Network Working Group F. Bersani Request for Comments: 4764 France Telecom R&D Category: Experimental H. Tschofenig Siemens Networks GmbH & Co KG January 2007 The EAP-PSK Protocol: A Pre-Shared Key Extensible Authentication Protocol (EAP) Method Status of This Memo This memo defines an Experimental Protocol for the Internet community. It does not specify an Internet standard of any kind. Discussion and suggestions for improvement are requested. Distribution of this memo is unlimited. Copyright Notice Copyright (C) The IETF Trust (2007). IESG Note This RFC is not a candidate for any level of Internet Standard. The IETF disclaims any knowledge of the fitness of this RFC for any purpose and in particular notes that the decision to publish is not based on IETF review for such things as security, congestion control, or inappropriate interaction with deployed protocols. The RFC Editor has chosen to publish this document at its discretion. Readers of this document should exercise caution in evaluating its value for implementation and deployment. See RFC 3932 for more information. The IESG thinks that this work is related to IETF work done in WGs EMU and EAP, but this does not prevent publishing. Abstract This document specifies EAP-PSK, an Extensible Authentication Protocol (EAP) method for mutual authentication and session key derivation using a Pre-Shared Key (PSK). EAP-PSK provides a protected communication channel when mutual authentication is successful for both parties to communicate over. This document describes the use of this channel only for protected exchange of result indications, but future EAP-PSK extensions may use the channel for other purposes. EAP-PSK is designed for authentication over insecure networks such as IEEE 802.11. Bersani & Tschofenig Experimental [Page 1] RFC 4764 EAP-PSK January 2007 Table of Contents 1. Introduction ....................................................4 1.1. Design Goals for EAP-PSK ...................................4 1.1.1. Simplicity ..........................................4 1.1.2. Wide Applicability ..................................5 1.1.3. Security ............................................5 1.1.4. Extensibility .......................................5 1.2. Terminology ................................................5 1.3. Conventions ................................................8 1.4. Related Work ...............................................9 2. Protocol Overview ..............................................12 2.1. EAP-PSK Key Hierarchy .....................................13 2.1.1. The PSK ............................................13 2.1.2. AK .................................................14 2.1.3. KDK ................................................14 2.2. The TEK ...................................................15 2.3. The MSK ...................................................15 2.4. The EMSK ..................................................15 2.5. The IV ....................................................15 3. Cryptographic Design of EAP-PSK ................................15 3.1. The Key Setup .............................................16 3.2. The Authenticated Key Exchange ............................19 3.3. The Protected Channel .....................................23 4. EAP-PSK Message Flows ..........................................25 4.1. EAP-PSK Standard Authentication ...........................26 4.2. EAP-PSK Extended Authentication ...........................28 5. EAP-PSK Message Format .........................................31 5.1. EAP-PSK First Message .....................................32 5.2. EAP-PSK Second Message ....................................34 5.3. EAP-PSK Third Message .....................................36 5.4. EAP-PSK Fourth Message ....................................39 6. Rules of Operation for the EAP-PSK Protected Channel ...........41 6.1. Protected Result Indications ..............................41 6.1.1. CONT ...............................................42 6.1.2. DONE_SUCCESS .......................................43 6.1.3. DONE_FAILURE .......................................43 6.2. Extended Authentication ...................................43 7. IANA Considerations ............................................45 7.1. Allocation of an EAP-Request/Response Type for EAP-PSK ....45 7.2. Allocation of EXT Type Numbers ............................45 8. Security Considerations ........................................46 8.1. Mutual Authentication .....................................46 8.2. Protected Result Indications ..............................47 8.3. Integrity Protection ......................................48 8.4. Replay Protection .........................................48 8.5. Reflection Attacks ........................................48 8.6. Dictionary Attacks ........................................49 Bersani & Tschofenig Experimental [Page 2] RFC 4764 EAP-PSK January 2007 8.7. Key Derivation ............................................49 8.8. Denial-of-Service Resistance ..............................51 8.9. Session Independence ......................................51 8.10. Exposition of the PSK ....................................52 8.11. Fragmentation ............................................52 