TCPM WG J. Touch Internet Draft USC/ISI Obsoletes: 2385 A. Mankin Intended status: Proposed Standard Johns Hopkins Univ. Expires: August 2009 R. Bonica Juniper Networks February 16, 2009 The TCP Authentication Option draft-ietf-tcpm-tcp-auth-opt-03.txt Status of this Memo This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html This Internet-Draft will expire on August 16, 2009. Copyright Notice Copyright (c) 2009 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 in effect on the date of publication of this document (http://trustee.ietf.org/license-info). Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Touch Expires August 16, 2009 [Page 1] Internet-Draft The TCP Simple Authentication Option February 2009 Abstract This document specifies the TCP Authentication Option (TCP-AO), which obsoletes the TCP MD5 Signature option of RFC-2385 (TCP MD5). TCP-AO specifies the use of stronger Message Authentication Codes (MACs), protects against replays even for long-lived TCP connections, and provides more details on the association of security with TCP connections than TCP MD5. TCP-AO is compatible with either static master key configuration or an external, out-of-band master key management mechanism; in either case, TCP-AO also protects connections when using the same master key across repeated instances of a connection, using connection keys derived from the master key. The result is intended to support current infrastructure uses of TCP MD5, such as to protect long-lived connections (as used, e.g., in BGP and LDP), and to support a larger set of MACs with minimal other system and operational changes. TCP-AO uses its own option identifier, even though used mutually exclusive of TCP MD5 on a given TCP connection. TCP-AO supports IPv6, and is fully compatible with the requirements for the replacement of TCP MD5. Table of Contents 1. Contributors...................................................3 2. Introduction...................................................3 2.1. Executive Summary.........................................4 2.2. Changes from Previous Versions............................5 2.2.1. New in draft-ietf-tcp-auth-opt-03....................6 2.2.2. New in draft-ietf-tcp-auth-opt-02....................6 2.2.3. New in draft-ietf-tcp-auth-opt-01....................7 2.2.4. New in draft-ietf-tcp-auth-opt-00....................8 2.2.5. New in draft-touch-tcp-simple-auth-03................9 2.2.6. New in draft-touch-tcp-simple-auth-02................9 2.2.7. New in draft-touch-tcp-simple-auth-01................9 3. Conventions used in this document.............................10 4. The TCP Authentication Option.................................10 4.1. Review of TCP MD5 Option.................................10 4.2. TCP-AO Option............................................11 5. Preventing replay attacks within long-lived connections.......14 6. Computing connection keys from TSAD entries...................16 7. Security Association Management...............................17 8. TCP-AO Interaction with TCP...................................21 8.1. TCP User Interface.......................................21 8.2. TCP States and Transitions...............................22 8.3. TCP Segments.............................................22 8.4. Sending TCP Segments.....................................23 8.5. Receiving TCP Segments...................................24 Touch Expires August 16, 2009 [Page 2] Internet-Draft The TCP Simple Authentication Option February 2009 8.6. Impact on TCP Header Size................................25 9. Connection Key Establishment and Duration Issues..............26 9.1. Master Key Reuse Across Socket Pairs.....................27 9.2. Master Key Use Within a Long-lived Connection............27 10. Obsoleting TCP MD5 and Legacy Interactions...................27 11. Interactions with Middleboxes................................28 11.1. Interactions with non-NAT/NAPT Middleboxes..............28 11.2. Interactions with NAT/NAPT Devices......................29 12. Evaluation of Requirements Satisfaction......................29 13. Security Considerations......................................35 14. IANA Considerations..........................................37 15. References...................................................37 15.1. Normative References....................................37 15.2. Informative References..................................38 16. Acknowledgments..............................................40 1. Contributors This document evolved as the result of collaboration of the TCP Authentication Design team (tcp-auth-dt), whose members were (alphabetically): Mark Allman, Steve Bellovin, Ron Bonica, Wes Eddy, Lars Eggert, Charlie Kaufman, Andrew Lange, Allison Mankin, Sandy Murphy, Joe Touch, Sriram Viswanathan, Brian Weis, and Magnus Westerlund. The text of this document is derived from a proposal by Joe Touch and Allison Mankin [To06] (originally from June 2006), which was both inspired by and intended as a counterproposal to the revisions to TCP MD5 suggested in a document by Ron Bonica, Brian Weis, Sriran Viswanathan, Andrew Lange, and Owen Wheeler [Bo07] (originally from Sept. 2005) and in a document by Brian Weis [We05]. Russ Housley suggested L4/application layer management of the TSAD. Steve Bellovin motivated the KeyID field. Eric Rescorla suggested the use of ISNs in the connection key computation and ESNs to avoid replay attacks, and Brian Weis extended the computation to incorporate the entire connection ID and provided the details of the connection key computation. 2. Introduction The TCP MD5 Signature (TCP MD5) is a TCP option that authenticates TCP segments, including the TCP IPv4 pseudoheader, TCP header, and TCP data. It was developed to protect BGP sessions from spoofed TCP segments which could affect BGP data or the robustness of the TCP connection itself [RFC2385][RFC4953]. There have been many recent concerns about TCP MD5. Its use of a simple keyed hash for authentication is problematic because there Touch Expires August 16, 2009 [Page 3] Internet-Draft The TCP Simple Authentication Option February 2009 have been escalating attacks on the algorithm itself [Wa05]. TCP MD5 also lacks both key management and algorithm agility. This document adds the latter, but notes that TCP does not provide a sufficient framework for cryptographic key management, because SYN segments lack sufficient remaining space to support key coordination in-band (see Section 8.6). This document obsoletes the TCP MD5 option with a more general TCP Authentication Option (TCP-AO), to support the use of other, stronger hash functions, provide replay protection for long- lived connections and across repeated instances of a single connection, and to provide a more structured recommendation on external key management. The result is compatible with IPv6, and is fully compatible with requirements under development for a replacement for TCP MD5 [Be07]. This document is not intended to replace the use of the IPsec suite (IPsec and IKE) to protect TCP connections [RFC4301][RFC4306]. In fact, we recommend the use of IPsec and IKE, especially where IKE's level of existing support for parameter negotiation, session key negotiation, or rekeying are desired. TCP-AO is intended for use only where the IPsec suite would not be feasible, e.g., as has been suggested is the case to support some routing protocols, or in cases where keys need to be tightly coordinated with individual transport sessions [Be07]. Note that TCP-AO obsoletes TCP MD5, although a particular implementation may support both for backward compatibility. For a given connection, only one can be in use. TCP MD5-protected connections cannot be migrated to TCP-AO because TCP MD5 does not support any changes to a connection's security algorithm once established. 2.1. Executive Summary This document replaces TCP MD5 as follows [RFC2385]: o TCP-AO uses a separate option Kind for TCP-AO (TBD-IANA-KIND). o TCP-AO allows TCP MD5 to continue to be used concurrently for legacy connections. o TCP-AO replaces MD5's single MAC algorithm with MACs specified in a separate document and allows extension to include other MACs. Touch Expires August 16, 2009 [Page 4] Internet-Draft The TCP Simple Authentication Option February 2009 o TCP-AO allows rekeying during a TCP connection, assuming that an out-of-band protocol or manual mechanism coordinates the key change. In such cases, a key ID allows the efficient concurrent use of multiple keys. Note that TCP MD5 does not preclude rekeying during a connection, but does not require its support either. Further, TCP-AO supports rekeying with zero packet loss, whereas rekeying in TCP MD5 can lose packets in transit during the changeover or require trying multiple keys on each received segment during key use overlap because it lacks an explicit key ID. o TCP-AO provides automatic replay protection for long-lived connections using an extended sequence number. o TCP-AO ensures per-connection keys as unique as the TCP connection itself, using TCP's ISNs for differentiation, even when static master keys are used across repeated instances of a socket pair. o This document provides detail in how this option interacts with TCP's states, event processing, and user interface. o The TCP-AO option is 3 bytes shorter than TCP MD5 (15 bytes overall, rather than 18) in the default case (using a 96-bit MAC). This document differs from an IPsec/IKE solution in that TCP-AO as follows [RFC4301][RFC4306]: o TCP-AO does not support dynamic parameter negotiation. o TCP-AO uses TCP's socket pair (source address, destination address, source port, destination port) as a security parameter index, rather than using a separate field as a primary index (IPsec's SPI). o TCP-AO forces a change of computed MACs when a connection restarts, even when reusing a TCP socket pair (IP addresses and port numbers) [Be07]. o TCP-AO does not support encryption. o TCP-AO does not authenticate ICMP messages (some ICMP messages may be authenticated via IPsec, depending on the configuration). 