PANA Working Group Internet Draft D. Forsberg Nokia Y. Ohba Toshiba B. Patil Nokia H. Tschofenig Siemens A. Yegin DoCoMo USA Labs Document: draft-ietf-pana-pana-00.txt Expires: September 2003 March 2003 Protocol for Carrying Authentication for Network Access (PANA) Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of RFC2026. 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. Expires September 2003 1 PANA March 2003 Abstract This document defines the Protocol for Carrying Authentication for Network Access (PANA), a link-layer agnostic transport for Extensible Authentication Protocol (EAP) to enable network access authentication between clients and access networks. PANA can carry any authentication method that can be specified as an EAP method, and can be used on any link that can carry IP. PANA covers the client-to-network access authentication part of an overall secure network access framework, which additionally includes other protocols and mechanisms for service provisioning, access control as a result of initial authentication, and accounting. Table of Contents 1 Introduction...................................................2 2 Terminology....................................................3 3 Protocol Overview..............................................4 4 Protocol Details...............................................5 5 PANA Security Association Establishment.......................15 6 Authentication Method Choice..................................16 7 Filter Rule Installation......................................16 8 Data Traffic Protection.......................................17 9 Message Formats...............................................18 10 Open Issues...................................................18 11 Security Considerations.......................................18 12 References....................................................23 13 Acknowledgments...............................................25 14 Author's Addresses............................................25 15 Full Copyright Statement......................................35 1 Introduction Providing secure network access service requires access control based on the authentication and authorization of the clients and the access networks. Initial and subsequent client-to-network authentication provides parameters that are needed to police the traffic flow through the enforcement points. A protocol is needed to carry authentication methods between the client and the access network. IETF PANA Working Group has been chartered with the goal of designing a network-layer access authentication protocol. Link-layer authentication mechanisms are used as enablers of secure network access. A higher-layer authentication is deemed necessary when link-layer authentication mechanisms are either not available for lack of technology or deployment difficulties, or not able to meet the overall requirements, or when multi-layer (e.g., link-layer and network-layer) authentication is needed. Currently there is no standard network-layer solution for authenticating clients for network access. In the absence of such a solution, some inadequate standards-based solutions are deployed or non-standard ad-hoc Tschofenig (ed.) Expires September 2003 2 PANA March 2003 solutions are invented. [USAGE] Internet-Draft describes the problem statement in detail. Scope of this working group is identified as designing a link-layer agnostic transport for network access authentication methods. PANA Working Group has identified EAP [RFC2284] as the payload for this protocol and carrier for authentication methods. In other words, PANA will carry EAP which can carry various authentication methods. By the virtue of enabling transport of EAP above IP, any authentication method that can be carried as an EAP method is made available to PANA and hence to any link-layer technology. There is a clear division of labor between PANA, EAP and EAP methods. Defining new authentication methods, or deriving/distributing keys is outside the scope of PANA. Providing a secure channel that protects EAP and EAP methods against eavesdropping and spoofing is not an objective of the PANA design. While PANA is a fundamental part of a complete secure network access solution, its responsibility is limited to authentication and authorization of the client and the network. Providing access control is outside the scope of PANA. A separate provisioning protocol is needed for passing filtering information to access control nodes in the network. Additionally, mechanisms to provide data traffic protection in terms of authentication, integrity and replay protection, and encryption are outside the scope as well. Various environments and usage models for PANA are identified in the [USAGE] Internet-Draft. Potential security threats for network-layer access authentication protocol is discussed in [THREATS] draft. These two drafts have been essential in defining the requirements [PY+02] on the PANA protocol. Note that some of these requirements are imposed by the chosen payload, EAP [RFC2284]. This Internet-Draft makes an attempt for defining the PANA protocol based on the other drafts discussed above. Special care has been given to ensure the currently stated scope is observed and to keep the protocol as simple as possible. The current state of this draft is not complete, but it should be regarded as a work in progress. The authors made effort to capture the common understanding developed within the working group as much as possible. The design choices being made in this draft should not be considered as cast in stone. 2 Terminology This section describes some terms introduced in this document: PANA Session: PANA session is defined as the exchange of messages between the PaC and the PAA to authenticate a user(PaC) for network access. If the authentication is unsuccessful, the session is Tschofenig (ed.) Expires September 2003 3 PANA March 2003 terminated. The session is considered as active until there is a disconnect indication by the PaC or the PAA terminates it. Session Identifier: The device identifier is also used as the session identifier. This is used for indicating a disconnect or session revocation or for charging purposes. PANA Disconnect Indication: PANA session termination with explicit notification from a PaC sent to the PAA. The PDI also includes the session identifier. PANA Session Revocation: PANA session termination with explicit notification sent from the PAA to the PaC. The PSR includes the session identifier. PANA Security Association: The representation of the trust relation between the PaC and the PAA that is created at the end of the authentication phase (PH2). This security association includes the device identifier of the peer, and a shared key when available. The terms PaC, PAA, EP and Device Identifier can be found in [PY+02]. 3 Protocol Overview The PANA protocol involves two functional entities namely the PaC and the PAA. The EP, mentioned in the context with PANA, is a logical entity. There is, however, the option that the EP is not physically co-located with the PAA. In case that the PAA and the EP are co-located only an API is required instead of a separate protocol. In the case where the PAA is separated from the EP, a separate protocol will be used between the PAA and the EP for managing access control. The protocol and messaging between the PAA and EP for access authorization is outside the scope of this draft and will be dealt separately. The PANA protocol (PaC<->PAA) resides above the transport layer and the details are explained in Section 4.2. Although this document describes the interaction with a number of entities and with other protocol which enable network access authentication; the PANA protocol itself is executed between the PaC and the PAA. The protocol has three primary functions: 1. The PaC discovering the address of the PAA 2. The transport of EAP payloads between the PaC and the PAA 3. Access authorization by the PAA to the EP [Note that this aspect Tschofenig (ed.) Expires September 2003 4 PANA March 2003 is outside the scope of the PANA protocol.] The placement of the entities used in PANA largely depend on a certain architecture. The PAA may optionally interact with a AAA backend to authenticate the user (PaC). And in the case where the PAA and EP are co-located, step 3 mentioned above may not require a separate protocol. The figure below illustrates the interactions in a simplified manner: PaC EP PAA AAA --- --- --- --- PAA Discovery <---------------------o-----------------> (1) | PANA_REQUEST | ----------------------------------------> | AAA interaction |(2) -----------> | <----------- | PANA_RESPONSE | <--------------------------------------- | Authorization <----------------- (3) Figure 1: PANA Protocol The details of each of these aspects of the protocol are described in section 4 of this document. PANA supports authentication of a PaC using various EAP methods. The EAP method used depends on the level of security required for the EAP messaging itself. PANA does not secure the data traffic itself. However EAP methods that enable key exchange may allow other protocols to be bootstrapped for securing the data traffic. From a state machine aspect, PANA protocol consists of three phases 1. Discovery and initial handshake phase 2. Authentication phase 3. Termination phase In the first phase, an IP address of PAA is discovered and a PANA session is established between PaC and PAA. EAP messages are exchanged and a PANA SA is established in the second phase. The established PANA session as well as a PANA SA is deleted in the third phase. 