23c3 Bluetooth hacking revisited

Bluetooth Hacking revisited
+
Kevin Finistere & Thierry
Zoller
23C3 - 2006

Bluetooth – Please just turn it off
Turn off your BT please,
,no really.
Yeah

The Goal of this Talk ?
 The Goal of this talk is not to:
 Build myths
 Show off – and not show how
 The Goal of this talk is to :
 Raise awareness
 Make risks (more) transparent
 Paradigm Shift – Bluetooth is not only for toys
 Show cool stuff…

What are we talking about today ?
 [ 0x00 ] – Introduction : What is Bluetooth ?
 Sorry this is required. Crash course..
 [ 0x01 ] – Get ready to rumble : Extending the Range
 Extending the range of Bluetooth devices
 Building automated reconnaissance and attack devices
 Bluetooth War driving (GPS, 360° Camera)
 [ 0x02 ] – Implementation issues : Bypassing Security
 Attacking drivers, Attacking applications
 Owning Bluetooth VNC style
 Attacking Internal Networks and pivoting
 Bluetooth Pin to Bluetooth Passkey
 [ 0x03 ] – Protocol/Specification issues : Ceci n’est pas une pipe
 Cracking the Pin and the Link-key (BTCrack)
 Key management, 8 bit Encryption, Collisions
 Tracking the un-trackeable
 Anti-Brute-forcing
 Random Number generators from hell

[ 0x00 ] Introduction
 Bluetooth - a few tidbits:
 Operates on the non-regulated ISM band : 2,4Ghz
 In general 79 Channels (Except France, Spain)
 Frequency Hopping (3200/sec, 1600/sec)
 Complete Framework with profiles and layers of protocols
 1 Billionth BT device sold in November 2006 (source SIG)
 Goals : Least cost cable replacement, low power usage

 The foundation – Protocol Stack
Hardware
Software
Redfang – read_remote_name()
L2ping

 “Typical” Bluetooth Scenario
Inquiry
Inquiry response
Paging (FHS)
Link establishment
Discovers
Profiles
Bluetooth
Access Point

 Inquiry - First Contact
 Predefined Hopping sequence
 FHS same for all devices
 Pass Paging parameters during Inquiry stage

 Paging - Frequency Hopping Synchronization
 Slaves always sync to the Master
 Paging initialisation :
 Slaves hop 1 Channel/sec
 Master hops 3200 times/sec
 Paging
 Both hop 1600 times/sec
 Piconet agrees to a Sequence based on parts
of the BD_ADDR and Clock-offset of the master.
(Nice fingerprint by the way)
 FH is the reason you can not easily sniff BT traffic. You have to sync to the
Master (or use a Spectral Analyzer and reconstruct afterwards – Good luck)

 The Bluetooth Profiles
 Represent a group and defines mandatory options
 Prevent compatibility issues, modular approach to BT extensions
 Vertical representation of BT layer usage, handled through SDP
Object Push Profile

 Different Bluetooth modes
 Discoverable modes
 Discoverable :
Sends inquiry responses to all inquiries.
 Limited discoverable:
Visible for a certain period of time (Implementation bug: Sony Ericsson T60..)
 Non-Discoverable:
Never answers an inquiry scan (in theory)
 Pairing modes :
 Non-pairable mode :
Rejects every pairing request (LMP_not_accepted) (Implementation bug: Plantronic
Headset..)
 Pairable mode :
Will pair up-on request

[ 0x01 ] Get ready to rumble
 Extending the Range

 Long Distance - Datasets
 Antrum Lake, water reflection
guarantees longer ranges.
 788 Meters
 An old Man stole my phone
during this test! I tracked
him with the yagi.

 Optimizing for Penetration (1)
 Integrated Linksys Dongle
 Integrated USB Cable
 Metal Parabola
 10 * Zoom
 Laser (to be done)
 Experiment : Went through a building found the device on
the other side IN another building.

 Optimizing for Penetration (2)
 Bundling (Parabola)
 Higher penetration through walls
 Glass is your friend
 On board embedded device. (NSLU2)
 Autonomous scan and attack toolkit
 automatically scans
 may attack devices
 saves all the results

 PerimeterWatch – Bluetooth Wardriving
 Perl Script by KF
 Searches Bluetooth Devices
 Takes 360° pictures
 GPS coordinates

[ 0x02 ] Implementation bugs
 Implementation Bugs – Bypassing security

 Menu du Jour :
 Eavesdropping on Laptops/Desktops
 Remotely controlling workstations
 Car Whisperer NG
 Owning internal Networks over Bluetooth
 Linkkey theft and abuse
 Widcomm Overflows
(Broadcom merger leaves lots of vuln users that can not patch) BTW
3.0.1.905 (../ attacks) and up to BTW 1.4.2.10 has overflows