2.2. Changes from Previous Versions [NOTE: to be omitted upon final publication as RFC] Touch Expires August 16, 2009 [Page 5] Internet-Draft The TCP Simple Authentication Option February 2009 2.2.1. New in draft-ietf-tcp-auth-opt-03 o Added a placeholder to discuss key change coordination in Section 9. o Moved discussion of required MAC algorithms and PRF to a separate document, indicated as RFC-TBD until assigned. Included the PRF in the TSAD master key tuple so that TCP-AO is PRF algorithm agile, and updated general PRF input format. o Revised the description the TSAD and impact to the TCP user interface. Removed the description of the TSAD API. Access to the API is assumed specific to the implementation, and not part of the protocol specification. o Clarified the different uses of the term key; includes master key (from the TSAD) and connection key (per-connection key, derived from the master via the PRF). o Explained the ESN pseudocode operation in detail. o Added a contributors section up front. o Update discussion of requirements to be sufficiently stand-alone; update list to correlate more directly to Be07 (so that Be07 can be dropped from consideration for publication). o Provided detail on size of typical options (motivating a small option). o Confirmed WG consensus on IETF-72 topic - no algorithm ID and T- bit (options excluded) locations in the header. o Confirmed WG consensus on IETF-72 topic - no additional header bits for in-band key change signaling (the "K" bit from [Bo07]). 2.2.2. New in draft-ietf-tcp-auth-opt-02 o List issue - Replay Protection: incorporated extended sequence number space, not using KeyID space. o List issue - Unique Connection Keys: ISNs are used to generate unique connection keys even when static keys used for repeated instances of a socket pair. o List issue - Header Format and Alignment: Moved KeyID to front. Touch Expires August 16, 2009 [Page 6] Internet-Draft The TCP Simple Authentication Option February 2009 o List issue - Reserved KeyID Value: Suggestion to reserve a single KeyID value for implementation optimization received no support on the WG list, so this was not changed. o List issue - KeyID Randomness: KeyIDs are not assumed random; a note was added that nonce-based filtering should be done on a portion of the MAC (incorporated into the algorithm), and that header fields should not be assumed to have cryptographic properties (e.g., randomness). o List issue - Support for NATs: preliminary rough consensus suggests that TCP-AO should not be augmented to support NAT traversal. Existing mechanisms for such traversal (UDP support) can be applied, or IPsec NAT traversal is recommended in such cases instead. o IETF-72 topic - providing algorithm ID and T-bit (options excluded) locations in the header: (No current consensus was reached on this topic, so no change was made.) o IETF-72 topic - providing additional header bits for in-band key change signaling (draft-bonica's "K" bit): (No current consensus was reached on this topic, so no change was made.) o Clarified TCP-AO as obsoleting TCP MD5. o Clarified the MAC Type as referring to the IANA registry of IKEv2 transforms, not the RFC establishing that registry. o Added citation to the Wang/Yu paper regarding attacks on MD5 Wa05 to replace reports in Be05 and Bu06. o Explained why option exclusion can't be changed during a connection. o Clarified that AO explicitly allows rekeying during a TCP connection, without impacting packet loss. o Described TCP-AO's interaction with reboots more clearly, and explained the need to clear out old state that persists indefinitely. 2.2.3. New in draft-ietf-tcp-auth-opt-01 o Require KeyID in all versions. Remove odd/even indicator of KeyID usage. Touch Expires August 16, 2009 [Page 7] Internet-Draft The TCP Simple Authentication Option February 2009 o Relax restrictions on key reuse: requiring an algorithm for nonce introduction based on ISNs, and suggest key rollover every 2^31 bytes (rather than using an extended sequence number, which introduces new state to the TCP connection). o Clarify NAT interaction; currently does not support omitting the IP addresses or TCP ports, both of which would be required to support NATs without any coordination. This appears to present a problem for key management - if the key manager knows the received addrs and ports, it should coordinate them (as indicated in Sec 8). o Options are included or excluded all-or-none. Excluded options are deleted, not just zeroed, to avoid the impact of reordering or length changes of such options. o Augment replay discussion in security considerations. o Revise discussion of IKEv2 MAC algorithm names. o Remove executive summary comparison to expired documents. o Clarified key words to exclude lower case usage. 2.2.4. New in draft-ietf-tcp-auth-opt-00 o List of TBD values, and indication of how each is determined. o Changed TCP-SA to TCP-AO (removed 'simple' throughout). o Removed proposed NAT mechanism; cited RFC-3947 NAT-T as appropriate approach instead. o Made several changes coordinated in the TCP-AUTH-DT as follow: o Added R. Bonica as co-author. o Use new TCP option Kind in the core doc. o Addresses the impact of explicit declines on security. o Add limits to TSAD size (2 <= TSAD <= 256). o Allow 0 as a legitimate KeyID. o Allow the WG to determine the two appropriate required MAC algorithms. Touch Expires August 16, 2009 [Page 8] Internet-Draft The TCP Simple Authentication Option February 2009 o Add TO-DO items. o Added discussion at end of Introduction as to why TCP MD5 connections cannot be upgraded to TCP-AO. 2.2.5. New in draft-touch-tcp-simple-auth-03 o Added support for NAT/NAPT. o Added support for IPv6. o Added discussion of how this proposal satisfies requirements under development, including those indicated in [Be07]. o Clarified the byte order of all data used in the MAC. o Changed the TCP option exclusion bit from a bit to a list. 2.2.6. New in draft-touch-tcp-simple-auth-02 o Add reference to Bellovin's need-for-TCP-auth doc [Be07]. o Add reference to SP4 [SDNS88]. o Added notes that TSAD to be externally implemented; this was compatible with the TSAD described in the previous version. o Augmented the protocol to allow a KeyID, required to support efficient overlapping keys during rekeying, and potentially useful during connection establishment. Accommodated by redesigning the TSAD. o Added the odd/even indicator for the KeyID. o Allow for the exclusion of all TCP options in the MAC calculation. 2.2.7. New in draft-touch-tcp-simple-auth-01 o Allows intra-session rekeying, assuming out-of-band coordination. o MUST allow TSAD entries to change, enabling rekeying within a TCP connection. o Omits discussion of the impact of connection reestablishment on BGP, because added support for rekeying renders this point moot. o Adds further discussion on the need for rekeying. Touch Expires August 16, 2009 [Page 9] Internet-Draft The TCP Simple Authentication Option February 2009 3. Conventions used in 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]. In this document, these words will appear with that interpretation only when in ALL CAPS. Lower case uses of these words are not to be interpreted as carrying RFC-2119 significance. 4. The TCP Authentication Option The TCP Authentication Option (TCP-AO) uses a TCP option Kind value of TBD-IANA-KIND. 4.1. Review of TCP MD5 Option For review, the TCP MD5 option is shown in Figure 1. +---------+---------+-------------------+ | Kind=19 |Length=18| MD5 digest... | +---------+---------+-------------------+ | | +---------------------------------------+ | | +---------------------------------------+ | | +-------------------+-------------------+ | | +-------------------+ Figure 1 The TCP MD5 Option [RFC2385] In the TCP MD5 option, the length is fixed, and the MD5 digest occupies 16 bytes following the Kind and Length fields, using the full MD5 digest of 128 bits [RFC1321]. The TCP MD5 option specifies the use of the MD5 digest calculation over the following values in the following order: 1. The TCP pseudoheader (IP source and destination addresses, protocol number, and segment length). 2. The TCP header excluding options and checksum. 3. The TCP data payload. Touch Expires August 16, 2009 [Page 10] Internet-Draft The TCP Simple Authentication Option February 2009 4. The connection key. 4.2. TCP-AO Option The new TCP-AO option provides a superset of the capabilities of TCP MD5, and is minimal in the spirit of SP4 [SDNS88]. TCP-AO uses a new Kind field, and similar Length field to TCP MD5, as well as a KeyID field as shown in Figure 2. +----------+----------+----------+----------+ | Kind | Length | KeyID | MAC | +----------+----------+----------+----------+ | MAC (con't) ... +----------------------------------... ...