LNCS 7322 - How to Break EAP-MD5

How to Break EAP-MD5
Fanbao Liu and Tao Xie
School of Computer, National University of Defense Technology, Changsha, 410073,
Hunan, P.R. China
liufanbao@gmail.com
Abstract. We propose an efficient attack to recover the passwords, used
to authenticate the peer by EAP-MD5, in the IEEE 802.1X network.
First, we recover the length of the used password through a method
called length recovery attack by on-line queries. Second, we crack the
known length password using a rainbow table pre-computed with a fixed
challenge, which can be done efficiently with great probability through
off-line computations. This kind of attack can also be implemented successfully even if the underlying hash function MD5 is replaced with SHA1 or even SHA-512.
Keywords: EAP-MD5, IEEE 802.1X, Challenge and Response, Length
Recovery, Password Cracking, Rainbow Table.
1
Introduction
IEEE 802.1X [6] is an IEEE Standard for port-based Network Access Control, which provides an authentication mechanism to devices wishing to attach
to a Local Area Network (LAN) or Wireless Local Area Network (WLAN).
IEEE 802.1X defines the encapsulation of the Extensible Authentication Protocol (EAP) [4] over IEEE 802 known as “EAP over LAN” (EAPoL). IEEE 802.1X
authentication involves three parties: a peer, an authenticator and an authentication server. The peer is a client device that wishes to attach to the LAN/WLAN.
The authenticator is a network device, such as a Wireless Access Point (WAP),
and the authentication server is typically a host running software supporting
the Remote Authentication Dial In User Service (RADIUS) and EAP protocols. EAP-MD5 [4], first used as Challenge-Handshake Authentication Protocol
(CHAP) [13] in Point-to-Point Protocol (PPP) [12,9] network for peer authentication, is a hash-based challenge and response protocol. EAP-MD5 applies the
secret prefix approach of hashing [14] to generate response R = MD5(P ||C),
where P is the shared password and C is the challenge.
In EAP-MD5 based wireless network, the peer provides user name and the
shared password, which is further hashed after appending the challenge sent
by the authentication server, to the authenticator to get right to access the
network. The authenticator forwards the peer’s credentials to the authentication
server, only if the server determines the peer’s credentials are valid, and then
the authenticator will allow the peer to access the network.
I. Askoxylakis, H.C. Pöhls, and J. Posegga (Eds.): WISTP 2012, LNCS 7322, pp. 49–57, 2012.
c IFIP International Federation for Information Processing 2012
50
F. Liu and T. Xie
EAP-MD5 provides the peer authentication but without mutual authentication, it is prone to attacks as man-in-the-middle (MITM). The shared password
between the peer and the authentication server is always chosen by the peer,
who prefers to choose a simple and memorable one, as a result, EAP-MD5 is
also vulnerable to dictionary attack just like other systems with human memorable passwords.
Our Contributions. We firstly propose an efficient method to break EAP-MD5
by efficiently cracking the shared password.
Based on the length recovery attack to MD4 [15], we apply this kind of attack
after minor modification to recover the length of password used in EAP-MD5,
with which we can reduce most of the extra costs of length guessing during
password cracking.
Further, based on the rainbow table and the chosen challenge attack, we
generate a series of rainbow tables using a fixed challenge combined with all
password candidates with fixed lengths.
We look up the received response from the corresponding rainbow table to
recover the known length password.
This paper is organized as follows. We introduce some preliminaries and background in section two. We present some related work about EAP-MD5 and
password cracking in second three. We introduce an efficient password cracking
method to EAP-MD5 in detail in section four. We conclude the paper in the last
section.
2
2.1
Preliminaries
Brief Description of MD5
MD5 [11,17,7,8] is a typical Merkle-Damgård structure hash function, which
takes a variable-length message M as input and outputs a 128-bit hash value.