4 Protocol Details Throughout the section, we use a notation "MESSAGENAME_ack" to represent a message which is used as an acknowledgment to a message "MESSAGENAME". In actual message formats, the two messages have the same message type and are distinguished by an acknowledgment flag (i.e., A-flag) in PANA header. Tschofenig (ed.) Expires September 2003 5 PANA March 2003 4.1. Common Processing Rules 4.1.1. Payload Encoding The payload of any PANA message consists of zero or more AVPs (Attribute Value Pairs). Brief description on the AVPs defined in this document is listed below. - Cookie AVP: contains a random value that is used for making initial handshake robust against blind resource consumption DoS attacks. - Data-Protection AVP: contains a flag which indicates if link-layer or network-layer ciphering should be initiated after PANA. - Device-Id AVP: contains a device identifier of the sender of the message. A device identifier is represented as a pair of device identifier type and device identifier value. Either a layer-2 address or an IP address is used for the device identifier value. - EAP AVP: contains an EAP PDU. - MAC AVP: contains a Message Authentication Code that protects a PANA message PDU. - Revocation-Status AVP: contains the reason of session revocation. 4.1.2. Transport Layer Protocol PANA uses UDP as its transport layer protocol. The UDP port number is TBD. All messages except for PANA_discovery are always unicast. PaC MAY use unspecified IP address for communicating with PAA. 4.1.3. Fragmentation PANA does not provide fragmentation of PANA messages. Instead, it relies on fragmentation provided by EAP methods and IP layer when needed. 4.1.4. Sequence Number and Retransmission PANA uses sequence numbers to provide ordered delivery of EAP messages. The design involves use of two sequence numbers to prevent some of the DoS attacks on the sequencing scheme. Every PANA packet include one transmitted sequence number (tseq) and one received sequence number (rseq) in the PANA header. See Appendix for detailed explanation on why two sequence numbers are needed. The two sequence number fields have the same length of N (TBD: possibly 32) bits and appear in PANA header. tseq starts from initial sequence number (ISN) and is monotonically increased by 1. The serial number arithmetic defined in [RFC1982] is used for Tschofenig (ed.) Expires September 2003 6 PANA March 2003 sequence number operation. The ISNs are exchanged between PaC and PAA during the discovery and initial handshake phase (see section "Discovery and Initial Handshake Phase"). The rules that govern the sequence numbers in other phases are described as follows. o When a message is sent, a new sequence number is placed on the tseq field of message regardless of whether it is sent as a result of retransmission or not. When a message is sent, rseq is copied from the tseq field of the last accepted message. o When a message is received, it is considered as valid in terms of sequence numbers if and only if (i) its tseq is greater than the tseq of the last accepted message and (ii) its rseq falls in the range between the tseq of the last acknowledged message + 1 and the tseq of the last transmitted message. PANA relies on EAP-layer retransmission for retransmitting EAP Request based on timer. Other PANA layer messages that require a response from the communicating peer are retransmitted based on timer at PANA-layer until a response is received (in which case the retransmission timer is stopped) or the number of retransmission reaches the maximum value (in which case the PANA session MUST be deleted immediately). For PANA-layer retransmission, the retransmission timer SHOULD be calculated as described in [RFC2988] to provide congestion control (TBD: default timer and maximum retransmission count suggestions). 4.1.5. Message Authentication Code A PANA message can contain a MAC (Message Authentication Code) AVP for cryptographically protecting the message. When a MAC AVP is included in a PANA message, the value field of the MAC AVP is calculated in the following way: MAC AVP value = PRF(PANA_MAC_Key, PANA_PDU) where PANA_PDU is the PANA message including the PANA header, with the MAC AVP value field first initialized to 0. The default algorithm used for the PRF function is TBD (possibly HMAC-SHA1). PANA_MAC_Key MUST be derived from the Master Session Key (MSK) and thus MUST be a part of the EAP key hierarchy [Ab02]. Detailed derivation algorithm is TBD. 4.1.6. Message Validity Check When a PANA message is received, the message is considered to be invalid at least when one of the following conditions are not met: o Each field in the message header contains a valid value including sequence number, message length, message type, version number, flags, etc. Tschofenig (ed.) Expires September 2003 7 PANA March 2003 o When a device identifier of the communication peer is bound to the PANA session, it matches the device identifier carried in MAC and/or IP header(s). o The message type is one of the expected types in the current state. o The message payload contains a valid set of AVPs allowed for the message type and there is no missing AVP that needs to be included in the payload. o Each AVP is decoded correctly. o When a MAC AVP is included, the AVP value matches the MAC value computed against the received message. o When a Device-Id AVP is included, the AVP is valid if the device identifier type contained in the AVP matches the expected one (this check is for PAA only) and the device identifier value contained in the AVP matches the value extracted from the lower-layer encapsulation header corresponding to the device identifier type contained in the AVP. Invalid messages MUST be discarded in order to provide robustness against DoS attacks and an unprotected. (TBD: in addition, a non-acknowledged error notification message MAY be returned to the sender.) 4.2. Discovery and Initial Handshake Phase When a PaC attaches to a network, and knows that it has to discover PAA for PANA, it can send a PANA_discovery message to a well-known link local multicast address (TBD) over UDP. The source address may be unspecified IP address if the PaC has not configured an address yet. In all PAA_discovery messages, both tseq and rseq fields of the header are set to zero (0). PANA PAA discovery assumes that PaC and PAA are one hop away from each other. If PaC knows the IP address of the PAA (some pre-configuration), it can unicast the PANA_discovery message to that address. PAA answers to the PANA_discovery message with a PANA_start message. When the PAA receives such a request, or upon receiving some lower layer indications of a new PaC, PAA can unicast a PANA_start message. The destination address may be unspecified IP address, but the L2 destination would be a unicast address (something for the implementations to deal with). This message announces the PAA to the PaC. There can be multiple PAAs on the link. The result does not depend on which PAA PaC chooses. By default PaC chooses the PAA that sent the first response. Tschofenig (ed.) Expires September 2003 8 PANA March 2003 PaC may also choose to start sending packets before getting authenticated. In that case, the network should detect this and send an unsolicited PANA_start message to PaC. EP is the node that can detect such activity. If EP and PAA are co-located, then an internal mechanism (e.g. API) between the EP module and the PAA module on the same host can prompt PAA to start PANA. In case they are separate, there needs to an explicit message to prompt PAA. Upon detecting the need to authenticate a client, EP can send a PAC_discovery message to the PAA on behalf of the PaC. This message carries a device identifier of the PaC in a Device-Id AVP. So that, PAA can send the unsolicited PANA_start message directly to the PaC. If the link between the EP and PAA is not secure, the PAC_discover message sent from EP to PAA MUST be protected by using, e.g., IPsec. PANA_start message contains a cookie carried in a Cookie AVP in the payload, respectively. The rseq field of the header is set to zero (0). The tseq field of the header contains the initial sequence number. The cookie is used for preventing the PAA from resource consumption DoS attacks by blind attackers. The cookie is computed in such a way as not to require any saved per-session state to recognize its valid cookie when a particular message sent by the PaC in response to the PANA_start message arrives. The exact algorithms and syntax used for generating cookies does not affect interoperability and hence is not specified here. An example algorithm is described below. Cookie = | HMAC_SHA1( | ) where is a randomly generated secret known only to the PAA, is an index used for choosing the secret for generating the cookie and '|' indicates concatenation. The secret-version should be changed frequently enough to prevent replay attacks. When a PaC receives the PANA_start message, it responds with a PANA_start message. The PANA_start message sent from the PaC contains the initial sequence numbers in the tseq and rseq fields of the PANA header, a copy of the received Cookie as the PANA payload. When the PAA receives the PANA_start message from the PaC, it verifies the cookie. The cookie is considered as valid if the received cookie has the expected value. If the computed cookie is valid, the protocol enters the authentication phase. Otherwise, it MUST silently discard the received message. PANA_start