Shifting the Focus of WiFi Security:

Beyond cracking your neighbor's WEP key

Who are we and why do you care?

Thomas “Mister_X” d'Otreppe de BouvetteFounder of Aircrack-ng

Rick “Zero_Chaos” FarinaAircrack-ng Team MemberEmbedded Development

DISCLAIMER:

Some of the topics in this presentation may be used to break the law in new and exciting ways…

of course we do not recommend breaking the law and it is your responsibility to check your local laws and abide by them.

DO NOT blame us when a three letter organization knocks on your door.

Contest

Find the AP We have hidden an AP somewhere in the

airwaves Report the frequency of operation and

mac address to win (Insiders and friends are not eligible)

Spoils (first winner only)

Find the AP before the end of the talkFull price of Ubiquiti SRC wifi card

Find the AP before 1pm$50 towards a nice Atheros card

Find the AP after 1pmHearty handshake and a pat on the back

History of WEP Attacks / Why it doesn’t work

Passively Sniff for a long timeSlow, not enough data, impatientNo more weak ivs

Replay/Injection AttacksFast but very noisySimple signaturesAP features that try to block (PSPF)

History of WPA Attacks / Why it doesn’t work

Pre-shared keyRequires catching both sides of a quick

handshakeMust be in range of client and AP

EnterpriseNearly impossible to crack passivelyMost EAP types are difficult (at best) to MiTM

The Well Guarded Door

Nearly 100% of attacks focus on the AP APs are getting more and more secure New features built into AP

PSPF / Client IsolationStrong Authentication / EncryptionLightweight controller based architecture

APs are no longer the unguarded back doorWell deployed with fore thought for securityWell developed industry best practices

Take the Path of Least ResistanceAttack the Clients!

Tools have slowly appeared recently Difficult to use Odd requirements to make function

Attacking Client WEP Key

Wep0ff Caffe-Latte Hirte Attack

Attacking Client WPA Key

WPA-PSKNo public implementation

WPA-ENTFreeradius-wpe (thanks Brad and Josh!)Requires hardware AP

Attacking the Client

Many Separate Tools Difficult to configure Typically sparsely documented Odd requirements and configurations

Until now…

Introducing Airbase-ng

Full monitor mode AP simulation, needs no extra hardware

Merges many tools into one Also works in Ad-hoc mode New and improved, simplified implementations Easy, fast, deadly (to encryption keys at least)

Airbase-ng Abilities

Evil Twin / Honey Pot Karma WEP attacks WPA-PSK attacks WPA-Enterprise attacks (coming soon)

Airbase-ngFeatures Soft AP

WEP• Open/Shared auth• Caffe Latte• Hirte attack

Capture WPA/WPA2 handshake

Manipulate and resend packets

Encrypt/Decrypt packets

Airbase-ng Features Filtering to avoid disturbing nearby networks

AP FiltersBSSIDsESSIDs

Client filtersMAC Filtering (allow/disallow)

Airbase-ng Abilities

WPA Handshake capture: airbase-ng -W 1 -c 5 -z 2 -I 102 --essid myAP rausb0

Script to manipulate packets: airbase-ng –Y both rausb0 then start replay.py at1

Soft AP: airbase-ng –y –e myAP –c 5 –I 102 rausb0 ifconfig at0 up 192.168.0.254 ping/ssh/… it from the client

What are you, a blackhat?

No seriously, this doesn’t promise a win There are ways to defend as well APs are finally being configured securely,

now clients must be as well

Simple Defenses

Proper Secure Client Configurations Check the right boxes GPO

A Step Beyond Crazy

WiFi Frequencies .11b/g 2412-2462 (US) .11a 5180-5320, 5745-5825 (US)

Does this look odd to anyone else?Does the card really not have the ability to

use 5320-5740?

Licensed Bands Some vendors carry licensed radios Special wifi cards for use by military and

public safety Typically expensive Requires a license to even purchase Frequencies of 4920 seem surprisingly

close to 5180

Can we do this cheaper?

Atheros and others sometimes support more channels

Allows for 1 radio to be sold for many purposes.

Software controls allowed frequencies

Who Controls the Software?

Sadly, typically the chipset vendors Most wifi drivers in linux require binary

firmware This firmware controls regulatory

compliance as well as purposing

What can we do?

