This document is obsolete. Please
refer to RFC 9293.
Internet Engineering Task Force (IETF) F. Gont Request for Comments: 6528 SI6 Networks / UTN-FRH Obsoletes: 1948 S. Bellovin Updates: 793 Columbia University Category: Standards Track February 2012 ISSN: 2070-1721
Defending against Sequence Number Attacks
Abstract
This document specifies an algorithm for the generation of TCP Initial Sequence Numbers (ISNs), such that the chances of an off-path attacker guessing the sequence numbers in use by a target connection are reduced. This document revises (and formally obsoletes) RFC 1948, and takes the ISN generation algorithm originally proposed in that document to Standards Track, formally updating RFC 793.
Status of This Memo
This is an Internet Standards Track document.
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Further information on Internet Standards is available in Section 2 of RFC 5741.
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc6528.
Copyright Notice
Copyright (c) 2012 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Gont & Bellovin Standards Track [Page 1]
RFC 6528 Defending against Sequence Number Attacks February 2012
For a long time, the Internet has experienced a number of off-path attacks against TCP connections. These attacks have ranged from trust-relationship exploitation to denial-of-service attacks [CPNI-TCP]. Discussion of some of these attacks dates back to at least 1985, when Morris [Morris1985] described a form of attack based on guessing what sequence numbers TCP [RFC0793] will use for new connections between two known end-points.
In 1996, RFC 1948 [RFC1948] proposed an algorithm for the selection of TCP Initial Sequence Numbers (ISNs), such that the chances of an off-path attacker guessing valid sequence numbers are reduced. With the aforementioned algorithm, such attacks would remain possible if and only if the attacker already has the ability to perform "man-in- the-middle" attacks.
This document revises (and formally obsoletes) RFC 1948, and takes the ISN generation algorithm originally proposed in that document to Standards Track.
Section 2 provides a brief discussion of the requirements for a good ISN generation algorithm. Section 3 specifies a good ISN selection algorithm. Appendix A provides a discussion of the trust- relationship exploitation attacks that originally motivated the publication of RFC 1948 [RFC1948]. Finally, Appendix B lists the differences from RFC 1948 to this document.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].
Gont & Bellovin Standards Track [Page 2]
RFC 6528 Defending against Sequence Number Attacks February 2012
RFC 793 [RFC0793] suggests that the choice of the ISN of a connection is not arbitrary, but aims to reduce the chances of a stale segment from being accepted by a new incarnation of a previous connection. RFC 793 [RFC0793] suggests the use of a global 32-bit ISN generator that is incremented by 1 roughly every 4 microseconds.
It is interesting to note that, as a matter of fact, protection against stale segments from a previous incarnation of the connection is enforced by preventing the creation of a new incarnation of a previous connection before 2*MSL have passed since a segment corresponding to the old incarnation was last seen (where "MSL" is the "Maximum Segment Lifetime" [RFC0793]). This is accomplished by the TIME-WAIT state and TCP's "quiet time" concept (see Appendix B of [RFC1323]).
Based on the assumption that ISNs are monotonically increasing across connections, many stacks (e.g., 4.2BSD-derived) use the ISN of an incoming SYN segment to perform "heuristics" that enable the creation of a new incarnation of a connection while the previous incarnation is still in the TIME-WAIT state (see p. 945 of [Wright1994]). This avoids an interoperability problem that may arise when a node establishes connections to a specific TCP end-point at a high rate [Silbersack2005].
Unfortunately, the ISN generator described in [RFC0793] makes it trivial for an off-path attacker to predict the ISN that a TCP will use for new connections, thus allowing a variety of attacks against TCP connections [CPNI-TCP]. One of the possible attacks that takes advantage of weak sequence numbers was first described in [Morris1985], and its exploitation was widely publicized about 10 years later [Shimomura1995]. [CERT2001] and [USCERT2001] are advisories about the security implications of weak ISN generators. [Zalewski2001] and [Zalewski2002] contain a detailed analysis of ISN generators, and a survey of the algorithms in use by popular TCP implementations.
Simple random selection of the TCP ISNs would mitigate those attacks that require an attacker to guess valid sequence numbers. However, it would also break the 4.4BSD "heuristics" to accept a new incoming connection when there is a previous incarnation of that connection in the TIME-WAIT state [Silbersack2005].
