other two routers.
In today's climate, part of "security" means
"legally cover my butt". I use an SSID of
"private_keep_out" and don't broadcast it.
There is absolutely no way anyone is going to claim their PC
automatically connected to my network.
There are also another half-dozen b/g networks in range of my
house that are softer targets.
I just switched to 802.11n for wireless, using ndiswrapper for
the drivers under Linux until someone gets real drivers
written. The router is supported by DD-WRT, so I'm
going to try that this weekend.
Right now I use WPA2-PSK, with a 45 character key. I have
two printers, one laser and one label (USB), a server, VoIP phone
and TiVo connected via wire. There are 4 PCs connected via
wireless.
I do have MAC filtering turned on, just to weed out the amatuers,
and have assigned static IPs. DHCP is off. I also
have the router set to "N-only" mode, which means it
won't accept A, B or G connections. Just for
entertainment value, the router syslogs everything to the server.
Our research showed that the default settings for wireless APs
play a crucial role. We found for consumer wireless
access points (APs), such as Linksys, an encryption usage of 23%
overall. In contrast, 2Wire had an encryption usage of 96%
(2008). The explanation for this difference is largely due to
2Wire has a default setting for encryption turned on.
Most APs have defaults for security set to off to save
money in support costs. The paper can be found at:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=964950
Wi-fi security.
The whole wi-fi security issue is starting to crop up with increasing regularity in the mainstream media, but there is never any discussion, just warnings of the dangers and warnings of the consequences.
Most home wi-fi boxes offer three levels of "security" under the appropriate tab in the internal web server config page, none, WPA & WEP.
In various other tabs in the internal web server you can set things like channel number and network name, but chances are most of this is way over the head of the average user and they will instead use a set-up wizard and accept whatever defaults it suggests.
My home network consists of a cable modem feeding into a Belkin F5D7231 wireless hub, one cat5 in from the cable modem, and one cat5 out to the Planet DH1600 inside the Sun cabinet.
From the Planet there is cat5 to the streaming mp3 jukebox, to the lan file server, to the RAQ, to the networked colour laser, and one cat5 to the living room "multimedia" PC, and one cable running upstairs.
The Belkin provides wireless access to this laptop, to the girl's computer upstairs, and to the workshop (CNC controller) computer.
Everything in the Sun cabinet I have set up with fixed IP in the 192.168 range, the wireless stuff is DHCP assigned in a different part of the 192.168 range with permanent leases.
The Belkin also provides NAT and DynDNS to the RAQ, and I use OpenDNS nameservers for it.
Security is "none", but the Firewall tab on the internal web server I have MAC address filtering enabled, with the MAC address of every one of the above machines enabled.
From the security perspective I'm not so much worried about anyone stealing my bandwidth, I'd be more concerned about an outsider accessing correspondence or worse still deleting files.
There are two other wireless networks in range of this laptop, one is unsecured with a default name, one is secured with WPA and is called s****snetwork, but Network Stumbler would make easy meat out of either one. Both of them are often powered down.
My method is "picky", it is quite easy to create a set up where (which is actually the situation) one computer (in my case the workshop one) can access the LAN and play streaming mp3's from the LAN and so on, but not access the internet as a whole. In my case this is deliberate, but of course most people want something that they can set up by clicking "OK" a couple of times and everything just works.
I am not sure why, when using the word "security" with wi-fi the default response is "encryption", even though my RAQ is now all https, this is not a security response per se.
The default option when wishing to secure a server is not to encrypt all traffic, the default option is to restrict access.
I am not someone who has vast experience of configuring wi-fi networking, but one feature I have never ever seen on a wi-fi internal web server is a feature that detects all stations in range, and asks if you want to grant each one access, and if so, what type and level of access.
Nor have I ever seen a wi-fi router that flatly refused to function as a wi-fi router until you directly link to it via USB and change settings from factory default, specifically setting an admin password and a workgroup / network name.
This would be a LOT simpler than many of the wizards that do ship.
I was particularly disturbed, a couple of years ago, when I discovered that PCI and PCMCIA wi-fi cards weren't all hard wired, but that most had a small amount of non volatile memory in which settings were saved. I discovered this because I dual boot my laptop between WinXP and Debian, and for the longest time ran a Prism card for debian and a generic one for xp, then one day I discovered that running the windows wireless network setup wizards altered the data in this non volatile memory in such a way that the card would no longer work reliably with linux, and these cards do not come with buttons to reset to factory defaults.
It appears to be the mentality that since THIS is the machine I am using, then the device plugged into THIS machine is the one to run the wizards on, and everything views the whole world of wi-fi from that perspective, instead of the true perspective, which is that all the clients revolve around the AP, the AP is the centre of gravity, not whichever machine I happen to be sat at.
It is this mentality that gives us the broken security model we have, where traffic is encrypted by default, but brute force challenges can simply overcome them, whereas the method I use requires you to brute force spoof every possible MAC address before you can join the network.
I have watched competent coders fiddling with their devices for many minutes trying to get net access in my house, their first recourse is to see if there is a network broadcasting, there is, and their second recourse is to see if they can log on to it, they can't, and their third recourse is to just assume that since their machine didn't log them on automagically, they need a password from me.
This is despite the fact that even the default windows wizards and tools actually give them enough information to indicate that it isn't a password authentication problem, but a more fundamental authentication problem.
I suppose if I was truly paranoid I could add encryption as a SECOND level of defence after the first level of MAC address filtering, but quite frankly it is overkill and creates more problems that it is worth.