 Bluetooth PIN is really a Bluetooth Passkey
 Did you know ? A Bluetooth “Pin” can be more than digits…
 Not aware of any implementation, all use just digits
 Uses UTF8
 Max 16, UTF8 char may take some off
 Example :
 It’s like implementing NTLM with digits only….
 BTCrack would a lot more time if this would be “correctly” implemented
0xC3 0x84 0x72 0x6c 0x69 0x63 0x68Ärlich
0x30 0x31 0x032 0x330123
BT handlesUser enters

 CarWhisperer – Martin Herfurt
 Listen and Record Conversations
 Not that new, but what’s new :
 Works against Workstations
Example : Widcomm < BTW 4.0.1.1500 (No Pincode)
 Kevin did a real-time patch for it
 Remove the Class ID check
 Root Cause :
Paring mode, discoverable, hard coded Pin.

 HidAttack - Owning Bluetooth VNC Style
 HID = Human Interface Device
 Requires 2 HID (PSM) endpoints to act
as server
 2 implementations :
 Keyboard connects to the HID server
 HID server connects to the Keyboard
 You can control the Mouse and Keyboard HID just as you were in
front of the PC.
 Discovered by Collin Mulliner , fixed in hidd Bluez <2.25, Widcomm,
Toshiba not really tested. Yours?
 Code release today : www.mulliner.org/bluetooth/hidattack01.tar.gz
 Thanks Collin !

 Demo - Owning internal networks
 Apple
 OSX 10.3 Tiger
 OSX 10.4 Jaguar
Vanilla, delayed release
 Windows
 Widcomm, Toshiba,
Bluesoil, others ?
 Pocket PC
 Kevin: Apple asked me to not tell 10.4 was shipping vulnerable
 OSX 10.3.9 patched, OSX 10.4 shipped vulnerable patched a month
after OSX 10.3.9

 Demo – Remote Root over BT
 Vulnerability shown :
_Directory Traversal_ in un-authenticated
Obexserver (Patched)
 Cause :
User input validated client-side (except btftp)
 ObexFTP server directory traversal exploit & malicious InputManager & local
root exploit = remote login tty over rfcomm = 0WNAGE
 Was possible on Windows and Pocket PC and everything that has Toshiba or
Broadcom & Widcomm (estimate 90%), and most probably others too. But we
choose a MAC, because…we can.
 Points are :
- Macs are NOT invulnerable (far from that) - You can own internal networks
over Bluetooth

 Windows Widcomm - Buffer overflows

 Windows Widcomm - Buffer overflows
 Vulnerable versions known to us :
 Widcomm Stack up to 3.x is vuln
 Widcomm BTStackServer 1.4.2 .10
 Widcomm BTStackServer 1.3.2 .7
 Widcomm Bluetooth Communication Software 1.4.1 .03
 HP IPAQ 2215
 HP IPAQ 5450

[ 0x03 ] Protocol issues
They are just
implementation
Bugs*
*This is supposed to be a joke

 Menu du Jour :
 Why the Pin is not that important
 Unit Keys
 How to find non discoverable devices
 Random Number generators that may be from Hell
 Link Keys
 Reconstructing them
 Abusing them
 Re-force Pairing, Corruption
 Denial of Service

 The PIN is not really that useful
 The link key is !
 Here’s why :
 Pairing mode required for PIN
 The LK is enough to authenticate
 Encryption (E0) calculated from
the LK
 We can authenticate against both
sides with the same key
 Protocol 1.2 Authentication :

 Unit keys
 Generated by the device when starting up
 Based on a PRNG that may come from hell
 Permanently saved and cannot be changed
 Only has one key
 Problem :
 The SIG clearly does not recommend it’s use.
A B
Step1
A C
Step2

 How to find nondiscoverable devices passively
 From the man himself: Joshua Wright
 We knew read_remote_name(), now l2ping.
 Target : BD_Addr : 48-bit
4. Sniff on a preset channel and wait for devices to hop by , capture
the Bluetooth Preamble, extract the cannel access code (which
is based on 24 bits of the BD_addr)
5. Extract Error Correction field (baseband header – CRC 10bit
field)
6. Assume the first 8 bits 00
7. Brute force the remaining: 8bits
00:11:9F:C5:F1:AE

[ 0x03 ] Specification issues
 Random Number Generators from Hell
 Specification is not very clear about what to achieve or how to achieve
it
 The specification reads :
Each device has a pseudo-random number generator. Pseudo-random
numbers are used for many purposes within the security functions − for
instance, for the challenge-response scheme, for generating authentication and
encryption keys, etc.
Within this specification, the requirements placed on the random
numbers used are non-repeating and randomly generated
For example, a non-repeating value could be the output of a counter that
is unlikely to repeat during the lifetime of the authentication key, or a
date/time stamp.