-----------------+ ... MAC (con't) | ...-----------------+ Figure 2 The TCP-AO Option The TCP-AO defines the following fields: o Kind: An unsigned 1-byte field indicating the TCP-AO Option. TCP- AO uses a new Kind value of TBD-IANA-KIND. Because of how keys are managed (see Section 7), an endpoint will not use TCP-AO for the same connection in which TCP MD5 is used. >> A single TCP segment MUST NOT have more than one TCP-AO option. o Length: An unsigned 1-byte field indicating the length of the TCP- AO option in bytes, including the Kind, Length, KeyID, and MAC fields. >> The Length value MUST be greater than or equal to 3. >> The Length value MUST be consistent with the TCP header length; this is a consistency check and avoids overrun/underrun abuse. Values of 3 and other small values are of dubious utility (e.g., for MAC=NONE, or small values for very short MACs) but not specifically prohibited. Touch Expires August 16, 2009 [Page 11] Internet-Draft The TCP Simple Authentication Option February 2009 o KeyID: An unsigned 1-byte field is used to support efficient key changes during a connection and/or to help with key coordination during connection establishment, and will be discussed further in Section 4. Note that the KeyID has no cryptographic properties - it need not be random, nor are there any reserved values. o MAC: Message Authentication Field. Its contents are determined by the particulars of the security association. Typical MACs are 96- 128 bits (12-16 bytes), but any length that fits in the header of the segment being authenticated is allowed. >> Required support for TCP-AO MACs as defined in RFC-TBD; other MACs MAY be supported [RFC2403]. The MAC is computed over the following fields in the following order: 1. The extended sequence number (ESN), in network-standard byte order, as follows (described further in Section 5): +--------+--------+--------+--------+ | ESN | +--------+--------+--------+--------+ Figure 3 Extended sequence number The ESN for transmitted segments is locally maintained from a locally maintained SND.ESN value, for received segments, a local RCV.ESN value is used. The details of how these values are maintained and used is described in Sections 5, 8.4, and 8.5. 2. The TCP pseudoheader: IP source and destination addresses, protocol number and segment length, all in network byte order, prepended to the TCP header below. The pseudoheader is exactly as used for the TCP checksum in either IPv4 or IPv6 [RFC793][RFC2460]: +--------+--------+--------+--------+ | Source Address | +--------+--------+--------+--------+ | Destination Address | +--------+--------+--------+--------+ | zero | Proto | TCP Length | +--------+--------+--------+--------+ Figure 4 TCP IPv4 pseudoheader [RFC793] Touch Expires August 16, 2009 [Page 12] Internet-Draft The TCP Simple Authentication Option February 2009 +--------+--------+--------+--------+ | | + + | | + Source Address + | | + + | | + + +--------+--------+--------+--------+ | | + + | | + Destination Address + | | + + | | +--------+--------+--------+--------+ | Upper-Layer Packet Length | +--------+--------+--------+--------+ | zero | Next Header | +--------+--------+--------+--------+ Figure 5 TCP IPv6 pseudoheader [RFC2460] 3. The TCP header, by default including options, and where the TCP checksum and TCP-AO MAC fields are set to zero, all in network byte order 4. TCP data, in network byte order Note that the connection key is not included here; the MAC algorithm indicates how to use the connection key, e.g., as HMACs do in general [RFC2104][RFC2403]. The connection key is derived from the TSAD entry's master key as described in Sections 7, 8.4, and 8.5. By default, TCP-AO includes the TCP options in the MAC calculation because these options are intended to be end-to-end and some are required for proper TCP operation (e.g., SACK, timestamp, large windows). Middleboxes that alter TCP options en-route are a kind of attack and would be successfully detected by TCP-AO. In cases where the configuration of the connection's security association state indicates otherwise, the TCP options can be excluded from the MAC calculation. When options are excluded, all options - including TCP- AO - are skipped over during the MAC calculation (rather than being zeroed). Touch Expires August 16, 2009 [Page 13] Internet-Draft The TCP Simple Authentication Option February 2009 The TCP-AO option does not indicate the MAC algorithm either implicitly (as with TCP MD5) or explicitly. The particular algorithm used is considered part of the configuration state of the connection's security association and is managed separately (see Section 7). 5. Preventing replay attacks within long-lived connections TCP uses a 32-bit sequence number which may, for long-lived connections, roll over and repeat. This could result in TCP segments being intentionally and legitimately replayed within a connection. TCP-AO prevents replay attacks, and thus requires a way to differentiate these legitimate replays from each other, and so it adds a 32-bit extended sequence number (ESN) for transmitted and received segments. The ESN extends TCP's sequence number so that segments within a single connection are always unique. When TCP's sequence number rolls over, there is a chance that a segment could be repeated in total; using an ESN differentiates even identical segments sent with identical sequence numbers at different times in a connection. TCP-AO emulates a 64-bit sequence number space by inferring when to increment the high-order 32-bit portion (the ESN) based on transitions in the low-order portion (the TCP sequence number). TCP-AO thus maintains SND.ESN for transmitted segments, and RCV.ESN for received segments, both initialized as zero when a connection begins. The intent of these ESNs is, together with TCP's 32-bit sequence numbers, to provide a 64-bit overall sequence number space. For transmitted segments SND.ESN can be implemented by extending TCP's sequence number to 64-bits; SND.ESN would be the top (high- order) 32 bits of that number. For received segments, TCP-AO needs to emulate the use of a 64-bit number space, and correctly infer the appropriate high-order 32-bits of that number as RCV.ESN from the received 32-bit sequence number and the current connection context. The implementation of ESNs is not specified in this document, but one possible way is described here that can be used for either RCV.ESN, SND.ESN, or both. Consider an implementation with two ESNs as required (SND.ESN, RCV.ESN), and additional variables as listed below, all initialized to zero, as well as a current TCP segment field (SEG.SEQ): o SND.PREV_SEQ, needed to detect rollover of SND.SEQ Touch Expires August 16, 2009 [Page 14] Internet-Draft The TCP Simple Authentication Option February 2009 o RCV.PREV_SEQ, needed to detect rollover of RCV.SEQ o SND.ESN_FLAG, which indicates when to increment the SND.ESN o RCV.ESN_FLAG, which indicates when to increment the RCV.ESN When a segment is received, the following algorithm (written in C) computes the ESN used in the MAC; an equivalent algorithm can be applied to the "SND" side: # # ROLL is just shorthand ROLL = (RCV.PREV_SEQ > 0x7fff) && (SEG.SEQ < 0x7fff); # # set the flag when the SEG.SEQ first rolls over if ((RCV.ESN_FLAG == 0) && (ROLL)) { RCV.ESN = RCV.ESN + 1; RCV.ESN_FLAG = 1; } # # decide which ESN to use during rollover after incremented if ((RCV.ESN_FLAG == 1) && (ROLL)) { ESN = RCV.ESN - 1; # use the pre-increment value } else { ESN = RCV.ESN; # use the current value } # # reset the flag in the *middle* of the window if ((RCV.PREV_SEQ < 0x7fff) && (SEG.SEQ > 0x7fff)) { RCV.ESN_FLAG = 0; } # # save the current SEQ for the next time through the code RCV.PREV_SEQ = SEG.SEQ; In the above code, ROLL is true in the first line when the sequence number rolls over, i.e., when the new number is low (in the bottom half of the number space) and the old number is high (in the top half of the number space). The first time this happens, the ESN is incremented and a flag is set. The flag prevents the ESN from being incremented again until the flag is reset, which happens in the middle of the window (when the old number is in the bottom half and the new is in the top half). Because the receive window is never larger than half of the number space, it is impossible to both set and reset the flag at the same time - outstanding packets, regardless of reordering, cannot straddle both regions simultaneously. Touch Expires August 16, 2009 [Page 15] Internet-Draft The TCP Simple Authentication Option February 2009 6. Computing connection keys from TSAD entries TSAD entries, described in Section 7, include master keys which are used in conjunction with a TCP's connection ISNs to generate unique connection keys. This allows a static master key to be reused across different connections, or across different instances of connections within a socket pair, while maintaining unique connection keys. Unique connection keys are generated without relying on external key management properties. Given a master key tuple, the TCP socket pair, and the connection ISNs, the connection key used