M is first padded to be multiples of 512 bits, a ‘1’ added at the tail of M , followed
by ‘0’s until the bit length becomes 448 on modulo 512, finally, the length of
the unpadded message M is inserted to the last 64 bits. Let t = |M |/512, the
padded M is further divided into chunks of 512-bit blocks (M0 , M1 , . . . , Mt−1 ).
The MD5 compression function (CF ) takes Mi and a 128-bit chaining variable
Hi , initialized to H0 , as input, and outputs Hi+1 . For example, H1 = CFH0 (M0 ),
and the hash result of MD5 is computed as (1).
MD5(M ) = Ht = CFHt−1 (Mt−1 )
(1)
However, we omit the details of CF, since it is irrelevan t to this paper, for more
details please refer [11].
Padding Rule. For two arbitrary distinct messages M and M , if their lengths
|M | = |M |, then, the padding bits of these two messages are just the same. Since
How to Break EAP-MD5
51
MD5, MD4 and SHA-1 et al. share the same padding procedure, the padding
rule is also applicable to them.
Extension Attack. For an unknown message M and the corresponding hash
R = MD5(M ). With the known length of M , compute the padding bits pad for
M , then for arbitrary message x, compute the padding bits pad1 for M ||pad||x,
then we can generate the hash value for M ||pad||x from computing CFR (x||pad1),
without any knowledge of M except its length.
2.2
EAP-MD5
Extensible Authentication Protocol (EAP) [4] is an authentication framework
frequently used in 802.1X wireless networks and PPP connections. EAP provides
some common functions and negotiation of authentication methods called EAP
methods, such as EAP-MD5.
EAP-MD5 [4] is analogous to CHAP [13]. EAP-MD5 differs from other EAP
methods in that it only provides peer authentication.
The authentication of EAP-MD5 in 802.1X network involves three parts, the
peer, the wireless access point (WAP, the authenticator) and the RADIUS server
(the authentication server). The peer first sends an “EAPoL-start” packet to
the WAP, and after then the WAP replies an “EAP-Request/ID” to ask for the
peer’s ID. After the WAP and RADIUS server both receive the peer’s ID, the
WAP forwards the EAP-MD5 random challenge string including a 1-byte id and
a variable length C, sent by the RADIUS server, to the peer. If the response
R sent by the peer is valid, then the peer is authorized to use the network,
otherwise, reject. At random interval, the server redoes the authentication, by
sending different challenges, to limit the time of exposure to any single attack.
The peer authentication process of EAP-MD5 in IEEE 802.1X is shown in
Fig. 1.
Fig. 1. The peer authentication of EAP-MD5
The response R of EAP-MD5 is calculated with MD5 using a 1-byte session
identifier id and a challenge C both sent by the authentication server, combined
52
F. Liu and T. Xie
with the shared password P . The calculation is shown in (2), where || means
concatenation.
R = MD5(id||P ||C)
(2)
With randomly chosen id and C, EAP-MD5 provides protection against playback
attack launched by malicious peer.
3
3.1
Related Work
EAP-MD5-Pass Attack
In 2008, Wright demonstrated that a program utility eapmd5pass [16] could
crack the passwords used in EAP-MD5 of the wireless network. The utility includes a eavesdropping and a dictionary attack tools against the EAP-MD5
protocol. However, the successful ratio of eapmd5pass totally depends on the
used dictionary and the complexity of the password.
3.2
Rainbow Table
Rainbow tables contain a connection between a hash value and its corresponding
password P , for passwords hashed by H(P ), where H is a hash function. By
having this connection pre-calculated for all possible hash values, a quick search
through the table for a desired hash value can reveal the password.
In 1980, Hellman [5] proposed a time-memory trade-off when applying a cryptanalytic attack. Hellman used hash chains, which is time consuming, to decrease
the memory requirements. In the way of hash chain, only the first and the last
elements of the chain are stored in the memory, which saves memory at the
cost of extra cryptanalysis time. Further, additional memory trade-off can be
achieved by reducing the hash values to “keys” with the help of a reduction
function. Instead of using a single reduction function R, Oechslin [10] suggested
using a sequence of related reduction functions R1 up to Rk−1 , in 2003. Oechslin
called the new chains rainbow chains, and the tables containing these chains
rainbow tables. Since then, this method has been widely used and implemented
in many popular password recovery tools [2,3,1], to break passwords hashed by
the scheme of h(P ).