exchange is needed before entering authentication phase even when the PaC is pre-configured with PAAs IP address and the PANA_discover is a unicast message. PANA_start message sent from PAA is never retransmitted. PANA_start message sent from PaC is retransmitted based on timer in the same manner as other messages retransmitted at PANA-layer. Tschofenig (ed.) Expires September 2003 9 PANA March 2003 PaC PAA Message(tseq,rseq)[AVPs] ------------------------------------------------------ -----> PANA_discover(0,0) <----- PANA_start(x,0)[Cookie] -----> PANA_start(y,x)[Cookie] (continued to authentication phase) (PANA_discover sent by PaC) Figure 2: Example Sequence for Discovery and Initial Handshake Phase PaC EP PAA Message(tseq,rseq)[AVPs] ------------------------------------------------------ ---->o (Data packet arrival or L2 trigger) ------> PANA_discover(0,0)[Device-Id] <------------ PANA_start(x,0)[Cookie] ------------> PANA_start(y,x)[Cookie] (continued to authentication phase) (PANA_discover sent by EP) Figure 3: Example Sequence for Discovery and Initial Handshake Phase 4.3. Authentication Phase The main task in authentication phase is to carry EAP messages between PaC and PAA. All EAP messages except for EAP Success/Failure messages are carried in PANA_auth messages. When an EAP Success/Failure message is sent from a PAA, the message is carried in PANA_success or PANA_failure messages. PANA_success and PANA_failure messages are acknowledged with PANA_success_ack and PANA_failure_ack messages, respectively. It is possible to carry multiple EAP sequences in a single PANA sequence, with using a PANA_success or a PANA_failure message as a delimiter of each EAP sequence. An EAP Success or an EAP Failure message is carried in a PANA_success or a PANA_failure message, respectively. A single PANA session can enable more than one EAP authentication. This is used to satisfy the separate NAP and ISP authentications scenario. Each EAP authentication is delineated from the subsequent one with a PANA_success or PANA_failure message. F-flag in the PANA header indicates if this was the final authentication from sender's perspective. If PAA enables two separate authentication, it should not set F-flag in the PANA_success or PANA_failure message after the first EAP method. This indicates PAA's willingness to offer another authentication method for NAP-ISP separation. PaC can respond with a F-flag unset, indicating PaC's willingness to go through a second authentication method. PaC can optionally decline by setting the F-flag, and this concludes the PANA authentication. If PAA does not offer two levels of authentication, then it sets the F-flag even at Tschofenig (ed.) Expires September 2003 10 PANA March 2003 the end of first EAP method. In that case PaC has no other option but to set the F-flag to mark the end of PANA authentication. Currently, use of multiple EAP methods in PANA is designed only for NAP-ISP authentication separation. It is not for arbitrary EAP method sequencing, or giving PaC another chance when an authentication method fails. NAP and ISP authentication are considered completely independent. Presence or success of one should not effect the other. Making a decision based on the success or failure of each authentication is a network policy issue. A PANA_success or PANA_failure message is only qualified to signal the result of immediately preceding authentication method. When an EAP method that is capable of deriving keys is used during the authentication phase and the keys are successfully derived, PANA_success, PANA_success_ack, PANA_failure, PANA_failure_ack messages MUST contain a MAC AVP. The PANA_success and PANA_success_ack message exchange also is used for binding device identifiers of the PaC and PAA to the PANA SA. To achieve this, PANA_success and PANA_success_ack messages SHOULD contain a device identifier of the PAA and PaC, respectively, in a Device-Id AVP. The PaC MUST use the same type of device identifier as contained in the PANA_success message. In this case, the device identifier type contained in the PANA_success message indicates the device identifier type that the PaC needs to use. Validity check on the device identifier MUST be performed for these messages (see section "Message Validity Check"). PANA_success message MAY also contain a Data-Protection AVP to indicate if link-layer or network-layer ciphering should be initiated after PANA. A bit flag is the only information carried in the AVP and it states whether PANA SA will be used with link-layer ciphers (e.g., WEP) or network-layer ciphers (e.g, IKE and IPsec). It does not carry any other information specific to ciphering methods at all. When the information is preconfigured on PaC and PAA this AVP can be omitted. It is assumed that at least PAA is aware of the security capabilities of the access network. PANA protocol does not specify how the PANA SA and the Data-Protection AVP will be used to provide per-packet protection for data traffic. PANA_success and PANA_failure messages MUST be retransmitted based on the retransmission rule described in section "Sequence Number and Retransmission". Tschofenig (ed.) Expires September 2003 11 PANA March 2003 PaC PAA Message(tseq,rseq)[AVPs] ------------------------------------------------- (continued from discovery and initial handshake phase) <----- PANA_auth(x+1,y)[EAP{Request}] -----> PANA_auth(y+1,x+1)[EAP{Response}] . . <----- PANA_auth(x+2,y+1)[EAP{Request}] -----> PANA_auth(y+2,x+2)[EAP{Response}] <----- PANA_success(x+3,y+2) // F-flag set [EAP{Success}, Device-Id, Data-Protection, MAC] -----> PANA_success_ack(y+3,x+3) [Device-Id, MAC] // F-flag set Figure 4: Example Sequence in Authentication Phase 4.4. Re-authentication There are two types of re-authentication supported by PANA. The first type of re-authentication is based on EAP by entering the authentication phase. In this case, the discovery and initial handshake phase MAY be omitted. If there is an established PANA SA, PANA_auth messages MAY be protected by adding a MAC AVP to each message. The second type of re-authentication is based on a single protected message exchange without entering the authentication phase. PANA_reauth/PANA_reauth_ack messages are used for this purpose. If there is an established PANA SA, both PaC and PAA can send a PANA_reauth message to the communicating peer whenever it needs to make sure the availability of the PANA SA on the peer and expect the peer to return a PANA_reauth_ack message. Both PANA_reauth / PANA_reauth_ack messages MUST be protected with a MAC AVP. Implementations MUST limit the rate of performing re-authentication for both types of re-authentication. Tschofenig (ed.) Expires September 2003 12 PANA March 2003 PaC PAA Message(tseq,rseq)[AVPs] ------------------------------------------------------ -----> PANA_reauth(q,p)[MAC] <----- PANA_reauth_ack(p+1,q)[MAC] Figure 5: Example Sequence for PaC-initiated Re-authentication PaC PAA Message(tseq,rseq)[AVPs] ------------------------------------------------------ <----- PANA_reauth(p,q)[MAC] -----> PANA_reauth_ack(q+1,p)[MAC] Figure 6: Example Sequence for PAA-initiated Re-authentication 4.5. Termination Phase A procedure for explicitly terminating a PANA session can be initiated either from PaC (i.e., disconnect indication) or from PAA (i.e., session revocation). PANA_disconnect/PANA_disconnect_ack and PANA_revocation/PANA_revocation_ack message exchanges are used for disconnect indication and session revocation procedures, respectively. A PANA_revocation message contains the reason of revocation in Revocation-Status AVP. When there is an established PANA SA established between the PaC and PAA, all messages exchanged during the termination phase MUST be protected with a MAC AVP. When the sender of PANA_disconnect or PANA_revocation message receives a valid acknowledgment, all states maintained for the PANA session MUST be deleted immediately. PaC PAA Message(tseq,rseq)[AVPs] ------------------------------------------------------ -----> PANA_disconnect(q,p)[MAC] <----- PANA_disconnect_ack(p+1,q)[MAC] Figure 7: Example Sequence for Disconnect Indication PaC PAA Message(tseq,rseq)[AVPs] ------------------------------------------------------ <----- PANA_revocation(p,q)[Revocation-Status,MAC] -----> PANA_revocation_ack(q+1,p)[MAC] Figure 8: Example Sequence for Session Revocation 4.6. Illustration of a Complete Message Sequence A complete PANA message sequence is illustrated in Figure 5.6. The example assumes the following scenario. Tschofenig (ed.) Expires September 2003 13 PANA March 2003 - PaC multicasts PANA_discover message - ISNs used by PAA and PaC are x and y, respectively. - A single EAP sequence is used in authentication phase. - A single EAP authentication method is used in the EAP sequence. - The EAP authentication method derives keys. PANA SA is established based on the keys. - After PANA SA is established, all messages are integrity protected with MAC AVP. - Re-authentication based on PANA_reauth/PANA_reauth_ack exchange is performed. - PANA session is terminated as a result of disconnect indication from PaC. PaC PAA Message(tseq,rseq)[AVPs] ----------------------------------------------------- // Discovery and initial handshake phase -----> PANA_discover(0,0) <----- PANA_start(x,0)[Cookie] -----> PANA_start(y,x)[Cookie] // Authentication phase <----- PANA_auth(x+1,y)[EAP] -----> PANA_auth(y+1,x+1)[EAP] <----- PANA_auth(x+2,y+1)[EAP] -----> PANA_auth(y+2,x+2)[EAP] <----- PANA_success(x+3,y+2) // F-flag set [EAP, Device-Id, Data-Protection, MAC] -----> PANA_success_ack(y+3,x+3) // F-flag set [Device-Id, MAC] // Re-authentication <----- PANA_reauth(x+4,y+3)[MAC] -----> PANA_reauth_ack(y+4,x+4)[MAC] // Termination phase -----> PANA_disconnect(y+5,x+4)[MAC] <----- PANA_disconnect_ack(x+5,y+5)[MAC] Figure 9: A Complete Message 4.7. Device ID choice PaC has to pick a device identifier to provide for PANA exchanges. In this version of the specification, device ID is considered to be Tschofenig (ed.) Expires September 2003 14 PANA March 2003 fixed. Future versions might enable changing it during a PANA session. A PaC will configure an IP address before PANA if it can. It might either have a pre-configured IP address, or have to obtain one via dynamic methods such as DHCP or stateless address autoconfiguration. Dynamic methods may or may not succeed depending on the local security policy. In networks where the PaCs need to use PANA prior to address configuration, EPs will detect the PaCs attempt to get IP address and help PAA to initiate authentication. Either an IP address or link-layer address should be used as device DI at any time. The only case an IP address should be used as device ID is when IPsec will be used for protecting data traffic after initial authentication. Any other time a link-layer address can be used by both PAA and PaC as device ID. It is assumed that PAA knows the security mechanisms being provided or required on the access network (e.g., physical security, link-layer ciphers prior to PANA, link-layer ciphers enabled after PANA, IPsec). When IPsec is the choice of data ciphering, PAA should provide its IP address as device ID, and expect the PaC to provide its IP address if it has one. In all other cases, link-layer addresses can be provided from both sides. [TBD: can we allow IP address allocation after PANA and still be able to use IPsec?] 4.8. PaC Implications - PaC state machine. [TBD] 4.9. PAA Implications - PAA state machine. [TBD] 5 PANA Security Association Establishment When PANA is used over an already established secure channel, such as physically secured wires or ciphered link-layers, we can reasonably assume that man-in-the-middle attack or service theft is not possible [THREATS]. Anywhere else where there is no secure channel prior to PANA, the protocol needs to protect itself against such attacks. The device identifier that is used during the authentication needs to be verified at the end of the authentication to prevent service theft and DoS attacks. Additionally, a free loader should be prevented from spoofing data packets by using the device identifier of an already authorized legitimate client. Both of these requirements necessitate generation of a security association between the PaC and the PAA at the end of the authentication. This can only be done when the authentication method used can generate cryptographic keys. Use of secret keys can prevent attacks which would otherwise be very easy to launch by eavesdropping on and spoofing traffic over a insecure links. Tschofenig (ed.) Expires September 2003 15 PANA March 2003 PANA relies on EAP and the EAP methods to provide a session key in order to establish a PANA security association. An example of such a method is EAP-TLS [EAPTLS], whereas EAP-MD5 [RFC2284] is an example of a method that cannot create such keying material. The choice of EAP method becomes important, as already discussed in the next section. This keying material is already used within PANA during the final handshake. This handshake ensures that the device identifier that is bound to the PaC at the end of the authentication process is not coming from a man-in-the-middle, but from the legitimate PaC. Knowledge of the same keying material on both PaC and the PAA helps prove this. The other use of the keying material will be discussed in sections 7 and 8. 6 Authentication Method Choice Authentication methods' capabilities and therefore applicability to various environments differ among them. Not all methods provide support for mutual authentication, key derivation or distribution, and DoS attack resiliency that are necessary for operating in insecure networks. Such networks might be susceptible to eavesdropping and spoofing, therefore a stronger authentication method needs to be used to prevent attacks on the client and the network. The authentication method choice is a function of the underlying security of the network (e.g., physically secured, shared link, etc.). It is the responsibility of the user and the network operator to pick the right method for authentication. PANA carries EAP regardless of the EAP method used. It is outside the scope of PANA to mandate, recommend, or limit use of any authentication methods. PANA cannot increase the strength of a weak authentication method to make it suitable for an insecure environment. There are some EAP- based approaches to achieve this goal [PEAP][TTLS]. PANA can carry these EAP encapsulating methods but it does not concern itself with how they achieve protection for the weak methods (i.e., their EAP method payloads). 7 Filter Rule Installation PANA protocol provides client authentication and authorization functionality for securing network access. The other component of a complete solution is the access control which ensures that only authenticated and authorized clients can gain access to the network. PANA enables access control by identifying legitimate clients and generating filtering information for access control mechanisms. Getting this filtering information to the EPs (enforcement points) and performing filtering are outside the scope of PANA. Access control can be achieved by placing EPs in the network for policing the traffic flow. EPs should prevent data traffic from and to any unauthorized client unless it's PANA traffic. When a client is authenticated and authorized, PAA should notify EP(s) and ask for Tschofenig (ed.) Expires September 2003 16 PANA March 2003 changing filtering rules to allow traffic for a recently authorized client. There needs to be a protocol between PAA and EP(s) when these entities are not co-located. PANA Working Group will not be defining a new protocol for this interaction. Instead, it will (preferably) identify one of the existing protocols that can fit the requirements. Possible candidates include but not limited to COPS, SNMP, DIAMETER. This task is similar to what MIDCOM Working Group is trying to achieve, therefore some of the MIDCOM's output might be useful here. EPs location in the network topology should be appropriate for performing access control functionality. The closest IP-capable access device to the client devices is the logical choice. PAA and EPs on an access network should be aware of each other as this is necessary for access control. Generally this can be achieved by manual configuration. Dynamic discovery is another possibility, but this is clearly outside the scope of PANA. Filtering rules generally include device identifiers for a client, and also cryptographic keying material when needed. Such keys are needed when attackers can eavesdrop and spoof on the device identifiers easily. They are used with link-layer or network-layer ciphering to provide additional protection. For issues regarding data-origin authentication see Section 8. 8 Data Traffic Protection Protecting data traffic of authenticated and authorized clients from others is another component of providing a complete secure network access solution. Authentication, integrity and replay protection of data packets are needed to prevent spoofing when the underlying network is not physically secured. Encryption is needed when eavesdropping is a concern in the network. When the network is physically secured, or the link-layer ciphering is already enabled prior to PANA, data traffic protection is already in place. In other cases, enabling link-layer ciphering or network- layer ciphering might rely on PANA authentication. The user and network have to make sure an appropriate EAP method that can generate required keying materials is used. Once the keying material is available, it needs to be provided to the EP(s) for use with ciphering. Network-layer ciphering, i.e., IPsec, can be used when data traffic protection is required but link-layer ciphering capability is not available. Note that a simple shared secret generated by an EAP method is not readily usable by IPsec for authentication and encryption of IP packets. Fresh and unique session key derived from the EAP method is still insufficient to produce an IPsec SA since both traffic selectors and other IPsec SA parameters are missing. The shared secret can be used in conjunction with a key management protocol like IKE [RFC2409] to turn a simple shared secret into the required IPsec SA. The details of this mechanism is outside the scope of PANA protocol, and it can be outlined in a separate Tschofenig (ed.) Expires September 2003 17 PANA March 2003 Internet-Draft. PANA provides bootstrapping functionality for such a mechanism by carrying EAP methods that can generate initial keying material. Using network-layer ciphers should be regarded as a substitute for link-layer ciphers when the latter is not available. IKE involves several message exchanges which can incur additional delay in getting basic IP connectivity for a mobile device. Such a latency is inevitable when there is no other alternative and this level of protection is required. Network-layer ciphering can also be used in addition to link-layer ciphering if the added benefits outweigh its cost to the user and the network. 