in the MAC algorithm is computed as follows, truncated to the same length as the master key, using a pseudorandom function (PRF): Conn_key = PRF(TSAD_master_key, input) where input = 0 + "TCP-AO" + connblock + TSAD_master_key_len The components of the input are concatenated as a single byte string (the string concatenation operator is shown here as "+"). The initial zero of the input is a single byte, "TCP-AO" is a null-terminated string, connblock is defined below, and TSAD_master_key_len is the length of the TSAD master key in bytes, as stored in the TSAD entry. The PRF to be used for a given master key is indicated in the TDAD master key tuple, and details of the PRF are provided in [RFC-TBD]. The connection block (connblock) is defined as follows (IP addresses are correspondingly longer for IPv6 addresses): +--------+--------+--------+--------+ | Source IP | +--------+--------+--------+--------+ | Destination IP | +--------+--------+--------+--------+ | Source Port | Dest. Port | +--------+--------+--------+--------+ | Source ISN | +--------+--------+--------+--------+ | Destination ISN | +--------+--------+--------+--------+ Figure 6 Connection block used for connection key generation "Source" and "destination" are defined by the direction of the segment being MAC'd; for incoming packets, source is the remote side, Touch Expires August 16, 2009 [Page 16] Internet-Draft The TCP Simple Authentication Option February 2009 whereas for outgoing packets source is the local side. This further ensures that connection keys generated for each direction are unique. For SYN segments (segments with the SYN set, but the ACK not set), the destination ISN is not known. For these segments, the connection key is computed using the connection block shown above, in which the Destination ISN value is zero. For all other segments, the ISN pair is used when known. If the ISN pair is not known, e.g., when sending a RST after a reboot, the segment should be sent without authentication; if authentication was required, the segment cannot have been MAC'd properly anyway and would have been dropped on receipt. >> TCP-AO SYN segments (SYN set, no ACK set) MUST use a destination ISN of zero (whether sent or received); all other segments use the known ISN pair. >> Segments sent in response to connections for which the ISNs are not known SHOULD NOT use TCP-AO. Once a connection is established, a connection key would typically be cached to avoid recomputing it on a per-segment basis (e.g., in the TCP Transmission Control Block, i.e, the TCB [RFC793]). The use of both ISNs in the connection key computation ensures that segments cannot be replayed across repeated connections reusing the same socket pair (provided the ISN pair does not repeat, which is unlikely because both endpoints should select ISNs pseudorandomly [RFC1948], their 32-bit space avoids repeated use except under reboot, and reuse assumes both sides repeat their use on the same connection). In general, a SYN would be MAC'd using a destination ISN of zero (whether sent or received), and all other segments would be MAC'd using the ISN pair for the connection. There are other cases in which the destination ISN is not known, but segments are emitted, such as after an endpoint reboots, when is possible that the two endpoints would not have enough information to authenticate segments. In such cases, TCP's timeout mechanism will allow old state to be cleared to enable new connections, except where the user timeout is disabled; it is important that implementations are capable of detecting excesses of TCP connections in such a configuration and can clear them out if needed to protect its memory usage [Je07]. 7. Security Association Management TCP-AO relies on a TCP Security Association Database (TSAD), which indicates whether a TCP connection requires TCP-AO, and its parameters when so. The TSAD is described as an explicit component of Touch Expires August 16, 2009 [Page 17] Internet-Draft The TCP Simple Authentication Option February 2009 TCP-AO to enable external (master) key management mechanisms - automatic or manual - to interact with TCP-AO as needed. TSAD entries are assumed to exist at the endpoints where TCP-AO is used, in advance of the connection: 1. TCP connection identifier (ID), i.e., socket pair - IP source address, IP destination address, TCP source port, and TCP destination port [RFC793]. TSAD entries are uniquely determined by their TCP connection ID, which is used to index those entries. A TSAD entry may allow wildcards, notably in the source port value. >> There MUST be no more than one matching TSAD entry per direction for a fully-instantiated (no wildcards) TCP connection ID. 2. For each of inbound (for received TCP segments) and outbound (for sent TCP segments) directions for this connection (except as noted): a. TCP option flag. When 0, this flag allows default operation, i.e., TCP options are included in the MAC calculation, with TCP-AO's MAC field zeroed out. When 1, all options (including TCP-AO) are excluded from all MAC calculations (skipped over, not simply zeroed). >> The TCP option flag MUST default to 0 (i.e., options not excluded). >> The TCP option flag MUST NOT change during a TCP connection. The TCP option flag cannot change during a connection because TCP state is coordinated during connection establishment. TCP lacks a handshake for modifying that state after a connection has been established. b. An extended sequence number (ESN). The ESN enables each segment's MAC calculation to have unique input data, even when payload data is retransmitted and the TCP sequence number repeats due to wraparound. The ESN is initialized to zero upon connection establishment. Its use in the MAC calculation is described in Section 4.2, and its management is described in Section 5. Touch Expires August 16, 2009 [Page 18] Internet-Draft The TCP Simple Authentication Option February 2009 c. An ordered list of zero or more master key tuples. Each tuple is defined as the set as follows: >> Components of a TSAD master key tuple MUST NOT change during a connection. Keeping the tuple components static ensures that the KeyID uniquely determines the properties of a packet; this supports use of the KeyID to determine the packet properties. >> The set of TSAD master key tuples MAY change during a connection, but KeyIDs of those tuples MUST NOT overlap. I.e., tuple parameter changes MUST be accompanied by master key changes. i. KeyID. The value as used in the TCP-AO option; used to differentiate connection keys in concurrent use that are derived from different master keys. >> A TSAD implementation MUST support at least two KeyIDs per connection per direction, and MAY support up to 256. >> A KeyID MUST support any value, 0-255 inclusive. There are no reserved KeyID values. KeyID values are assigned arbitrarily. They can be assigned in sequence, or based on any method mutually agreed by the connection endpoints (e.g., using an external master key management mechanism). >> KeyIDs MUST NOT be assumed to be randomly assigned. Touch Expires August 16, 2009 [Page 19] Internet-Draft The TCP Simple Authentication Option February 2009 ii. MAC type. Indicates the MAC used for this connection, as defined in [RFC-TBD]. This includes the MAC algorithm (e.g., HMAC-SHA1, AES-CMAC, etc.) and the length of the MAC as truncated to (e.g., 96, 128, etc.). >> A MAC type of "NONE" MUST be supported, to indicate that authentication is not used in this direction; this allows asymmetric use of TCP-AO. >> At least one direction (inbound/outbound) SHOULD have a non-"NONE" MAC in practice, but this MUST NOT be strictly required by an implementation. >> When the outbound MAC is set to values other than "NONE", TCP-AO MUST occur in every outbound TCP segment for that connection; when set to NONE or when no tuple exists, TCP-AO MUST NOT occur in those segments. >> When the inbound MAC is set to values other than "NONE", TCP-AO MUST occur in every inbound TCP segment for that connection; when set to "NONE" or when no tuple exists, TCP-AO SHOULD NOT be added to those segments, but MAY occur and MUST be ignored. iii. Master key length. Indicates the length of the master key in bytes. iv. Master key. A byte sequence used for generating connection keys, this may be derived from a separate shared key by an external protocol over a separate channel. This sequence is used in network-standard byte order in the connection key generation algorithm described in Section 6. v. PRF. A pseudorandom function used for the geneation of a connection key from the master key tuple, as described in Section 6. The specific functions used are described in [RFC-TBD]. It is anticipated that TSAD entries for TCP connections in states other than CLOSED can be indexed in the TCP TCB for convenience, but that the index would reference a separate database with entries for all connections to an IP address (see Section 9.1 for notes on the latter. This means that for a particular endpoint (i.e., IP address) there would be exactly one database that is consulted by all pending connections, the same way that there is only one table of TCBs (a database can support multiple endpoints, but an endpoint is Touch Expires August 16, 2009 [Page 20] Internet-Draft The TCP Simple Authentication Option February 2009 represented in only one database). Multiple databases could be used to support virtual hosts, i.e., groups of interfaces. Note that the TCP-AO fields omit an explicit algorithm ID; that algorithm is already specified by the TCP connection ID and stored in the TSAD. Also note that this document does not address how TSAD entries are created by users/processes; it specifies how they must be destroyed corresponding to connection states, but users/processes may destroy entries as well. It is presumed that a TSAD entry affecting a particular connection cannot be destroyed during an active connection - or, equivalently, that its parameters are copied to TSAD entries local to the connection (i.e., instantiated) and so changes would affect only new connections. The TSAD could be managed by a separate application protocol. 