8.12. Channel Binding ..........................................53 8.13. Fast Reconnect ...........................................53 8.14. Identity Protection ......................................53 8.15. Protected Ciphersuite Negotiation ........................55 8.16. Confidentiality ..........................................55 8.17. Cryptographic Binding ....................................55 8.18. Implementation of EAP-PSK ................................55 9. Security Claims ................................................56 10. Acknowledgments ...............................................57 11. References ....................................................57 11.1. Normative References .....................................57 11.2. Informative References ...................................58 Appendix A. Generation of the PSK from a Password - Discouraged ...62 Bersani & Tschofenig Experimental [Page 3] RFC 4764 EAP-PSK January 2007 1. Introduction 1.1. Design Goals for EAP-PSK The Extensible Authentication Protocol (EAP) [3] provides an authentication framework that supports multiple authentication methods. This document specifies an EAP method, called EAP-PSK, that uses a Pre-Shared Key (PSK). EAP-PSK was developed at France Telecom R&D in 2003-2004. It is published as an RFC for the general information of the Internet community and to allow independent implementations. Because PSKs are of frequent use in security protocols, other protocols may also refer to a PSK or contain this word in their name. For instance, Wi-Fi Protected Access (WPA) [48] specifies an authentication mode called "WPA-PSK". EAP-PSK is distinct from these protocols and should not be confused with them. Design goals for EAP-PSK were: o Simplicity: EAP-PSK should be easy to implement and deploy without any pre-existing infrastructure. It should be available quickly because recently-released protocols, such as IEEE 802.11i [27], employ EAP in a different threat model than PPP [44] and thus require "modern" EAP methods. o Wide applicability: EAP-PSK should be suitable to authenticate over any network, and in particular over IEEE 802.11 [28] wireless LANs. o Security: EAP-PSK should be conservative in its cryptographic design. o Extensibility: EAP-PSK should be easily extensible. 1.1.1. Simplicity For the sake of simplicity, EAP-PSK relies on a single cryptographic primitive, AES-128 [7]. Restriction to such a primitive, and in particular, not using asymmetric cryptography like Diffie-Hellman key exchange, makes EAP- PSK: Bersani & Tschofenig Experimental [Page 4] RFC 4764 EAP-PSK January 2007 o Easy to understand and implement while avoiding cryptographic negotiations. o Lightweight and well suited for any type of device, especially those with little processing power and memory. However, as further discussed in Section 8, this prevents EAP-PSK from offering advanced features such as identity protection, password support, or Perfect Forward Secrecy (PFS). This choice has been deliberately made as a trade-off between simplicity and security. For the sake of simplicity, EAP-PSK has also chosen a fixed message format and not a Type-Length-Value (TLV) design. 1.1.2. Wide Applicability EAP-PSK has been designed in a threat model where the attacker has full control over the communication channel. This is the EAP threat model that is presented in Section 7.1 of [3]. 1.1.3. Security Since the design of authenticated key exchange is notoriously known to be hard and error prone, EAP-PSK tries to avoid inventing any new cryptographic mechanism. It attempts instead to build on existing primitives and protocols that have been reviewed by the cryptographic community. 1.1.4. Extensibility EAP-PSK explicitly provides a mechanism to allow future extensions within its protected channel (see Section 3.3). Thanks to this mechanism, EAP-PSK will be able to provide more sophisticated services as the need to do so arises. 1.2. Terminology Authentication, Authorization, and Accounting (AAA) Please refer to [10] for more details. AES-128 A block cipher specified in the Advanced Encryption Standard [7]. Authentication Key (AK) A 16-byte key derived from the PSK that the EAP peer and server use to mutually authenticate. Bersani & Tschofenig Experimental [Page 5] RFC 4764 EAP-PSK January 2007 AKEP2 An authenticated key exchange protocol; please refer to [14] for more details. Backend Authentication Server An entity that provides an authentication service to an Authenticator. When used, this server typically executes EAP methods for the Authenticator. (This terminology is also used in [26], and has the same meaning in this document.) CMAC Cipher-based Message Authentication Code. It is the authentication mode of operation of AES recommended by NIST in [8]. Extensible Authentication Protocol (EAP) Defined in [3]. EAP Authenticator (or simply Authenticator) The end of the EAP link initiating the EAP authentication methods. (This terminology is also used in [26], and has the same meaning in this document.) EAP peer (or simply peer) The end of the EAP link that responds to the Authenticator. (In [26], this end is known as the Supplicant.) EAP server (or simply server) The entity that terminates the EAP authentication with the peer. When there is no Backend Authentication