9 Message Formats Bits and bytes on the wire... 10 Open Issues The following list describes some open issues for PANA: - Should the PANA protocol provide downgrade protection? - How extensible or flexible should the device identifier be? - Should the PANA protocol support a modify message to be able to alter state? This would, for example, be useful in case of IP address change without mobility (e.g. in IPv6 for privacy reasons). - The PANA SA needs a session key and either this session key is derived from the EAP method as part of the EAP key derivation framework or within PANA. 11 Security Considerations The PANA protocol provides ordered delivery for EAP messages. If an EAP method that provides session keys is used, a PANA SA is created. The EAP Success/Failure message is one of the signaling messages which is integrity protected with this PANA SA. The PANA protocol does not provide security protection for the initial EAP message exchange. Integrity protection can only be provided after the PANA SA has been established. Thus, PANA re-authentication, revocation and disconnect notifications can be authenticated, integrity and replay protected. In certain environments (e.g. on a shared link) the EAP method selection is an important issue. The PANA framework described in this document covers the discussion of different protocols which are of interest for a protocol between the PaC and the PAA (typically referred as the PANA protocol). The PANA itself consists of a sequence of steps which are executed to complete the network access authentication procedure. Some of these steps are optional. The following execution steps have been identified as being relevant for PANA. They security considerations will be discussed in detail subsequently. Tschofenig (ed.) Expires September 2003 18 PANA March 2003 a) Discovery message exchange In general it is difficult to prevent a vulnerabilities of the discovery protocol since the initial discovery are unsecured. To prevent very basic attacks an adversary should not be able to cause state creation with discovery messages at the PAA. This is prevented by re-using a cookie concept (see [RFC2522]) which allows the responder to be stateless in the first message exchange. Because of the architectural assumptions made in PANA (i.e. the PAA is the on the same link as the PaC) the return-routability concept does not provide additional protection. Hence it is difficult to prevent this threat entirely. Furthermore it is not possible to shift heavy cryptographic operations to the PaC at the first few messages since the computational effort depends on the EAP method. The usage of client-puzzles as introduced by P. Nikander et. al. in [AN+00] is under investigation. Resistance against blind DoS attacks (i.e. attacks by off-path adversaries) is achieved with sequence numbers and cookies. Since PAA and PaC are one IP hop away from each other, PANA messages can be filtered whenever messages arrive at interfaces where they are not expected. b) EAP over PANA message exchange The EAP derived session key is used to create a PANA security association. Since the execution of an EAP method might require a large number of roundtrips and no other session key is available it is not possible to secure the EAP message exchange itself. Hence an adversary can both eavesdrop the EAP messages and is also able to inject arbitrary messages which might confuse both the PaC and the PAA. The threats caused by this ability heavily depend on the EAP state machine. Since especially the PAA is not allowed to discard packets and packets have to be stored or forwarded to an AAA infrastructure some risk of DoS attacks exists. Eavesdropping EAP packets might cause problems when (a) the EAP method is weak and enables dictionary or replay attacks or even allows an adversary to learn the long-term password directly. Furthermore, if the optional EAP Identity payload is used then it allows the adversary to learn the identity of the PaC. In such a case a privacy problem is prevalent. To prevent these threats Section 6 suggests using proper EAP methods for particular environments. Depending on the usage environment an EAP authentication has to be used for example which supports user identity confidentiality, protection against dictionary attacks and session key establishment. It is therefore the responsibility of the network operators and end users to choose the proper EAP method. Tschofenig (ed.) Expires September 2003 19 PANA March 2003 PANA does not protect the EAP method exchange, but provides ordered delivery with sequence numbers. Sequence numbers and cookies provide resistance against blind DoS attacks. c) PANA SA establishment Once the EAP message authentication is finished a fresh and unique session key is available to the PaC and the PAA. This assumes that the EAP method allows session key derivation and that the generated session key has a good quality. For further discussion about the importance of the session key generation refer to the next subsection (c) about compound authentication. The session key available for the PaC is established as part of the authentication and key exchange procedure of the selected EAP method. The PAA obtains the session key via the AAA infrastructure (if used). Draft [CFB02] describes how a session key is securely carried (i.e. CMS protected) between AAA servers. Security issues raised with this session key transport are described in [WHC02]. The establishment of a PANA SA is required in environments where no physical or link layer security is available. The PANA SA allows subsequently exchanged messages to experience cryptographic protection. For the current version of the document an Integrity object is defined which is based on Diameter objects. The Integrity Object supports data-origin authentication, replay protection based on sequence numbers and integrity protection based on a keyed message digest. Confidentiality protection is not provided. The session keys (one for each direction) used for this object has to be provided by the EAP method. For this version of the document it is assumed that no negotiation of algorithms and parameters takes place. Instead HMAC-SHA1 is used per-default. A different algorithm such as HMAC-MD5 might be used as an option. The used algorithm is indicated in the header of the Integrity object. To select the security association for signaling message protection the Session ID. The keyed message digest included in the Integrity object will include all fields of the PANA signaling message including the sequence number field of the packet. The protection of subsequent signaling messages prevents an adversary from acting as a man-in-the-middle adversary, from injecting packets, from replaying messages and from modifying the content of the exchanged packets. This prevents subsequently described threats. If an entity (PAA or PaC) looses its state (especially the current sequence number) then the entire PANA protocol has to be restarted. No re-synchronization procedure is provided. The lifetime of the PANA SA has to be bound to the refresh interval with an additional tolerance period. To provide fast re- authentication a separate security association (e.g. one stored at the local AAA server) should be used. By fast re-authentication we mean a new PANA protocol execution which does not involve the entire AAA communication. The ability to trigger such a protocol execution Tschofenig (ed.) Expires September 2003 20 PANA March 2003 depends on the given EAP method and on the policy of the local network requesting authentication. d) Enabling weak legacy authentication methods in insecure networks Some of the authentication methods are not strong enough to be used in insecure networks where attackers can easily eavesdrop and spoof on the link. They may not be able to produce much needed keying material either. An example would be using EAP-MD5 over wireless links. Use of such legacy methods can be enabled by carrying them over a secure channel. There are EAP methods which are specifically designed for this purpose, such as EAP-TTLS [TTLS] and PEAP [PEAP]. PANA can carry these EAP tunneling methods which can carry the legacy methods. PANA does not do anything special for this case. The EAP tunneling method will have to produce keying material for PANA SA when needed. There are certain MitM vulnerabilities with tunneling EAP methods [MITM]. Solving these problems are outside the scope of PANA. e) Preventing downgrading attacks EAP supports a number of different EAP methods for authentication and therefore it might be required to agree on a specific mechanism. An unprotected negotiation mechanism is supported in EAP and a secure negotiation procedure for the GSS-API methods. The support of the GSS-API as an EAP method is described in [AS02]. A protected negotiation is supported by the GSS-API with RFC 2478 [RFC2478]. If desired, such a protection can also be offered by PANA by repeating the list of supported EAP methods protected with the PANA SA. This type of protection is similar to the protected negotiation described in [RFC3329]. This issue requires further investigation especially since the EAP protocol runs in most cases different endpoints than the PANA protocol. f) Device Identifier exchange As part of the authorization procedure a Device Identifier has to be installed at the EP by the PAA. The PaC provides the Device Identifier information to the PAA secured with the PANA SA. Section 6.2.4 of [THREATS] describes a threat where an adversary modifies the Device