8. TCP-AO Interaction with TCP The following is a description of how TCP-AO affects various TCP states, segments, events, and interfaces. This description is intended to augment the description of TCP as provided in RFC-793, and its presentation mirrors that of RFC-793 as a result [RFC793]. 8.1. TCP User Interface The TCP user interface supports active and passive OPEN, SEND, RECEIVE, CLOSE, STATUS and ABORT commands. TCP-AO does not alter this interface as it applies to TCP, but some commands or command sequences of the interface need to be modified to support TCP-AO. TCP-AO does not specify the details of how this is achieved. TCP-AO requires the TCP user interface be extended to allow the TSAD to be configured, as well as to allow an ongoing connection to manage which KeyID tuples are active. The TSAD needs to be configured prior to connection establishment, and possibly changed during a connection: >> TCP OPEN, or the sequence of commands that configure a connection to be in the active or passive OPEN state, MUST be augmented so that a TSAD entry can be configured. >> A TCP-AO implmentation MUST allow TSAD entries for ongoing TCP connections (i.e., not in the CLOSED state) to be modified. Parameters not used to index a connection MAY be modified; parameters used to index a connection MUST NOT be modified. Touch Expires August 16, 2009 [Page 21] Internet-Draft The TCP Simple Authentication Option February 2009 The TSAD information of a connection needs to be available for confirmation; this includes the ability to read the connection key: >> TCP STATUS SHOULD be augmented to allow the TSAD entry of a current or pending connection to be read (for confirmation). Senders need to be able to determine when the outgoing KeyID changes; this change immediately affects all subsequent outgoing segments (i.e., it need not be synchronized with the data of the SEND call, if indicated therein): >> TCP SEND, or a sequence of commands resulting in a SEND, MUST be augmented so that the KeyID of a TSAD entry can be indicated. It may be useful to change the sender-side active KeyID even when no data is being sent, which can be achieved by sending a zero-length buffer or by using a non-send interface (e.g., socket options in Unix), depending on the implementation. It is also useful for the receive side to indicate the recent KeyID received; although there could be a number of such KeyIDs, the KeyIDs are not expected to change quickly so any recent sample of a received KeyID is sufficient: >> TCP RECEIVE, or the sequence of commands resulting in a RECEIVE, MUST be augmented so that the KeyID of a recently received segment is available to the user out-of-band (e.g., as an additional parameter to RECEIVE, or via a STATUS call). 8.2. TCP States and Transitions TCP includes the states LISTEN, SYN-SENT, SYN-RECEIVED, ESTABLISHED, FIN-WAIT-1, FIN-WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK, TIME-WAIT, and CLOSED. >> A TSAD entry MAY be associated with any TCP state. >> A TSAD entry MAY underspecify the TCP connection for the LISTEN state. Such an entry MUST NOT be used for more than one connection progressing out of the LISTEN state. 8.3. TCP Segments TCP includes control (at least one of SYN, FIN, RST flags set) and data (none of SYN, FIN, or RST flags set) segments. Note that some control segments can include data (e.g., SYN). Touch Expires August 16, 2009 [Page 22] Internet-Draft The TCP Simple Authentication Option February 2009 >> All TCP segments MUST be checked against the TSAD for matching TCP connection IDs. >> TCP segments matching TSAD entries with non-NULL MACs without TCP- AO, or with TCP-AO and whose MACs and KeyIDs do not validate MUST be silently discarded. >> TCP segments with TCP-AO but not matching TSAD entries MUST be silently accepted; this is required for equivalent function with TCPs not implementing TCP-AO. >> Silent discard events SHOULD be signaled to the user as a warning, and silent accept events MAY be signaled to the user as a warning. Both warnings, if available, MUST be accessible via the STATUS interface. Either signal MAY be asynchronous, but if so they MUST be rate-limited. Either signal MAY be logged; logging SHOULD allow rate- limiting as well. All TCP-AO processing occurs between the interface of TCP and IP; for incoming segments, this occurs after validation of the TCP checksum. For outgoing segments, this occurs before computation of the TCP checksum. Note that the TCP-AO option is not negotiated. It is the responsibility of the receiver to determine when TCP-AO is required and to enforce that requirement. 8.4. Sending TCP Segments The following procedure describes the modifications to TCP to support TCP-AO when a segment departs. 1. Check the segment's TCP connection ID against the TSAD 2. If there is NO TSAD entry, omit the TCP-AO option. Proceed with computing the TCP checksum and transmit the segment. 3. If there is a TSAD entry with zero master key tuples, omit the TCP-AO option. Proceed with computing the TCP checksum and transmit the segment. 4. If there is a TSAD entry and a master key tuple and the outgoing MAC is NONE, omit the TCP-AO option. Proceed with computing the TCP checksum and transmit the segment. 5. If there is a TSAD entry and a master key tuple and the outgoing MAC is not NONE: Touch Expires August 16, 2009 [Page 23] Internet-Draft The TCP Simple Authentication Option February 2009 a. Augment the TCP header with the TCP-AO, inserting the appropriate Length and KeyID based on the indexed TSAD entry. Update the TCP header length accordingly. b. Determine SND.ESN as described in Section 5. c. Determine the connection key from the indexed TSAD entry as described in Section 6. d. Compute the MAC using the indexed TSAD entry and data from the segment as specified in Section 4.2, including the TCP pseudoheader and TCP header. Include or exclude the options as indicated by the TSAD entry's TCP option exclusion flag. e. Insert the MAC in the TCP-AO field. f. Proceed with computing the TCP checksum on the outgoing packet and transmit the segment. 8.5. Receiving TCP Segments The following procedure describes the modifications to TCP to support TCP-AO when a segment arrives. 1. Check the segment's TCP connection ID against the TSAD. 2. If there is NO TSAD entry, proceed with TCP processing. 3. If there is a TSAD entry with zero master key tuples, proceed with TCP processing. 4. If there is a TSAD entry with a master key tuple and the incoming MAC is NONE, proceed with TCP processing. 5. If there is a TSAD entry with a master key tuple and the incoming MAC is not NONE: a. Check that the segment's TCP-AO Length matches the length indicated by the indexed TSAD. i. If Lengths differ, silently discard the segment. Log and/or signal the event as indicated in Section 8.3. b. Use the KeyID value to index the appropriate connection key for this connection. Touch Expires August 16, 2009 [Page 24] Internet-Draft The TCP Simple Authentication Option February 2009 i. If the TSAD has no entry corresponding to the segment's KeyID, silently discard the segment. c. Determine the segment's RCV.ESN as described in Section 5. d. Determine the segment's connection key from the indexed TSAD entry as described in Section 6. e. Compute the segment's MAC using the indexed TSAD entry and portions of the segment as indicated in Section 4.2. Again, if options are excluded (as per the TCP option exclusion flag), they are skipped over (rather than zeroed) when used as input to the MAC calculation. i. If the computed MAC differs from the TCP-AO MAC field value, silently discard the segment. Log and/or signal the event as indicated in Section 8.3. f. Proceed with TCP processing of the segment. It is suggested that TCP-AO implementations validate a segment's Length field before computing a MAC, to reduce the overhead incurred by spoofed segments with invalid TCP-AO fields. Additional reductions in MAC validation overhead can be supported in the MAC algorithms, e.g, by using a computation algorithm that prepends a fixed value to the computed portion and a corresponding validation algorithm that verifies the fixed value before investing in the computed portion. Such optimizations would be contained in the MAC algorithm specification, and thus are not specified in TCP-AO explicitly. Note that the KeyID cannot be used for connection validation per se, because it is not assumed random. 