Fortunately, most linux users don’t like closed source binaries

For many reasons, fully open sourced drivers are being developed

As these drivers become stable, we can start to play

Let’s Play…

Madwifi-ng is driven by a binary HAL Ath5k is the next gen fully open source

driver Kugutsumen released a patch for

“DEBUG” regdomain Allows for all *officially* supported

channels to be tuned to

Fun Comments in ath5k /* Set this to 1 to disable regulatory

domain restrictions for channel tests. * WARNING: This is for debuging only

and has side effects (eg. scan takes too * long and results timeouts). It's also

illegal to tune to some of the * supported frequencies in some

countries, so use this at your own risk, * you've been warned. */

Comments (cont)

/* * XXX The tranceiver supports frequencies from 4920 to 6100GHz * XXX and from 2312 to 2732GHz. There are problems with the * XXX current ieee80211 implementation because the IEEE * XXX channel mapping does not support negative channel * XXX numbers (2312MHz is channel -19). Of course, this * XXX doesn't matter because these channels are out of range * XXX but some regulation domains like MKK (Japan) will * XXX support frequencies somewhere around 4.8GHz. */

New Toys

Yesterday .11b/g 2412-2462 (US) .11a 5180-5320, 5745-5825 (US)

Today .11b/g 2192-2732 (DEBUG) .11a 4800-6000 (DEBUG)

What is on these new freq?

2180.000 - 2200.000 Fixed Point-to-point (n-p)2200.000 - 2290.000 DoD2300.000 - 2310.000 Amateur2390.000 - 2450.000 Amateur2450.000 - 2500.000 Radio location2500.000 - 2535.000 Fixed SAT2500.000 - 2690.000 Fixed Point-to-point (n-p), Instructional TV2655.000 - 2690.000 Fixed SAT2690.000 - 2700.000 Radio Astronomy2700.000 - 2900.000 DoD

Freq (cont)

4400.000 - 4990.000 DoD4990.000 - 5000.000 Meteo - Radio Astronomy5250.000 - 5650.000 Radio Location - Coastal Radar5460.000 - 5470.000 Radio Nav - General5470.000 - 5650.000 Meteo - Ground-based Radar5650.000 - 5925.000 Amateur5800.000 ISM5925.000 - 6425.000 Common Carrier and Fixed SAT

Spectrum Analyzer

Fully tested frequenciesSadly they wouldn’t let me borrow the SA

Warning: This may differ from card to card I’ve already lost a few wifi cards…

Limitations

Many real licensed implementations are broken Card reports channel 1 but is actually on

4920MHz This is done to make is easy to use existing

drivers This breaks many open source applications

Airodump-ng

Airodump-ng now supports a list of frequencies to scan rather than channels

Only channels are shown in display, may be wrong

Strips vital header information off of packet so data saved from extended channels is useless

Kismet

At time of writing is unable to handle most of the extended channels

Displays channels not frequencies Does save usable pcap files*

Improvement Needed

Sniffers are too trusting, they believe what they see

Never intended to deal with oddly broken implementations such as channel number fudging

Sniffers need to be improved to report more reality, and less assumptions

Improvements made!

After this talk was submitted, changes started happening

Kismet-newcore fully supports fun channels

Displays frequencies that packets are received on

Airodump-ng updates are being made now for release soon

Final Thoughts

Remember everyone here is a white hat Please use your new found knowledge for

good not evil In the United States it is LEGAL to monitor

all radio frequencies Have fun…

WEP cloaking Old hardware like wireless barcode

scanners

Insert chaff in the air to fool cracking tools

Good idea butUse half bandwidth => 300kb/sec with 11MbitSometimes packets doesn’t need to be filtered

to be cracked

How to break it? No public documentation => analyze capture

files Every data packet is cloaked (at least packets

from the AP protected) Cloaked Packet size is the same as the original

packet Plays with Sequence Numbers. In most cases,

not the same as the original packet (cloaked SN = original +2 to -2)

Only data packets are cloaked (at least type 2, subtype 0)

Signal is not the same as the access point

Implementation No idea of the implementation => don’t

care about key used by the sensor (if any) or data used in cloaked packets (real or fake).

Apply filters to remove cloaked packetsSignalSequence numbersBase analysis on packets know not to be

cloakedCombine filters in a different order

Implementation We know that all management and control frames

are uncloaked. Base filter:

If any packet with an unknown status has the same SN as one of the uncloaked packets then it’s cloaked

Signal filter: Get the average signal from uncloaked packets Allow a small margin of error Packets outside the margin should be cloaked

Implementation

Code release soon, check the subversion.

Thanks

Updated Slide Presentation can be found at: http://www.aircrack-ng.org/defcon16.ppt

Bibliographyhttp://www.willhackforsushi.com/FreeRADIUS-WPE.

htmlWe will complete this and post this weekend