Identifier to gain unauthorized access to the network. The installation of the Device Identifier at the EP (independently whether the EP is co-located with the PAA or not) has to be accomplished in a secure manner. These threats are, however, not part of the PANA protocol itself since the protocol is not PANA specific. g) Triggering a data protection protocol Recent activities in the EAP working group try to create a common framework for key derivation which is described in [Ab02]. This Tschofenig (ed.) Expires September 2003 21 PANA March 2003 framework is also relevant for PANA in various ways. First, a PANA security association needs to be created. Additionally it might be necessary to trigger a protocol which allows link layer and network layer data protection to be established. As an example see Section 1 of [Ab02] with [802.11i] and [802.11] as an example. Furthermore, a derived session key might help to create the pre-requisites for network layer protection (for example IPsec). As motivated in Section 6.4 of [THREATS] it might be necessary to establish either a link layer or a network layer protection to prevent certain thefts in certain scenarios. Threats specific to the establishment of a link layer or a network layer security association are outside the scope of PANA. The interested reader should refer to the relevant working groups such as IPsec or Midcom. h) Periodic refresh messages Network access authentication is done for a very specific purpose and often charging procedures are involved which allow restricting network resource usage based on some policies. In mobility environments it is always possible that an end host suddenly disconnects without transmitting a disconnect message. If network access authentication as part of PANA is executed only at the beginning then an adversary can gain advantage of the installed packet filters to submit and receive data packets. Also for the network operator it might be desirable to enforce a disconnect based on some external events (e.g. because of insufficient funds, etc.). An additional motivation for detecting a disconnected end host is the ability to release resources (i.e. garbage collection). The PAA can remove per-session state information including installed security association, packet filters etc. Different procedures can be used for disconnect indication. PANA cannot assume link layer disconnect indication. Hence this functionality has to be provided at a higher layer. With this version of the draft we suggest to apply the soft-state principle found at other protocols (such as RSVP [RFC2205]). Soft-state means that session state is kept alive as long as refresh messages refresh the state. If no new refresh messages are provided then the state automatically times out and resources are released. This process includes stopping accounting procedures. Based on the different environments where PANA could be used it is difficult to fix a refresh interval. Hence a default refresh interval of 30 seconds is suggested. Additionally there is the possibility to negotiation this interval once the PANA security association is established. A policy at the PAA and the PaC would ensure that the refresh interval is selected with a value which is either too high or too low. There is certainly a tradeoff between Tschofenig (ed.) Expires September 2003 22 PANA March 2003 the refresh interval and the bandwidth consumption. To reduce the bandwidth consumption a small PANA message consisting only of a session identifier and the Integrity object is used. The session identifier refers to the state that has to be refreshed. Some environments do not need PANA refresh messages to detect orphan states. For these environments the refresh interval should be set to zero which effectively disables the usage of refresh messages. In case of IPsec protection a dead-peer mechanism can be used to detect inactivity (see [HBR03]). Refresh messages are sent from the PaC to the PAA. From a security point of view an adversary must not be able to inject, modify or replay refresh messages nor must he be able to change the refresh interval (e.g. setting it to zero) without detection. Hence these messages experience cryptographic protection. i) Tear-Down message The PANA protocol supports the ability for both the PaC and the PAA to transmit a tear-down message. This message causes state removal, a stop of the accounting procedure and removes the installed packet filters. It is obvious that such a message must be protected to prevent an adversary from deleting state information and thereby causing denial of service attacks. 12 References [802.11] I. S. 802.11-1997, "Information technology - telecommunications and information exchange between systems - local and metropolitan area networks - specific requirements part 11: Wireless lan medium access control (mac) and physical layer (phy) specifications," tech. rep., 1997. [RFC2522] P. Karn and W. Simpson, "Photuris: Session-key management protocol," RFC 2522, Internet Engineering Task Force, Mar. 1999. [PEAP] H. Andersson, S. Josefsson, G. Zorn, et al. , "Protected extensible authentication protocol (PEAP)," Internet Draft, Internet Engineering Task Force, Feb. 2002. Work in progress. [Ab02] B. Aboba, "The EAP session key problem," Internet Draft, Internet Engineering Task Force, Feb. 2002. Work in progress. [802.11i] I. D. 802.11i/D2, "Draft supplement to standard for telecommunications and information exchange between systems - lan/man specific requirements - part 11: Wireless medium access control (mac) and physical layer (phy) specifications: Specification for enhanced security," tech. rep., 2001. [AS02] Aboba, B., Simon, D.: "EAP GSS Authentication Protocol", , (work in progress), April, 2002. Tschofenig (ed.) Expires September 2003 23 PANA March 2003 [CFB02] P. Calhoun, S. Farrell, and W. Bulley, "Diameter CMS security application," Internet Draft, Internet Engineering Task Force, Mar. 2002. Work in progress. [RFC2284] Blunk, L. and J. Vollbrecht, "PPP Extensible Authentication Protocol (EAP)", RFC 2284, March 1998. [RFC2716] Aboba, B., and D. Simon, "PPP EAP TLS Authentication Protocol", RFC 2716, October 1999. [HBR03] G. Huang, S. Beaulieu, and D. Rochefort, "A traffic-based method of detecting dead ike peers," internet draft, Internet Engineering Task Force, 2003. Work in progress. [RFC2409] Harkins, D. and D. Carrel, "The Internet Key Exchange (IKE)", RFC 2409, November 1998. [IKEv2] Kaufman, C.: "Internet Key Exchange (IKEv2) Protocol", , (work in progress), February, 2003. [MITM] N. Asokan, V. Niemi, and K. Nyberg, "Man-in-the-middle in tunneled authentication," in http://eprint.iacr.org/2002/163/ , 2002. [PANATLS] Y. Ohba, S. Baba, and S. Das, "Pana over tls," Internet Draft, Internet Engineering Task Force, 2002. Work in progress. [PEAP] H. Andersson, S. Josefsson, G. Zorn, et al. , "Protected [PIC] Y. Sheffer, H. Krawczyk, and B. Aboba, "PIC, a pre-IKE credential provisioning protocol," Internet Draft, Internet Engineering Task Force, Feb. 2002. Work in progress. [PL+03] J. Puthenkulam, V. Lortz, A. Palekar, D. Simon, and B. Aboba, "The compound authentication binding problem," internet draft, Internet Engineering Task Force, 2003. Work in progress. [AN+00] Aura, T., Nikander, P., Leiwo, J.: "DOS-resistant Authentication with Client Puzzles", in "Proc. Security Protocols Workshop 2000, Cambridge, UK", 2000. [PY+02] Penno, R., Yegin, A., Ohba, Y., Tsirtsis, G., Wang, C.: "Protocol for Carrying Authentication for Network Access (PANA) Requirements and Terminology", Internet-Draft, , (work in progress), October, 2002. [RFC2284bis] Blunk, L., Vollbrecht, J., Aboba, B., Carlson, J.: "Extensible Authentication Protocol (EAP)", < , (work in progress), January, 2003. [RFC1982] Elz, R., Bush, R.: "Serial Number Arithmetic", RFC 1982, August 1996. Tschofenig (ed.) Expires September 2003 24 PANA March 2003 [RFC2205] Braden, R., Zhang, L., Berson, S., Herzog, S., Jamin, S.: ƒResource ReSerVation Protocol (RSVP) Ï Version 1 Functional Specification", RFC 2205, September 1997. [RFC2478] E. Baize and D. Pinkas, "The simple and protected GSS-API negotiation mechanism," RFC 2478, Internet Engineering Task Force, Dec. 1998. [RFC2988] Paxson, V., Allman, M.: "Computing TCP's Retransmission Timer", RFC 2988, November, 2000. [RFC3329] Arkko, J., Torvinen, V., Camarillo, G., Niemi, A., Haukka, T.: "Security Mechanism Agreement for the Session Initiation Protocol (SIP)", RFC 3329, January, 2003. [SENAA] D. Forsberg and J. Rajahalme, "Secure network access authentication (senaa)," Internet Draft, Internet Engineering Task Force, 2002. Work in progress. [THREATS] Parthasarathy, M.: "PANA Threat Analysis and security requirements", , (work in progress), January, 2003. [TTLS] P. Funk and S. Blake-Wilson, "EAP tunneled TLS authentication protocol (EAP-TTLS)," Internet Draft, Internet Engineering Task Force, Mar. 2002. Work in progress. [USAGE] Ohba, Y., Das, S., Patil, B., Soliman, H., Yegin, A.: "Problem Statement and Usage Scenarios for PANA", , (work in progress), February, 2003. [WHC02] J. Walker, R. Housley, and N. Cam-Winget, "AAA key distribution," Internet Draft, Internet Engineering Task Force, Apr. 2002. Work in progress. 