8.6. Impact on TCP Header Size The TCP-AO option typically uses a total of 17-19 bytes of TCP header space. TCP-AO is no larger than and typically 3 bytes smaller than the TCP MD5 option (assuming a 96-bit MAC). Note that TCP option space is most critical in SYN segments, because flags in those segments could potentially increase the option space area in other segments. Because TCP ignores unknown segments, however, it is not possible to extend the option space of SYNs without breaking backward-compatibility. Touch Expires August 16, 2009 [Page 25] Internet-Draft The TCP Simple Authentication Option February 2009 TCP's 4-bit data offset requires that the options end 60 bytes (15 32-bit words) after the header begins, including the 20-byte header. This leaves 40 bytes for options, of which 15 are expected in current implementations (listed below), leaving at most 20 for TCP-AO. Assuming a 96-bit MAC, TCP-AO consumes 15 bytes, leaving up to 10 bytes for other options (depending on implementation dependant alignment padding, which could consume another 2 bytes at most). o SACK permitted (2 bytes) [RFC2018][RFC3517] o Timestamps (10 bytes) [RFC1323] o Window scale (3 bytes) [RFC1323] Although TCP option space is limited, we believe TCP-AO is consistent with the desire to authenticate TCP at the connection level for similar uses as were intended by TCP MD5. 9. Connection Key Establishment and Duration Issues The TCP-AO option does not provide a mechanism for connection key negotiation or parameter negotiation (MAC algorithm, length, or use of the TCP-AO option), or for coordinating rekeying during a connection. We assume out-of-band mechanisms for master key establishment, parameter negotiation, and rekeying. This separation of master key use from master key management is similar to that in the IPsec security suite [RFC4301][RFC4306]. We encourage users of TCP-AO to apply known techniques for generating appropriate master keys, including the use of reasonable master key lengths, limited connection key sharing, and limiting the duration of master key use [RFC3562]. This also includes the use of per- connection nonces, as suggested in Section 4.2. TCP-AO supports rekeying in which new master keys are negotiated and coordinated out-of-band, either via a protocol or a manual procedure [RFC4808]. New master key use is coordinated using the out-of-band mechanism to update the TSAD at both TCP endpoints. When only a single master key is used at a time, the temporary use of invalid master keys could result in packets being dropped; although TCP is already robust to such drops, TCP-AO uses the KeyID field to avoid such drops. The TSAD can contain multiple concurrent master keys, where the KeyID field is used to identify the master key that corresponds to the connection key used for a segment, to avoid the need for expensive trial-and-error testing of master keys in sequence. Touch Expires August 16, 2009 [Page 26] Internet-Draft The TCP Simple Authentication Option February 2009 TCP-AO does not currently provide an explicit key coordination mechanism. Such a mechanism is useful when new keys are installed, or when keys are changed, to determine when to commence using installed keys. Note that because TCP-AO uses directional keys, the receive- side keys can be installed in advance of the send side, avoiding the need for tight coordination between endpoints. The KeyID field is also useful in coordinating master keys used for new connections. A TSAD entry may be configured that matches the unbound source port, which would return a set of possible master keys. The KeyID would then indicate the specific master key, allowing more efficient connection establishment; otherwise, the master keys could have been tried in sequence. See also Section 9.1. Users are advised to manage master keys following the spirit of the advice for key management when using TCP MD5 [RFC3562], notably to use appropriate key lengths (12-24 bytes), to avoid sharing master keys among multiple BGP peering arrangements, and to change master keys every 90 days. This requires that the TSAD support monitoring and modification. 9.1. Master Key Reuse Across Socket Pairs Master keys can be reused across different socket pairs within a host, or across different instances of a socket pair within a host. In either case, replay protection is maintained. Master keys reused across different socket pairs cannot enable replay attacks because the TCP socket pair is included in the MAC, as well as in the generation of the connection key. Master keys reused across repeated instances of a given socket pair cannot enable replay attacks because the connection ISNs are included in the connection key generation algorithm, and ISN pairs are unlikely to repeat over useful periods. 9.2. Master Key Use Within a Long-lived Connection TCP-AO uses extended sequence numbers (ESNs) to prevent replay attacks within long-lived connections. Explicit master key rollover, accomplished by external means and indexed using the KeyID field, can be used to change keying material for various reasons (e.g., personnel turnover), but is not required to support long-lived connections. 10. Obsoleting TCP MD5 and Legacy Interactions TCP-AO obsoletes TCP MD5. As we have noted earlier: Touch Expires August 16, 2009 [Page 27] Internet-Draft The TCP Simple Authentication Option February 2009 >> TCP implementations MUST support TCP-AO. Systems implementing TCP MD5 only are considered legacy, and ought to be upgraded when possible. In order to support interoperation with such legacy systems until upgrades are available: >> TCP MD5 SHOULD be supported where interactions with legacy systems is needed. >> A system that supports both TCP-AO and TCP MD5 MUST use TCP-AO for connections unless not supported by its peer, at which point it MAY use TCP MD5 instead. >> A TCP implementation MUST NOT use both TCP-AO and TCP MD5 for a particular TCP connection, but MAY support TCP-AO and TCP MD5 simultaneously for different connections (notably to support legacy use of TCP MD5). The Kind value explicitly indicates whether TCP-AO or TCP MD5 is used for a particular connection in TCP segments. It is possible that the TSAD could be augmented to support TCP MD5, although use of a TSAD-like system is not described in RFC2385. It is possible to require TCP-AO for a connection or TCP MD5, but it is not possible to require 'either'. When an endpoint is configured to require TCP MD5 for a connection, it must be added to all outgoing segments and validated on all incoming segments [RFC2385]. TCP MD5's requirements prohibit the speculative use of both options for a given connection, e.g., to be decided by the other end of the connection. 11. Interactions with Middleboxes TCP-AO may interact with middleboxes, depending on their behavior [RFC3234]. Some middleboxes either alter TCP options (such as TCP-AO) directly or alter the information TCP-AO includes in its MAC calculation. TCP-AO may interfere with these devices, exactly where the device modifies information TCP-AO is designed to protect. 11.1. Interactions with non-NAT/NAPT Middleboxes TCP-AO supports middleboxes that do not change the IP addresses or ports of segments. Such middleboxes may modify some TCP options, in which case TCP-AO would need to be configured to ignore all options in the MAC calculation on connections traversing that element. Touch Expires August 16, 2009 [Page 28] Internet-Draft The TCP Simple Authentication Option February 2009 Note that ignoring TCP options may provide less protection, i.e., TCP options could be modified in transit, and such modifications could be used by an attacker. Depending on the modifications, TCP could have compromised efficiency (e.g., timestamp changes), or could cease correct operation (e.g., window scale changes). These vulnerabilities affect only the TCP connections for which TCP-AO is configured to ignore TCP options. 11.2. Interactions with NAT/NAPT Devices TCP-AO cannot interoperate natively across NAT/NAPT devices, which modify the IP addresses and/or port numbers. We anticipate that traversing such devices will require variants of existing NAT/NAPT traversal mechanisms, e.g., encapsulation of the TCP-AO-protected segment in another transport segment (e.g., UDP), as is done in IPsec [RFC2766][RFC3947]. Such variants can be adapted for use with TCP-AO, or IPsec NAT traversal can be used instead in such cases [RFC3947]. 12. Evaluation of Requirements Satisfaction TCP-AO satisfies all the current requirements for a revision to TCP MD5, as summarized below [Be07]. 