We can prevent sequence number guessing attacks by giving each connection -- that is, each four-tuple of (localip, localport, remoteip, remoteport) -- a separate sequence number space. Within
Gont & Bellovin Standards Track [Page 3]
RFC 6528 Defending against Sequence Number Attacks February 2012
each space, the ISN is incremented according to [RFC0793]; however, there is no obvious relationship between the numbering in different spaces.
An obvious way to prevent sequence number guessing attacks while not breaking the 4.4BSD heuristics would be to perform a simple random selection of TCP ISNs while maintaining state for dead connections (e.g. changing the TCP state transition diagram so that both end- points of all connections go to TIME-WAIT state). That would work but would consume system memory to store the additional state. Instead, we propose an improvement to the TCP ISN generation algorithm that does not require TCP to keep state for all recently terminated connections.
3. Proposed Initial Sequence Number Generation Algorithm
TCP SHOULD generate its Initial Sequence Numbers with the expression:
ISN = M + F(localip, localport, remoteip, remoteport, secretkey)
where M is the 4 microsecond timer, and F() is a pseudorandom function (PRF) of the connection-id. F() MUST NOT be computable from the outside, or an attacker could still guess at sequence numbers from the ISN used for some other connection. The PRF could be implemented as a cryptographic hash of the concatenation of the connection-id and some secret data; MD5 [RFC1321] would be a good choice for the hash function.
The result of F() is no more secure than the secret key. If an attacker is aware of which cryptographic hash function is being used by the victim (which we should expect), and the attacker can obtain enough material (i.e., ISNs selected by the victim), the attacker may simply search the entire secret-key space to find matches. To protect against this, the secret key should be of a reasonable length. Key lengths of 128 bits should be adequate. The secret key can either be a true random number [RFC4086] or some per-host secret. A possible mechanism for protecting the secret key would be to change it on occasion. For example, the secret key could be changed whenever one of the following events occur:
o The system is being bootstrapped (e.g., the secret key could be a combination of some secret and the boot time of the machine).
o Some predefined/random time has expired.
o The secret key has been used sufficiently often that it should be regarded as insecure at that point.
Gont & Bellovin Standards Track [Page 4]
RFC 6528 Defending against Sequence Number Attacks February 2012
Note that changing the secret would change the ISN space used for reincarnated connections, and thus could cause the 4.4BSD heuristics to fail; to maintain safety, either dead connection state could be kept or a quiet time observed for two maximum segment lifetimes before such a change.
It should be noted that while there have been concerns about the security properties of MD5 [RFC6151], the algorithm specified in this document simply aims at reducing the chances of an off-path attacker guessing the ISN of a new connection, and thus in our threat model it is not worth the effort for an attacker to try to learn the secret key. Since MD5 is faster than other "stronger" alternatives, and is used in virtually all existing implementations of this algorithm, we consider that use of MD5 in the specified algorithm is acceptable. However, implementations should consider the trade-offs involved in using functions with stronger security properties, and employ them if it is deemed appropriate.
Good sequence numbers are not a replacement for cryptographic authentication, such as that provided by IPsec [RFC4301] or the TCP Authentication Option (TCP-AO) [RFC5925]. At best, they are a palliative measure.
If random numbers are used as the sole source of the secret, they MUST be chosen in accordance with the recommendations given in [RFC4086].
A security consideration that should be made about the algorithm proposed in this document is that it might allow an attacker to count the number of systems behind a Network Address Translator (NAT) [RFC3022]. Depending on the ISN generators implemented by each of the systems behind the NAT, an attacker might be able to count the number of systems behind a NAT by establishing a number of TCP connections (using the public address of the NAT) and identifying the number of different sequence number "spaces". [Gont2009] discusses how this and other information leakages at NATs could be mitigated.
An eavesdropper who can observe the initial messages for a connection can determine its sequence number state, and may still be able to launch sequence number guessing attacks by impersonating that connection. However, such an eavesdropper can also hijack existing connections [Joncheray1995], so the incremental threat is not that high. Still, since the offset between a fake connection and a given real connection will be more or less constant for the lifetime of the secret, it is important to ensure that attackers can never capture
Gont & Bellovin Standards Track [Page 5]
RFC 6528 Defending against Sequence Number Attacks February 2012
such packets. Typical attacks that could disclose them include both eavesdropping and the variety of routing attacks discussed in [Bellovin1989].