13 Acknowledgments Place your name here 14 Author's Addresses Basavaraj Patil Nokia 6000 Connection Dr. Irving, TX. 75039 USA Phone: +1 972-894-6709 Email: Basavaraj.Patil@nokia.com Dan Forsberg Nokia Research Center P.O. Box 407 FIN-00045 NOKIA GROUP, Finland Tschofenig (ed.) Expires September 2003 25 PANA March 2003 Phone: +358 50 4839470 EMail: dan.forsberg@nokia.com Alper E. Yegin DoCoMo USA Labs 181 Metro Drive, Suite 300 San Jose, CA, 95110 USA Phone: +1 408 451 4743 Email: alper@docomolabs-usa.com Yoshihiro Ohba Toshiba America Research, Inc. P.O. Box 136 Convent Station, NJ, 07961-0136 USA Phone: +1 973 829 5174 Email: yohba@tari.toshiba.com Hannes Tschofenig Siemens Corporate Technology Otto-Hahn-Ring 6 81739 Munich Germany Email: Hannes.Tschofenig@siemens.com Appendix A. Adding sequence number to PANA for carrying EAP A.1. Why is sequence number needed for PANA to carry EAP? EAP [RFC2284bis] requires underlying transports to provide ordered-delivery of messages. If an underlying transport does not satisfy the ordering requirement, the following situation could happen: EAP Peer EAP Authenticator -------------------------------------------- 1. (got req 1) <------- Request ID=1 2. Response ID=1 ---+ | (timeout) 3. | +-- Request ID=1 | | +-|--> (got resp 1) 4. (got req 2) <----|-- Request ID=2 | 5. Response ID=2 -----|--> (got resp 2) | 6. (got req 1) <----+ 7. Response ID=1 --------> [discarded due to unexpected ID] Figure A.1 Undesirable scenario Tschofenig (ed.) Expires September 2003 26 PANA March 2003 In Figure A.1, the second EAP Request message with Identifier=1 arrives at the EAP peer after the third EAP Request message with Identifier=2. As a result, the EAP peer accepts the second EAP Request as a new EAP Request while it is just an old EAP Request that was already responded and the authentication might be totally messed up. This problem occurs due to the fact that EAP doesn't recognize duplicate packets in the scope of one EAP protocol run, but only in the scope of current and previous packet (i.e., request and response message matching). When EAP is running over PPP or IEEE 802 links, this is not a problem, because those link-layers have the ordering invariant characteristic. On the other hand, the PANA design has chosen UDP as its transport. Given that UDP does not provide ordered delivery of packets and PANA does not assume any specific link-layer technology to carry EAP, PANA messages need to have a sequence number. In the following text we describe two possible approaches for sequence number handling in PANA. The first one makes use of a single sequence number whereas the latter utilizes two. Finally a comparison between the two approaches is provided. The method described in Section A.3.1. (i.e., the dual sequence number with orderly-delivery method) is suggested as the preferred method for PANA transport. A.2. Single sequence number approach This section discusses several methods based on using a single sequence number for providing orderly message delivery. Sequence number handling for all methods discussed in Section A.2 must comply to the following rules: Rule 1: The sequence number starts from initial sequence number (ISN) and is monotonically increased by 1. The arithmetic defined in [RFC1982] is used for sequence number operation. Rule 2: When a PAA sends an EAP message passed from EAP layer to a PaC, a new sequence number is placed in the message, regardless of whether it is sent as a result of a retransmission at the EAP layer or not. Note: It might be possible to define other mechanisms for sequence number handling if it can be assumed that a PAA detects EAP retransmissions. However, such an assumption heavily depends on EAP implementation details in particular on EAP APIs, thus it was decided not to use such an assumption. A.2.1. Single sequence number with EAP retransmission method Again, the following rules must hold: Tschofenig (ed.) Expires September 2003 27 PANA March 2003 Rule 3: Use EAP layer retransmission for retransmitting EAP messages (based on a timer expiration). Rule 4: When the PaC receives a message from the PAA, it checks the sequence number and discards the message if the sequence number is not greater than that of the last accepted message. Rule 5: When the PAA receives a message from the PaC, it checks the sequence number and discards the message if the sequence number does not match a pending request message. PaC PAA Seq# Message -------------------------------------------- 1. <------- (x) PANA_auth[EAP Req ID=1] 2. ---+ (x) PANA_auth[EAP Res ID=1] | (retransmission timeout at EAP-layer) 3. | +-- (x+1) PANA_auth[EAP Req ID=1] | | +-|--> (discarded due to Rule 5) | (retransmission timeout at EAP-layer) 4. <----|-- (x+2) PANA_auth[EAP Req ID=1] | 5. -----|--> (x+2) PANA_auth[EAP Res ID=1] | 6. <----+ (discarded due to Rule 4) 7. <------- (x+3) PANA_auth[EAP Req ID=2] . . Figure 10: Example for Single sequence number with EAP retransmission method This method is vulnerable to a blind DoS attack on the sequence number since the PaC will accept quite a wide range of sequence numbers. For example, if an attacker blindly sends a bogus message to a legitimate PaC with a randomly chosen sequence number, it will be accepted by the PaC with 50% probability, and once this happens, all messages sent from the communicating PAA will be discarded as long as they have a sequence number smaller than the accepted value. The problem of this method leads to a requirement for PaC to have a narrow range of acceptable sequence numbers to make the blind DoS attack difficult. Note that the DoS attack cannot be prevented if the attacker is on the same IP link as PaC and able to eavesdrop the PANA conversation. However, the attacker needs to put itself in promiscuous mode and thus spend more resources to eavesdrop and launch the attack (in other words, non-blind DoS attack is still possible as long as sequence numbers are unprotected.) A.2.2. Single sequence number with PANA-layer retransmission method The next method is still based on using a single sequence number but Tschofenig (ed.) Expires September 2003 28 PANA March 2003 the PANA-layer takes the responsibility of retransmission. The method uses the following rules in addition to the common rules described in section A.2. Rule 3: Use PANA-layer retransmission for retransmitting both EAP and non-EAP messages (based on a timer expiration). EAP layer retransmission is turned off. Retransmission based on timer occurs both on PaC and PAA side, but not on both sides simultaneously. PAA does retransmission at least for PANA_revocation and PANA_reauth messages, otherwise PaC takes care of retransmission. Rule 4: When the PaC receives a message from the PAA, it accepts the message if the sequence number is equal to that of the last accepted message + 1. If the sequence number is equal to that of the last accepted message, the PaC retransmits the last transmitted message. Otherwise, it silently discards the message. Rule 5: When the PAA receives a message from the PaC, it accepts the message if the sequence number is equal to that of the last transmitted message. If the receiving sequence number is equal to that of the last transmitted message - 1, the PAA retransmits the last transmitted message and discard the received message. Otherwise, it silently discards the message. Rule 6: The PaC retransmits the last transmitted EAP Response until a new EAP Request message or an EAP Success/Failure message is received and accepted. Rule 7: PAA must keep the copy of the last transmitted message and must be able to retransmit it until either a valid message is received and accepted by the PAA or a timer expires. The timer is used if no new message will be sent from the PaC. Tschofenig (ed.) Expires September 2003 29 PANA March 2003 PaC PAA Seq# Message -------------------------------------------- 1. <-------- (x) PANA_auth[EAP Req ID=1] 2. ---+ (x) PANA_auth[EAP Resp ID=1] | (retransmission timeout at PaC) 3. ---|----> (x) PANA_auth[EAP Resp ID=1] 4. | +--- (x+1) PANA_auth[EAP Req ID=2] | | +-|--> (duplicate detected) 5. <----|--- (x+1) PANA_auth[EAP Req ID=2] | 6. -----|--> (x+1) PANA_auth[EAP Resp ID=2] | <----|--- (x+2) PANA_auth[EAP Req ID=3] 7. -----|--> (x+2) PANA_auth[EAP Resp ID=3] <----+ (discarded by PaC) (retransmission timeout at PaC) 8. --------> (x+2) PANA_auth[EAP Resp ID=3] 9. lost<---- (x+3) PANA_auth[EAP Succ ID=3] (retransmission timeout at PaC) 10.---->lost (x+2) PANA_auth[EAP Resp ID=3] (retransmission timeout at PaC) 11.--------> (x+2) PANA_auth[EAP Resp ID=3] 12.<-------- (x+3) PANA_succ[EAP Succ ID=3] (retransmission timer stopped at PaC) (deletion timeout at PAA) (message (x+3) deleted at PAA) 13.lost<---- (x+4) PANA_revocation (retransmission timeout at PAA) 14.<-------- (x+4) PANA_revocation 15.---->lost (x+4) PANA_revocation_ack (retransmission timeout at PAA) 16.<-------- (x+4) PANA_revocation 17.