1. Protected Elements A solution to revising TCP MD5 should protect (authenticate) the following elements. This is supported - see Section 4.2. a. TCP pseudoheader, including IPv4 and IPv6 versions. Note that we do not allow optional coverage because IP addresses define a connection. If they can be coordinated across a NAT/NAPT, the sender can compute the MAC based on the received values; if not, a tunnel is required, as noted in Section 11.2. b. TCP header. Note that we do not allow optional port coverage because ports define a connection. If they can be coordinated across a NAT/NAPT, the sender can compute the MAC based on the received values; if not, a tunnel is required, as noted in Section 11.2. Touch Expires August 16, 2009 [Page 29] Internet-Draft The TCP Simple Authentication Option February 2009 c. TCP options. Note that TCP-AO allows exclusion of TCP options from coverage, to enable use with middleboxes that modify options (except when they modify TCP-AO itself). See Section 11. d. TCP payload data. 2. Option Structure Requirements A solution to revising TCP MD5 should use an option with the following structural requirements. This is supported - see Section 4.2. a. Privacy. The option should not unnecessarily expose information about the TCP-AO mechanism. The additional protection afforded by keeping this information private may be of little value, but also helps keep the option size small. TCP-AO exposes only the master key index, MAC, and overall option length on the wire. Note that short MACs could be obscured by using longer option lengths but specifying a short MAC length (this is equivalent to a different MAC algorithm, and is specified in the TSAD entry). See Section 4.2. b. Allow optional per connection. The option should not be required on every connection; it should be optional on a per connection basis. This is supported - see Sections 8.3, 8.4, and 8.5. c. Require non-optional. The option should be able to be specified as required for a given connection. This is supported - see Sections 8.3, 8.4, and 8.5. Touch Expires August 16, 2009 [Page 30] Internet-Draft The TCP Simple Authentication Option February 2009 d. Standard parsing. The option should be easily parseable, i.e., without conditional parsing, and follow the standard RFC 793 option format. This is supported - see Section 4.2. e. Compatible with Large Windows and SACK. The option should be compatible with the use of the Large Windows and SACK options. This is supported - see Section 8.6. The size of the option is intended to allow use with Large Windows and SACK. See also Section 2.1, which indicates that TCP-AO is 3 bytes shorter than TCP MD5 in the default case, assuming a 96-bit MAC. 3. Cryptography requirements A solution to revising TCP MD5 should support modern cryptography capabilities. a. Baseline defaults. The option should have a default that is required in all implementations. TCP-AO uses a default required algorithm as specified in [RFC- TBD], as noted in Section 4.2. b. Good algorithms. The option should use algorithms considered accepted by the security community, which are considered appropriately safe. The use of non-standard or unpublished algorithms should be avoided. TCP-AO uses MACs as indicated in [RFC-TBD]. The PRF is also specified in [RFC-TBD]. The PRF input string follows the typical design (in Section 6). Touch Expires August 16, 2009 [Page 31] Internet-Draft The TCP Simple Authentication Option February 2009 c. Algorithm agility. The option should support algorithms other than the default, to allow agility over time. TCP-AO allows any desired algorithm, subject to TCP option space limitations, as noted in Section 4.2. The TSAD allows separate connections to use different algorithms, both for the MAC and the PRF. d. Order-independent processing. The option should be processed independently of the proper order, i.e., they should allow processing of TCP segments in the order received, without requiring reordering. This avoids the need for reordering prior to processing, and avoids the impact of misordered segments on the option. This is supported - see Sections 8.3, 8.4, and 8.5. Note that pre-TCP processing is further required, because TCP segments cannot be discarded solely based on a combination of connection state and out-of-window checks; many such segments, although discarded, cause a host to respond with a replay of the last valid ACK, e.g. [RFC793]. See also the derivation of the ESN, which is reconstituted at the receiver using a demonstration algorithm that avoids the need for reordering (in Section 5). e. Security parameter changes require key changes. The option should require that the key change whenever the security parameters change. This avoids the need for coordinating option state during a connection, which is typical for TCP options. This also helps allow "bump in the stack" implementations that are not integrated with endpoint TCP implementations. TSAD parameters that should not change during a connection (by defininition, e.g., TCP connection ID, receiver TCP connection ID, TCP option exclusion list) cannot change. Other parameters change only when a master key is changed, using the master key tuple mechanism in the TSAD. See Section 7. Touch Expires August 16, 2009 [Page 32] Internet-Draft The TCP Simple Authentication Option February 2009 4. Keying requirements. A solution to revising TCP MD5 should support manual keying, and should support the use of an external automated key management system (e.g., a protocol or other mechanism). Note that TCP-AO does not specify a master key management system, but does indicate a proposed interface to the TSAD, allowing a completely separate master key system, as noted in Section 7. a. Intraconnection rekeying. The option should support rekeying during a connection, to avoid the impact of long-duration connections. This is supported by the KeyID and multiple master key tuples in a TSAD entry; see Section 7. b. Efficient rekeying. The option should support rekeying during a connection without the need to expend undue computational resources. In particular, the options should avoid the need to try multiple keys on a given segment. This is supported by the use of the KeyID. See Section 9. c. Automated and manual keying. The option should support both automated and manual keying. The use of a separate TSAD allows external automated and manual keying. See Section 9. This capability is enhanced by the generation of unique per-connection keys, which enables use of manual master keys with automatically generated connection keys as noted in Section 6. d. Key management agnostic. The option should not assume or require a particular key management solution. This is supported by use of a separate TSAD. See Section 9.1. Touch Expires August 16, 2009 [Page 33] Internet-Draft The TCP Simple Authentication Option February 2009 5. Expected Constraints A solution to revising TCP MD5 should also abide by typical safe security practices. a. Silent failure. Receipt of segments failing authentication must result in no visible external action and must not modify internal state, and those events should be logged. This is supported - see Sections 8.3, 8.4, and 8.5. b. At most one such option per segment. Only one authentication option can be permitted per segment. This is supported by the protocol requirements - see Section 4.2. c. Outgoing all or none. Segments out of a TCP connection are either all authenticated or all not authenticated. This is supported - see Section 8.4. d. Incoming all checked. Segments into a TCP connection are always checked to determine whether their authentication should be present and valid. This is supported - see Section 8.5. e. Non-interaction with TCP MD5. The use of this option for a given connection should not preclude the use of TCP MD5, e.g., for legacy use, for other connections. This is supported - see Section 10. Touch Expires August 16, 2009 [Page 34] Internet-Draft The TCP Simple Authentication Option February 2009 f. Optional ICMP discard. The option should allow certain ICMPs to be discarded, notably Type 3, Codes 2-4. This is supported - see Section 13. g. Maintain TCP connection semantics, in which the socket pair alone defines a TCP association and all its security parameters. This is supported - see Sections 7 and 11. 13. Security Considerations Use of TCP-AO, like use of TCP MD5 or IPsec, will impact host performance. Connections that are known to use TCP-AO can be attacked by transmitting segments with invalid MACs. Attackers would need to know only the TCP connection ID and TCP-AO Length value to substantially impact the host's processing capacity. This is similar to the susceptibility of IPsec to on-path attacks, where the IP addresses and SPI would be visible. For IPsec, the entire SPI space (32 bits) is arbitrary, whereas for routing protocols typically only the source port (16 bits) is arbitrary. As a result, it would be easier for an off-path attacker to spoof a TCP-AO segment that could cause receiver validation effort. However, we note that between Internet routers both ports could be arbitrary (i.e., determined a- priori out of band), which would constitute roughly the same off-path antispoofing protection of an arbitrary SPI. TCP-AO, like TCP MD5, may inhibit connectionless resets. Such resets typically occur after peer crashes, either in response to new connection attempts or when data is sent on stale connections; in either case, the recovering endpoint may lack the connection key required (e.g., if lost during the crash). This may result in time- outs, rather than more responsive recovery after such a crash. As noted in Section 6, such cases may also result in persistent TCP state for old connections that cannot be cleared, and so implementations should be capable of detecting an excess of such connections and clearing their state if needed to protect memory utilization [Je07]. TCP-AO does not include a fast decline capability, e.g., where a SYN- ACK is received without an expected TCP-AO option and the connection is quickly reset or aborted. Normal TCP operation will retry and timeout, which is what should be expected when the intended receiver is not capable of the TCP variant required anyway. Backoff