4
Breaking EAP-MD5
Since MD5 and MD4 share the same padding procedure, the length recovery
attack to MD4(P ||C) [15] is also applicable to EAP-MD5. After the length of
the used password is recovered, we search the corresponding rainbow table for
the received response, which is generated by using a pre-chosen challenge. So, we
divide the password cracking of EAP-MD5 into two parts, the first is to recover
the length of the used password, and the second is to crack the password with
already known length, using rainbow table technologies, based on fixed length
passwords, and pre-chosen challenges.
How to Break EAP-MD5
4.1
53
On-Line Length Recovery Attack to EAP-MD5
In the wireless network, we can easily collect enough information, by eavesdropping, about the access point and the peer, such as a pair of (C, R), id, the MAC
addresses of both, the user name and SSID of the access point, et al. We can generate faked packets, through masquerading as the access point, and send them
to the peer; the peer will take it for granted that the packets are sent from the
very access point, for he doesn’t authenticate the access point. We omit the role
of the authentication server, since it only interacts with the access point WAP.
We implement the length recovery attack to EAP-MD5 with pre-chosen challenge by masquerading as the access point (the authenticator), in the wireless
network shown as follows.
1. Pre-choose a fixed id and a fixed challenge C.
2. Eavesdrop from the network and find a communicating pair, a peer and a
server (the authenticator, access point). Collect information about them.
3. Masquerade as the server, send the pre-chosen id and C to the peer. Get
the corresponding response R = MD5(id||P ||C) from the peer, and reply an
affirmative signal.
4. Initiate the length l of the password P with the guessing l = 1 character.
5. Generate the padding pad for id||P ||C with the guessing length l of the
password.
6. Generate C = C||pad||r, where r is a randomly generated string, and send
C with the known id to the peer by masquerading as the access point.
7. Receive R from the peer, reply an affirmative signal. Generate pad1 for C and check if R = CFR (r||pad1) holds. If so, l is the very length. Otherwise,
set l = l + 1, go to step 5.
Packet Injection. We send the faked packet with the challenge information
C, by masquerading as the access point. Since the peer doesn’t authenticate
the access point S in EAP-MD5, it will responds any packet from the pretended
server S . Both S and S will receive response R . Since S doesn’t send C , it will
silently discard the response R . We let the pretending S reply an affirmative
signal to the peer to end this verification session.
To recover the password length l, we need to send l + 1 faked packets with
different challenges and l + 1 packets of affirmation. The time of the length
recovery attack to EAP-MD5 is negligible, since the server has full control over
the time intervals to send challenge to verify the peer, and the peer will reply
any challenge at any time from the claimed server.
4.2
Off-Line Fixed Length Password Cracking with Password
Pre-computation
Passwords Pre-computation. We can utilize the security hole that challenges
can be sent by unauthenticated server, to launch a password pre-computation
cracking. The overall strategy is shown as follows.
54
F. Liu and T. Xie
1. Passwords pre-computation.
(a) We first choose a string, satisfying all the restrictions placed on the EAPMD5 challenges, as a fixed challenge C. And choose a byte as the fixed
id.
(b) Set the length of password candidates be l = 5.1
(c) Use the fixed challenge C and the fixed id, we compute the corresponding
hash values of all possible passwords P s with fixed length l in the form
of MD5(id||P ||C), and resort those results in a rainbow table2 .
(d) Check if l = 8 holds, if so, go to step 2a. Otherwise, set l = l + 1, go to
step 1c.
2. Break passwords in EAP-MD5.
(a) Launch an on-line length recovery attack with the fixed challenge C and
id. We can immediately get the very length of the user’s password, which
is discussed in last subsection.