[ 0x03 ] Specification issues
 Random Number Generators from Hell
 Remember the Clock inside each Device ?
 Remember that we can get the clock-offset with an simple non-authenticated
inquiry ?
 RND do not look very random, had no time left to investigate fully, looks
horrible.
 They don’t trust it themselves :
The reason for using the output of and not directly
choosing a random number as the key*, is to avoid
possible problems with degraded randomness due
to a poor implementation of the random number
generator within the device.
*What a great idea that would have been…

 Introducing BTCrack
 First presented at Hack.lu 2006
 Released for 23C3
 Cracks PIN and Link key
 Requires values from a Pairing sniff
 Imports CVS Data
Available for download here now:
http://www.nruns.com/security_tools.php

 History
 Ollie Whitehouse - 2003
 Presents weaknesses of the pairing process and how it may be used
crack the PIN
 Shaked and Wool - 2005
 Implemented and optimised the attack
 Found ways to re-initiate pairing
 Thierry Zoller – 2006
 Win32 implementation, first public release
 Tremendous help from somebody that will recognize himself

 Speed - Dual-Core P4-2GHZ
 BTcrack v0.3 (Hack.lu)
 22.000 keys per second
 BTcrack v0.5
 BTcrack v1.0
 Thanks to Eric Sesterhenn
 Optimised for caching,
cleaning code, static funcs,
removing Junk
 ICC
Results :
• 4 digit pin : 0.035 seconds
• 9 digit pin : 1318 seconds

 BT Crack – Behind the scenes (1)
Step1
Generates (RAND)
K = E22(RAND, PIN, PIN_LEN)
Device A Device B
Step1
K = E22(RAND, PIN, PIN_LEN)
Rand
Step2
Generates (RANDA)
CA = RANDA xor K
Step2
Generates (RANDB)
CB = RANDB xor K
CA
CB
Step3
RANDB=CA xor K
LKA=E21(RANDA, ADDRA)
LKB=E21(RANDB,ADDRB)
LKAB=LKA xor LKB
Step3
RANDB=CA xor K
LKA=E21(RANDA, ADDRA)
LKB=E21(RANDB,ADDRB)
LKAB=LKA xor LKB
Step4
SRESA =
E1(CH_RANDA,ADDRB,LKAB)
Step4
SRESB =
E1(CH_RANDA,ADDRB,LKAB)
CH_RANDA
SRESB
Step5
SRESA = SRESB
E22 = Connection key
E21 = Device key

 BT Crack – Behind the scenes
Pin =-1;
Do
{
PIN++;
CR_K=E22(RAND, PIN, length(PIN));
CR_RANDA = CA xor CR_K;
CR_RANDB = CB xor CR_K;
CR_LKA = E21 (CR_RANDA, ADDRA);
CR_LKB = E21 (CR_RANDB, ADDRB);
CR_LKAB = CR_LKA xor CR_LKB;
CR_SRES = (CH_RAND, ADDRB, CR_LKAB);
}
while (CR_SRES == SRES)
 Right : Shaked and Wool logic
 Top : Pseudo code by Tomasz Rybicki
Hackin9 04/2005

 BT Crack – Demo

 Link keys – What can I do with them ?
 Authenticated to both devices Master & Slave with the same link key
 Dump them from any Linux, Mac, Windows machine
 Create a encrypted hidden stealth channel, plant the linkkey
 You can decrypt encrypted traffic with the linkkey
 How to force repairing ?
 Shaked and Wool proposed:
 Injection of LMP_Not_Accepted spoofing the Master
 Before the master sends Au_rand, inject In_rand to the slave
 Before the master sends Au_rand, inject random SRES messages
 We propose :
 Use bdaddr to change the Bd_Addr to a member, connect to the master
with a unknown linkkey.

[ 0x04 ] Kick-Out
 Sooooo now we have :
 A quick and reliable way to get the BD_ADDR
 A way to crack the Pin and the keys
 What's left ?
 The sniffer. It costs around 13.000$, you can get it on eBay
sometimes for the 1/10 of the amount.
 Assignment : Go and make one for everybody.

[ 0x04 ] Kick-Out
 Things to Remember :
 Bluetooth might be a risk for your Company
 Risk assessment is rather complex
 Don’t accept every file you are being send, just click NO.
 Disable Bluetooth if not required
 Pair in “secure” places (SIG Recommendations)
 Don’t use Unit Keys
 Hold your Bluetooth vendor accountable for vulnerabilities
 Delete your pairings
 Use BT 2.0 and “Simple Paring”

23c3 Bluetooth hacking revisited

More Related Content

What's hot (20)

Similar to 23c3 Bluetooth hacking revisited (20)

More from Thierry Zoller (9)

Recently uploaded (20)

23c3 Bluetooth hacking revisited