--------> (x+4) PANA_revocation_ack (retransmission timer stopped at PAA) Figure 11: Example for Single sequence number with PANA-layer retransmission method This method has an advantage of eliminating EAP layer retransmission by providing reliability at the PANA layer. Retransmission at the EAP layer has a problem with determining an appropriate retransmission timer value, which occurs when the lower-layer is unreliable. In this case an EAP authenticator cannot distinguish between (i) EAP Request or EAP Response message loss (in this case the retransmission timer should be calculated based on network characteristics) and (ii) long latency for EAP Response generation due to e.g., user input etc. (in this case the retransmission timer should be calculated based on user or application characteristics). In general, the retransmission timer for case (ii) is longer than that for case (i). If case (i) happens while the retransmission timer is calculated based on user or application characteristics, then it might frustrate an end user since the completion of the Tschofenig (ed.) Expires September 2003 30 PANA March 2003 authentication procedure takes unnecessarily long. If case (ii) happens while the retransmission timer is calculated based on network characteristics (i.e., RTT), then unnecessarily traffic is generated by retransmission. Note that in this method a PaC still cannot distinguish case (i) and case (iii) the EAP authenticator or a backend authentication server is taking time to generate an EAP Request. A problem of this method is that it is based on the assumption that EAP authenticator does not send a new EAP message until an EAP Response to the outstanding EAP Request is received. However, this assumption does not hold at least EAP Success/Failure message which does not need the outstanding EAP Request to be responded before sending the EAP Success/Failure message. This would require timer-based retransmission not only at PaC side but also at PAA side. Another problem occurs when a new EAP message overrides the outstanding EAP Request, the PaC cannot assume any more that the sequence number of the next message to be accepted is the last accepted message + 1. So the PaC needs to accept a range of sequence numbers, instead of a single sequence number. These two additional things would increase the complexity of this method. A.3. Dual sequence number approach Based on the analysis of previous schemes, it is recognized that two sequence numbers are needed anyway, one for each direction. Two different methods are proposed based on this approach. Both methods have the following rules in common. Rule 1: A PANA packet carries two sequence numbers: transmitted sequence number (tseq) and received sequence number (rseq). tseq starts from initial sequence number (ISN) and is monotonically increased by 1. The arithmetic defined in [RFC1982] is used for sequence number operation. It is assumed that the two sequence numbers have the same length for simplicity. Rule 2: When PAA or PAC sends a new message, a new sequence number is placed on the tseq field of message. Every transmitted message is given a new sequence number. Rule 3: When a message is sent from PaC or PAA, rseq is copied from the tseq field of the last accepted message. Rule 4: For messages which experience a PANA layer retransmission, the retransmission timer is stopped when the message is acknowledged. It is possible to carry multiple EAP sequences in a single PANA sequence, with using EAP Success/Failure message as a delimiter of each EAP sequence. In this case, EAP Success/Failure message needs to be reliably delivered. Tschofenig (ed.) Expires September 2003 31 PANA March 2003 A.3.1. Dual sequence number with orderly-delivery method This method relies on EAP layer retransmission for EAP messages. This method is referred to as orderly-delivery method. The following rules are used in addition to the common rules. Rule 5: Use the EAP-layer retransmission for retransmitting EAP Requests (based on a timer expiration). For other PANA layer messages that require a response from the peer, PANA layer has its own mechanism to retransmit the request until it gets a response or gives up. A new tseq value is always used when sending any message even when it is retransmitted at PANA layer. Rule 6: When a message is received, it is accepted if (i) the tseq value is greater than the tseq of the last accepted message and (ii) the rseq falls in the range between the tseq of the last acknowledged message + 1 and the tseq of the last transmitted message. Otherwise, the received message is discarded. PaC PAA (tseq,rseq) Message -------------------------------------------------- 1. <------- (x,y) PANA_auth[EAP Req, ID=1] 2. -------> (y+1,x) PANA_auth[EAP Resp, ID=1] 3. <------- (x+1,y+1) PANA_auth[EAP Req, ID=2] 4. --->lost (y+2,x+1) PANA_auth[EAP Resp, ID=2] (retransmission timeout at EAP layer) 5. <------- (x+2,y+1) PANA_auth[EAP Req, ID=2] 6. -------> (y+3,x+2) PANA_auth[EAP Resp, ID=2] 7. lost<--- (x+3,y+3) PANA_auth[EAP Req, ID=3] (retransmission timeout at EAP layer) 8. +---- (x+4,y+3) PANA_auth[EAP Req, ID=3] | (retransmission timeout at EAP layer) 9. <--|---- (x+5,y+3) PANA_auth[EAP Req, ID=3] 10.---|---> (y+4,x+5) PANA_auth[EAP Resp, ID=3] | <--+ (out of order. discarded) 11.lost<--- (x+6,y+4) PANA_succ[EAP Succ, ID=3] (retransmission timeout at PAA) 12.<------- (x+7,y+4) PANA_succ[EAP Succ, ID=3] 13.--->lost (y+5,x+7) PANA_succ_ack (retransmission timeout at PAA) 14.<------- (x+8,y+4) PANA_succ[EAP Succ, ID=3] (dupicate detected by PaC) 15.-------> (y+6,x+8) PANA_succ_ack Figure 12: Example for Dual sequence number with orderly-delivery method A.3.2. Dual sequence number with reliable-delivery method Tschofenig (ed.) Expires September 2003 32 PANA March 2003 This method relies solely on PANA layer retransmission for all messages. This method is referred to as reliable-delivery method. The following additional rules are applied in addition to the common rules. Rule 5: Use the PANA layer retransmission for retransmitting all messages (based on a timer expiration). EAP retransmission is turned off. Rule 6: Either an ACK message is used for acknowledgment or an acknowledgment can be piggybacked with data. ACK messages are not retransmitted. An ACK message is sent if no the acknowledgement cannot be piggybacked with a data within a given time frame W. Rule 7: When a message is received, it is accepted if (i) the tseq value is greater than the tseq of the last accepted message and (ii) the rseq falls in the range between the tseq of the last acknowledged message and the tseq of the last transmitted message. Otherwise, the received message is discarded. Rule 8: When a duplicate message is received, the last transmitted message is retransmitted if the received message is not an ACK. A message is considered as duplicate if its tseq value is equal to the tseq of the last accepted message. Tschofenig (ed.) Expires September 2003 33 PANA March 2003 PaC PAA (tseq,rseq) Message -------------------------------------------------- 1. <------- (x,y) PANA_auth[EAP Req, ID=1] (user input ongoing) 2. -------> (y+1,x) PANA_ACK (user input completed) 3. -------> (y+2,x) PANA_auth[EAP Resp, ID=1] 4. <------- (x+1,y+2) PANA_auth[EAP Req, ID=2] 5. --->lost (y+3,x+1) PANA_auth[EAP Resp, ID=2] (retransmission timeout at PAA) 6. <------- (x+1,y+2) PANA_auth[EAP Req, ID=2] (duplicate detected by PaC) 7. -------> (y+3,x+1) PANA_auth[EAP Resp, ID=2] 8. lost<--- (x+2,y+3) PANA_auth[EAP Req, ID=3] (retransmission timeout at PaC) 9. -------> (y+3,x+1) PANA_auth[EAP Resp, ID=2] (duplicate detected at PAA) 10.<------- (x+2,y+3) PANA_auth[EAP Req, ID=3] 11.---+ (y+4,x+2) PANA_auth[EAP Resp, ID=3] | (retransmission timeout at PAA) 12.<--|---- (x+2,y+3) PANA_auth[EAP Req, ID=3] | (duplicate detected at PaC) 13.---|---> (y+4,x+2) PANA_auth[EAP Resp, ID=3] 14.<--|---- (x+3,y+4) PANA_succ[EAP Succ, ID=3] 15.---|---> (y+5,x+3) PANA_ACK +---> (out of order. discarded) Figure 13: Example for Dual sequence number with reliable-delivery method A.3.3 Comparison of the dual sequence number methods The orderly-delivery method is simpler than the reliable-delivery method in that the former does not allow sending a separate ACK while the latter does. In terms of authentication performance, the reliable-delivery method is better than the orderly-delivery method in that the former gives more detailed status of the link than the latter, e.g., an entity can know whether a request has reached the communicating peer without before receiving a response. The reliable-delivery can reduce retransmission traffic and communication delay that would occur if there is no reliability, as described in section A.2.2. A.4 Consensus Although it is recognizable that the reliable-delivery method would be important in terms of improvement of overall authentication latency, we believe that this is a performance problem of EAP and not a problem of PANA. It is agreed that solving the EAP problem is not the scope of PANA and simplicity is more important factor in the PANA design. Tschofenig (ed.) Expires September 2003 34 PANA March 2003 As a consequence, the orderly-delivery method is chosen as the message transport part of PANA. 15 Full Copyright Statement Copyright (C) The Internet Society (2000). All Rights Reserved. This document and translations of it may be copied and furnished to others, and derivative works that comment on or otherwise explain it or assist in its implementation may be prepared, copied, published and distributed, in whole or in part, without restriction of any kind, provided that the above copyright notice and this paragraph are included on all such copies and derivative works. However, this document itself may not be modified in any way, such as by removing the copyright notice or references to the Internet Society or other Internet organizations, except as needed for the purpose of developing Internet standards in which case the procedures for copyrights defined in the Internet Standards process must be followed, or as required to translate it into languages other than English. The limited permissions granted above are perpetual and will not be revoked by the Internet Society or its successors or assigns. This document and the information contained herein is provided on an "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Acknowledgement Funding for the RFC Editor function is currently provided by the Internet Society. Tschofenig (ed.) Expires September 2003 35