is not Touch Expires August 16, 2009 [Page 35] Internet-Draft The TCP Simple Authentication Option February 2009 optimized because it would present an opportunity for attackers on the wire to abort authenticated connection attempts by sending spoofed SYN-ACKs without the TCP-AO option. TCP-AO is intended to provide similar protections to IPsec, but is not intended to replace the use of IPsec or IKE either for more robust security or more sophisticated security management. TCP-AO does not address the issue of ICMP attacks on TCP. IPsec makes recommendations regarding dropping ICMPs in certain contexts, or requiring that they are endpoint authenticated in others [RFC4301]. There are other mechanisms proposed to reduce the impact of ICMP attacks by further validating ICMP contents and changing the effect of some messages based on TCP state, but these do not provide the level of authentication for ICMP that TCP-AO provides for TCP [Go07]. >> A TCP-AO implementation MUST allow the system administrator to configure whether TCP will ignore incoming ICMP messages of Type 3 (destination unreachable) Codes 2-4 (protocol unreachable, port unreachable, and fragmentation needed - 'hard errors') intended for connections that match TSAD entries with non-NONE inbound MACs. An implementation SHOULD allow ignored ICMPs to be logged. This control affects only ICMPs that currently require 'hard errors', which would abort the TCP connection [RFC1122]. This recommendation is intended to be similar to how IPsec would handle those messages [RFC4301]. TCP-AO includes the TCP connection ID (the socket pair) in the MAC calculation. This prevents different concurrent connections using the same connection key (for whatever reason) from potentially enabling a traffic-crossing attack, in which segments to one socket pair are diverted to attack a different socket pair. When multiple connections use the same master key, it would be useful to know that packets intended for one ID could not be (maliciously or otherwise) modified in transit and end up being authenticated for the other ID. The ID cannot be zeroed, because to do so would require that the TSAD index was unique in both directions (ID->key and key->ID). That requirement would place an additional burden of uniqueness on master keys within endsystems, and potentially across endsystems. Although the resulting attack is low probability, the protection afforded by including the received ID warrants its inclusion in the MAC, and does not unduly increase the MAC calculation or master key management system. The use of any security algorithm can present an opportunity for a CPU DOS attack, where the attacker sends false, random segments that the receiver under attack expends substantial CPU effort to reject. Touch Expires August 16, 2009 [Page 36] Internet-Draft The TCP Simple Authentication Option February 2009 In IPsec, such attacks are reduced by the use of a large Security Parameter Index (SPI) and Sequence Number fields to partly validate segments before CPU cycles are invested validated the Integrity Check Value (ICV). In TCP-AO, the socket pair performs most of the function of IPsec's SPI, and IPsec's Sequence Number, used to avoid replay attacks, isn't needed in all cases due to TCP's Sequence Number, which is used to reorder received segments. TCP already protects itself from replays of authentic segment data as well as authentic explicit TCP control (e.g., SYN, FIN, ACK bits, but even authentic replays could affect TCP congestion control [Sa99]. TCP-AO does not protect TCP congestion control from such attacks due to the cumbersome nature of layering a windowed security sequence number within TCP in addition to TCP's own sequence number; when such protection is desired, users are encouraged to apply IPsec instead. Further, it is not useful to validate TCP's Sequence Number before performing a TCP-AO authentication calculation, because out-of-window segments can still cause valid TCP protocol actions (e.g., ACK retransmission) [RFC793]. It is similarly not useful to add a separate Sequence Number field to the TCP-AO option, because doing so could cause a change in TCP's behavior even when segments are valid. 14. IANA Considerations [NOTE: This section be removed prior to publication as an RFC] The TCP-AO option defines no new namespaces. The TCP-AO option requires that IANA allocate a value from the TCP option Kind namespace, to be replaced for TCP-IANA-KIND throughout this document. To specify MAC and PRF algorithms, TCP-AO refers to a separate document that may involve IANA actions [RFC-TBD]. 15. References 15.1. Normative References [RFC793] Postel, J., "Transmission Control Protocol," STD-7, RFC-793, Standard, Sept. 1981. [RFC1122] Braden, R., "Requirements for Internet Hosts -- Communication Layers," RFC-1122, Oct. 1989. Touch Expires August 16, 2009 [Page 37] Internet-Draft The TCP Simple Authentication Option February 2009 [RFC2018] Mathis, M., Mahdavi, J., Floyd, S. and A. Romanow, "TCP Selective Acknowledgement Options", RFC-2018, Proposed Standard, April 1996. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP-14, RFC-2119, Best Current Practice, March 1997. [RFC2385] Heffernan, A., "Protection of BGP Sessions via the TCP MD5 Signature Option," RFC-2385, Proposed Standard, Aug. 1998. [RFC2403] Madson, C., R. Glenn, "The Use of HMAC-MD5-96 within ESP and AH," RFC-2403, Proposed Standard, Nov. 1998. [RFC2460] Deering, S., Hinden, R., "Internet Protocol, Version 6 (IPv6) Specification," RFC-2460, Proposed Standard, Dec. 1998. [RFC3517] Blanton, E., Allman, M., Fall, K., and L. Wang, "A Conservative Selective Acknowledgment (SACK)-based Loss Recovery Algorithm for TCP", RFC-3517, Proposed Standard, April 2003. [RFC4306] Kaufman, C., "Internet Key Exchange (IKEv2) Protocol," RFC-4306, Proposed Standard, Dec. 2005. [RFC-TBD] Lebovitz, G., "MAC Algorithms for TCP-AO," RFC-TBD, date TBD. 15.2. Informative References [Be07] Eddy, W., (ed), S. Bellovin, J. Touch, R. Bonica, "Problem Statement and Requirements for a TCP Authentication Option," draft-bellovin-tcpsec-01, (work in progress), Jul. 2007. [Bo07] Bonica, R., B. Weis, S. Viswanathan, A. Lange, O. Wheeler, "Authentication for TCP-based Routing and Management Protocols," draft-bonica-tcp-auth-06, (work in progress), Feb. 2007. [Go07] Gont, F., "ICMP attacks against TCP," draft-ietf-tcpm-icmp- attacks-04, (work in progress), Oct. 2008. [Je07] Jethanandani, M., and M. Bashyam, "TCP Robustness in Persist Condition," draft-mahesh-persist-timeout-02, (work in progress), Oct. 2007. Touch Expires August 16, 2009 [Page 38] Internet-Draft The TCP Simple Authentication Option February 2009 [RFC1321] Rivest, R., "The MD5 Message-Digest Algorithm," RFC-1321, Informational, April 1992. [RFC1323] Jacobson, V., R. Braden, D. Borman, "TCP Extensions for High Performance," RFC-1323, May 1992. [RFC1948] Bellovin, S., "Defending Against Sequence Number Attacks," RFC-1948, Informational, May 1996. [RFC2104] Krawczyk, H., Bellare, M., Canetti, R., "HMAC: Keyed- Hashing for Message Authentication," RFC-2104, Informational, Feb. 1997. [RFC2766] Tsirtsis, G., Srisuresh, P., "Network Address Translation - Protocol Translation (NAT-PT)," RFC-2766, Proposed Standard, Feb. 2000. [RFC3234] Carpenter, B., S. Brim, "Middleboxes: Taxonomy and Issues," RFC-3234, Informational, Feb. 2002. [RFC3562] Leech, M., "Key Management Considerations for the TCP MD5 Signature Option," RFC-3562, Informational, July 2003. [RFC3947] Kivinen, T., B. Swander, A. Huttunen, V. Volpe, "Negotiation of NAT-Traversal in the IKE," RFC-3947, Proposed Standard, Jan. 2005. [RFC4301] Kent, S., K. Seo, "Security Architecture for the Internet Protocol," RFC-4301, Proposed Standard, Dec. 2005. [RFC4808] Bellovin, S., "Key Change Strategies for TCP-MD5," RFC-4808, Informational, Mar. 2007. [RFC4953] Touch, J., "Defending TCP Against Spoofing Attacks," RFC-4953, Informational, Jul. 2007. [Sa99] Savage, S., N. Cardwell, D. Wetherall, T. Anderson, "TCP Congestion Control with a Misbehaving Receiver," ACM Computer Communications Review, V29, N5, pp71-78, October 1999. [SDNS88] Secure Data Network Systems, "Security Protocol 4 (SP4)," Specification SDN.401, Revision 1.2, July 12, 1988. [To06] Touch, J., A. Mankin, "The TCP Simple Authentication Option," draft-touch-tcpm-tcp-simple-auth-03, (expired work in progress), Oct. 2006. Touch Expires August 16, 2009 [Page 39] Internet-Draft The TCP Simple Authentication Option February 2009 [Wa05] Wang, X., H. Yu, "How to break MD5 and other hash functions," Proc. IACR Eurocrypt 2005, Denmark, pp.19-35. [We05] Weis, B., "TCP Message Authentication Code Option," draft- weis-tcp-mac-option-00, (expired work in progress), Dec. 2005. 16. Acknowledgments Alfred Hoenes, Charlie Kaufman, and Adam Langley provided substantial feedback on this document. This document was prepared using 2-Word-v2.0.template.dot. Authors' Addresses Joe Touch USC/ISI 4676 Admiralty Way Marina del Rey, CA 90292-6695 U.S.A. Phone: +1 (310) 448-9151 Email: touch@isi.edu URL: http://www.isi.edu/touch Allison Mankin Johns Hopkins Univ. Washington, DC U.S.A. Phone: 1 301 728 7199 Email: mankin@psg.com URL: http://www.psg.com/~mankin/ Ronald P. Bonica Juniper Networks 2251 Corporate Park Drive Herndon, VA 20171 U.S.A. Email: rbonica@juniper.net Touch Expires August 16, 2009 [Page 40]