(b) We look up in the corresponding pre-computed rainbow table, generated
in the phase of passwords pre-computation, for the first received response
R replying from the peer, according to the already recovered password
length. If it hits, we get the peer’s very password.
Given a simple example of one password pre-computation attack to EAP-MD5,
we choose a string “<jantie lee@mukouren.com>” as a fixed challenge C. We also
choose “x” as the fixed id. There are 95 printable characters to be used as password candidates in total, including digits, punctuations and letters with lower
and upper cases. Then for a password with exact six characters, there may be
956 ≈239 password candidates. We compute the hash results of R=MD5 (x||P||<
jantie lee@mukouren.com>), using all 239 passwords with six characters in the
brute force strategy, and get all corresponding “responses”, eventually, then resort these results into a rainbow table.
Finally, we listen to the network and find a communicating pair. We impersonate the server and send the pre-chosen id “x” and the challenge string
“<jantie lee@mukouren.com>” to the peer, for recovering his password’s length.
After the length of the peer’s password is recovered, we look up in the corresponding rainbow table or a common table for the first received response.
Obviously, if the peer uses a password with six characters, we can immediately
get the very answer, for we have pre-computed all possible passwords.
Response Checking. To check whether a given hash value is in the precomputed rainbow table, we first generate a chain from this value. For each
password that we encounter in the chain, we check to see if it is at the end of
the chain. When we find a match, we know that the hash is part of this chain.
Thus we generate the complete chain again and get the very password.
The Rainbow Table. We use the rainbow tables to pre-compute all hashes of
the possible passwords with fixed lengths, based on the pre-chosen challenge and
1
2
We can directly generate a table mapping all possible candidates whose lengths are
less than 5.
We do not “store” their corresponding hash results, since those results are reduced
by the reduction functions.
How to Break EAP-MD5
55
id. Here, we implement the rainbow table of passwords with six characters, which
takes about 3 days to be completed on an ordinary PC with Intel 2.G CPU. The
details of complexity of such rainbow table are shown in Table 1, where s means
second. We also evaluate the overall complexity of generating rainbow tables of
passwords with seven and eight characters, respectively. It seems infeasible to
complete the rainbow table of passwords with eight characters on an ordinary
PC. However, if utilize the parallelism computing with one thousand PCs, this
work can be done soon. More important, an attacker of EAP-MD5 may only want
to use the network as a legal user, hence, he needs to break a peer’s passwords
less than eight characters.
Table 1. The Complexity of Generating the Rainbow Table of MD5(id||P ||C)
Char Set Characters Passwords Space Rainbow Table Size Hit Time
95
8
958 ≈252
576 GB
1108 s
95
7
957 ≈245
64 GB
829 s
95
6
956 ≈239
9 GB
18 s
Overall Complexity. The overall complexity of such attack consists of three
parts, the off-line pre-computation, the on-line query and the off-line table looking up. The time of the on-line query is negligible, since we have full control
over the time intervals to challenge the peer. The time of the off-line table
looking up varies from seconds to minutes. The complexity of the rainbow table pre-computation depends on the chosen length of the passwords and the
pre-chosen challenge. Since EAP-MD5 employs human-memorable passwords, it
means that we can compute all of the passwords within eight characters to cover
most of the ordinary using passwords. Moreover, we point out that once the
pre-computations are done, thus rainbow tables can always be reused for this
kind of attack. In [16], to crack a password, the repeated dictionary attack must
be performed each time. Hence, the attack in [16] is not a reusable one. More
important, we will not try to crack passwords with more characters than these
in our pre-computed rainbow tables, hence, our attack is a clever one.
Nowadays, the computation technology is dramatically improving, which includes the memory and storage enlargement, CPU processing speed increasing,
and the network computing strengthening. Parallel computing allows anybody
owning a modern personal computer in a network to break such cryptographic
systems, which were believed to be secure.
5
Conclusion
In this paper, we propose an efficient and universal way to break EAP-MD5.