Off-path attacks against TCP connections require the attacker to guess or know the four-tuple (localip, localport, remoteip, remoteport) that identifies the target connection. TCP port number randomization [RFC6056] reduces the chances of an attacker of guessing such a four-tuple by obfuscating the selection of TCP ephemeral ports, therefore contributing to the mitigation of such attacks. [RFC6056] provides advice on the selection of TCP ephemeral ports, such that the overall protection of TCP connections against off-path attacks is improved.
[CPNI-TCP] contains a discussion of all the currently known attacks that require an attacker to know or be able to guess the TCP sequence numbers in use by the target connection.
Matt Blaze and Jim Ellis contributed some crucial ideas to RFC 1948, on which this document is based. Frank Kastenholz contributed constructive comments to that memo.
The authors of this document would like to thank (in chronological order) Alfred Hoenes, Lloyd Wood, Lars Eggert, Joe Touch, William Allen Simpson, Tim Shepard, Wesley Eddy, Anantha Ramaiah, and Ben Campbell for providing valuable comments on draft versions of this document.
Fernando Gont wishes to thank Jorge Oscar Gont, Nelida Garcia, and Guillermo Gont for their love and support, and Daniel Bellomo and Christian O'Flaherty for their support in his Internet engineering activities.
Fernando Gont's attendance to IETF meetings was supported by ISOC's "Fellowship to the IETF" program.
[CPNI-TCP] CPNI, "Security Assessment of the Transmission Control Protocol (TCP)", http://www.gont.com.ar/ papers/tn-03-09-security-assessment-TCP.pdf, 2009.
[Gont2009] Gont, F. and P. Srisuresh, "Security implications of Network Address Translators (NATs)", Work in Progress, October 2009.
[Joncheray1995] Joncheray, L., "A Simple Active Attack Against TCP", Proc. Fifth Usenix UNIX Security Symposium, 1995.
[Morris1985] Morris, R., "A Weakness in the 4.2BSD UNIX TCP/IP Software", CSTR 117, AT&T Bell Laboratories, Murray Hill, NJ, 1985.
[RFC0854] Postel, J. and J. Reynolds, "Telnet Protocol Specification", STD 8, RFC 854, May 1983.
[RFC1034] Mockapetris, P., "Domain names - concepts and facilities", STD 13, RFC 1034, November 1987.
Gont & Bellovin Standards Track [Page 7]
RFC 6528 Defending against Sequence Number Attacks February 2012
[RFC1948] Bellovin, S., "Defending Against Sequence Number Attacks", RFC 1948, May 1996.
[RFC3022] Srisuresh, P. and K. Egevang, "Traditional IP Network Address Translator (Traditional NAT)", RFC 3022, January 2001.
[RFC4120] Neuman, C., Yu, T., Hartman, S., and K. Raeburn, "The Kerberos Network Authentication Service (V5)", RFC 4120, July 2005.
[RFC4251] Ylonen, T. and C. Lonvick, "The Secure Shell (SSH) Protocol Architecture", RFC 4251, January 2006.
[RFC4301] Kent, S. and K. Seo, "Security Architecture for the Internet Protocol", RFC 4301, December 2005.
[RFC4954] Siemborski, R. and A. Melnikov, "SMTP Service Extension for Authentication", RFC 4954, July 2007.
[RFC5321] Klensin, J., "Simple Mail Transfer Protocol", RFC 5321, October 2008.
[RFC5925] Touch, J., Mankin, A., and R. Bonica, "The TCP Authentication Option", RFC 5925, June 2010.
[RFC5936] Lewis, E. and A. Hoenes, "DNS Zone Transfer Protocol (AXFR)", RFC 5936, June 2010.
[RFC6151] Turner, S. and L. Chen, "Updated Security Considerations for the MD5 Message-Digest and the HMAC-MD5 Algorithms", RFC 6151, March 2011.
[Shimomura1995] Shimomura, T., "Technical details of the attack described by Markoff in NYT", http://www.gont.com.ar/docs/post-shimomura- usenet.txt, Message posted in USENET's comp.security.misc newsgroup, Message-ID: <3g5gkl$5j1@ariel.sdsc.edu>, 1995.
[Silbersack2005] Silbersack, M., "Improving TCP/IP security through randomization without sacrificing interoperability", EuroBSDCon 2005 Conference.
Gont & Bellovin Standards Track [Page 8]
RFC 6528 Defending against Sequence Number Attacks February 2012
[USCERT2001] US-CERT, "US-CERT Vulnerability Note VU#498440: Multiple TCP/IP implementations may use statistically predictable initial sequence numbers", http://www.kb.cert.org/vuls/id/498440, 2001.