We first recover the length of the used password by on-line queries, and then
56
F. Liu and T. Xie
we apply the advantage of rainbow table to crack the known length password
off-line. Though the 802.1X wireless network may not use this authentication
method any more, the operating systems supporting EAP-MD5 are vulnerable
to this kind of attack. This attack is also applicable even the underlying hash
function of EAP-MD5 is replaced with SHA-1 or even SHA-2, since these hash
functions share the same padding rule and have the same weakness.
Acknowledgement. We thank the anonymous reviewers for their valuable
comments. This work was partially supported by the program “Core Electronic
Devices, High-end General Purpose Chips and Basic Software Products” in
China (No. 2010ZX01037-001-001), and supported by the 973 program of China
under contract 2007CB311202, and by National Science Foundation of China
through the 61070228 project.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
OphCrack, http://ophcrack.sourceforge.net/
Rainbow Tables, http://rainbowtables.shmoo.com/
RainbowCrack, http://project-rainbowcrack.com/
Aboba, B., Blunk, L., Vollbrecht, J., Carlson, J., Levkowetz, H.: Extensible
Authentication Protocol (EAP). RFC 3748 (Proposed Standard) (June 2004),
http://www.ietf.org/rfc/rfc3748.txt, (updated by RFC 5247)
Hellman, M.: A cryptanalytic time-memory trade-off. IEEE Transactions on Information Theory 26(4), 401–406 (1980)
IEEE: IEEE 802.1x: IEEE Standards for Local and Metropolitan Area Networks:
Port based Network Access Control (2001)
Liu, F.: On the security of digest access authentication. In: Proc. 14th IEEE Int.
Conf. on Computational Science and Engineering, CSE 2011 and 11th Int. Symp.
on Pervasive Systems, Algorithms, and Networks, I-SPAN 2011 and 10th IEEE
Int. Conf. on IUCC 2011, pp. 427–434 (2011)
Liu, F., Liu, Y., Xie, T.: Fast Password Recovery Attack: Application to APOP.
Cryptology ePrint Archive, Report 2011/248 (2011), http://eprint.iacr.org/
Liu, F., Xie, T., Feng, Y., Feng, D.: On the Security of PPPoE Network. Security
and Communication Networks, p. n/a (2012),
http://dx.doi.org/10.1002/sec.512
Oechslin, P.: Making a Faster Cryptanalytic Time-Memory Trade-Off. In: Boneh,
D. (ed.) CRYPTO 2003. LNCS, vol. 2729, pp. 617–630. Springer, Heidelberg (2003)
Rivest, R.: The MD5 Message-Digest Algorithm. RFC 1321 (Informational) (April
1992),
http://www.ietf.org/rfc/rfc1321.txt (updated by RFC 6151)
Simpson, W.: The Point-to-Point Protocol (PPP). RFC 1661 (Standard) (July
1994), http://www.ietf.org/rfc/rfc1661.txt (updated by RFC 2153)
Simpson, W.: PPP Challenge Handshake Authentication Protocol (CHAP). RFC
1994 (Draft Standard) (August 1996), http://www.ietf.org/rfc/rfc1994.txt
(updated by RFC 2484)
How to Break EAP-MD5
57
14. Tsudik, G.: Message authentication with one-way hash functions. SIGCOMM Comput. Commun. Rev. 22, 29–38 (1992)
15. Wang, L., Ohta, K., Kunihiro, N.: Password recovery attack on authentication protocol MD4 (Password||Challenge). In: Proceedings of the 2008 ACM Symposium
on Information, Computer and Communications Security, ASIACCS 2008, pp. 3–9.
ACM, New York (2008)
16. Wright, J.: EAPMD5PASS-AUDITING EAP-MD5 (2003),
http://www.willhackforsushi.com
17. Xie, T., Liu, F., Feng, D.: Could The 1-MSB Input Difference Be The Fastest
Collision Attack For MD5?. Cryptology ePrint Archive, Report 2008/391 (2008),
http://eprint.iacr.org/