[Wright1994] Wright, G. and W. Stevens, "TCP/IP Illustrated, Volume 2: The Implementation", Addison-Wesley, 1994.
[Zalewski2002] Zalewski, M., "Strange Attractors and TCP/IP Sequence Number Analysis - One Year Later", http://lcamtuf.coredump.cx/newtcp/, 2002.
Gont & Bellovin Standards Track [Page 9]
RFC 6528 Defending against Sequence Number Attacks February 2012
Appendix A. Address-Based Trust-Relationship Exploitation Attacks
This section discusses the trust-relationship exploitation attack that originally motivated the publication of RFC 1948 [RFC1948]. It should be noted that while RFC 1948 focused its discussion of address-based trust-relationship exploitation attacks on Telnet [RFC0854] and the various UNIX "r" commands, both Telnet and the various "r" commands have since been largely replaced by secure counterparts (such as SSH [RFC4251]) for the purpose of remote login and remote command execution. Nevertheless, address-based trust relationships are still employed nowadays in some scenarios. For example, some SMTP [RFC5321] deployments still authenticate their users by means of their IP addresses, even when more appropriate authentication mechanisms are available [RFC4954]. Another example is the authentication of DNS secondary servers [RFC1034] by means of their IP addresses for allowing DNS zone transfers [RFC5936], or any other access control mechanism based on IP addresses.
In 1985, Morris [Morris1985] described a form of attack based on guessing what sequence numbers TCP [RFC0793] will use for new connections. Briefly, the attacker gags a host trusted by the target, impersonates the IP address of the trusted host when talking to the target, and completes the three-way handshake based on its guess at the next ISN to be used. An ordinary connection to the target is used to gather sequence number state information. This entire sequence, coupled with address-based authentication, allows the attacker to execute commands on the target host.
Clearly, the proper solution for these attacks is cryptographic authentication [RFC4301] [RFC4120] [RFC4251].
The following subsection provides technical details for the trust- relationship exploitation attack described by Morris [Morris1985].
In order to understand the particular case of sequence number guessing, one must look at the three-way handshake used in the TCP open sequence [RFC0793]. Suppose client machine A wants to talk to rsh server B. It sends the following message:
A->B: SYN, ISNa
That is, it sends a packet with the SYN ("synchronize sequence number") bit set and an initial sequence number ISNa.
Gont & Bellovin Standards Track [Page 10]
RFC 6528 Defending against Sequence Number Attacks February 2012
B replies with
B->A: SYN, ISNb, ACK(ISNa)
In addition to sending its own ISN, it acknowledges A's. Note that the actual numeric value ISNa must appear in the message.
A concludes the handshake by sending
A->B: ACK(ISNb)
RFC 793 [RFC0793] specifies that the 32-bit counter be incremented by 1 in the low-order position about every 4 microseconds. Instead, Berkeley-derived kernels traditionally incremented it by a constant every second, and by another constant for each new connection. Thus, if you opened a connection to a machine, you knew to a very high degree of confidence what sequence number it would use for its next connection. And therein lied the vulnerability.
The attacker X first opens a real connection to its target B -- say, to the mail port or the TCP echo port. This gives ISNb. It then impersonates A and sends
Ax->B: SYN, ISNx
where "Ax" denotes a packet sent by X pretending to be A.
B's response to X's original SYN (so to speak)
B->A: SYN, ISNb', ACK(ISNx)
goes to the legitimate A, about which more anon. X never sees that message but can still send
Ax->B: ACK(ISNb')
using the predicted value for ISNb'. If the guess is right -- and usually it will be, if the sequence numbers are weak -- B's rsh server thinks it has a legitimate connection with A, when in fact X is sending the packets. X can't see the output from this session, but it can execute commands as more or less any user -- and in that case, the game is over and X has won.
There is a minor difficulty here. If A sees B's message, it will realize that B is acknowledging something it never sent, and will send a RST packet in response to tear down the connection. However, an attacker could send the TCP segments containing the commands to be
Gont & Bellovin Standards Track [Page 11]
RFC 6528 Defending against Sequence Number Attacks February 2012
executed back-to-back with the segments required to establish the TCP connection, and thus by the time the connection is reset, the attacker has already won.
In the past, attackers exploited a common TCP implementation bug to prevent the connection from being reset (see subsection "A Common TCP Bug" in [RFC1948]). However, all TCP implementations that used to implement this bug have been fixed for a long time.