Server, this term refers to the EAP Authenticator. Where the EAP Authenticator operates in pass-through mode, it refers to the Backend Authentication Server. EAX An authenticated-encryption with associated data mode of operation for block ciphers [4]. Extended Master Session Key (EMSK) Additional keying material derived between the EAP peer and server that is exported by the EAP method. The EMSK is reserved for future uses that are not defined yet and is not provided to a third party. Please refer to [9] for more details. EAP-PSK generates a 64-byte EMSK. Initialization Vector (IV) A quantity of at least 64 bytes, suitable for use in an initialization vector field, that is derived between the peer and EAP server. Since the IV is a known value in Bersani & Tschofenig Experimental [Page 6] RFC 4764 EAP-PSK January 2007 methods such as EAP-TLS [11], it cannot be used by itself for computation of any quantity that needs to remain secret. As a result, its use has been deprecated and EAP methods are not required to generate it. Please refer to [9] for more details. EAP-PSK does not generate an IV. Key-Derivation Key (KDK) A 16-byte key derived from the PSK that the EAP peer and server use to derive session keys (namely, the TEK, MSK, and EMSK). Message Authentication Code (MAC) Informally, the purpose of a MAC is to provide assurances regarding both the source of a message and its integrity [40]. IEEE 802.11i uses the acronym MIC (Message Integrity Check) to avoid confusion with the other meaning of the acronym MAC (Medium Access Control). Master Session Key (MSK) Keying material that is derived between the EAP peer and server and exported by the EAP method. In existing implementations, a AAA server acting as an EAP server transports the MSK to the Authenticator [9]. EAP-PSK generates a 64-byte MSK. Network Access Identifier (NAI) Identifier used to identify the communicating parties [2]. One Key CBC-MAC 1 (OMAC1) A method to generate a Message Authentication Code [29]. CMAC is the name under which NIST has standardized OMAC1. Perfect Forward Secrecy (PFS) The confidence that the compromise of a long-term private key does not compromise any earlier session keys. In other words, once an EAP dialog is finished and its corresponding keys are forgotten, even someone who has recorded all of the data from the connection and gets access to all of the long-term keys of the peer and the server cannot reconstruct the keys used to protect the conversation without doing a brute-force search of the session key space. EAP-PSK does not have this property. Bersani & Tschofenig Experimental [Page 7] RFC 4764 EAP-PSK January 2007 Pre-Shared Key (PSK) A Pre-Shared Key simply means a key in symmetric cryptography. This key is derived by some prior mechanism and shared between the parties before the protocol using it takes place. It is merely a bit sequence of given length, each bit of which has been chosen at random uniformly and independently. For EAP-PSK, the PSK is the long-term 16- byte credential shared by the EAP peer and server. Protected Result Indication Please refer to Section 7.16 of [3] for a definition of this term. This feature has been introduced because EAP- Success/Failure packets are unidirectional and are not protected. Transient EAP Key (TEK) A session key that is used to establish a protected channel between the EAP peer and server during the EAP authentication exchange. The TEK is appropriate for use with the ciphersuite negotiated between the EAP peer and server to protect the EAP conversation. Note that the ciphersuite used to set up the protected channel between the EAP peer and server during EAP authentication is unrelated to the ciphersuite used to subsequently protect data sent between the EAP peer and Authenticator [9]. EAP-PSK uses a 16-byte TEK for its protected channel, which is the only ciphersuite available between the EAP peer and server to protect the EAP conversation. This ciphersuite uses AES-128 in the EAX mode of operation. 1.3. Conventions All numbers presented in this document are considered in network-byte order. || denotes concatenation of strings (and not the logical OR). MAC(K, String) denotes the MAC of String under the key K (the algorithm used in this document to compute the MACs is CMAC with AES- 128; see Section 3.2). [String] denotes the concatenation of String with the MAC of String calculated as specified by the context. Hence, we have, with K specified by the context: [String]=String||MAC(K,String) ** denotes integer exponentiation. Bersani & Tschofenig Experimental [Page 8] RFC 4764 EAP-PSK January 2007 "i" denotes the unsigned binary representation on 16 bytes of the integer i in network byte order. Therefore, this notation only makes sense when i is between 0 and 2**128-1. denotes the unsigned binary representation on 4 bytes of the integer i in network byte order. Therefore, this notation only makes sense when i is between 0 and 2**32-1. 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 [1]. 1.4. Related Work At the time this document is written, only three EAP methods are standards track EAP methods per IETF terminology (see [17]), namely: o MD5-Challenge (EAP-Request/Response type 4), defined in [3], which uses a MD5 challenge similar to [45]. o OTP (EAP-Request/Response type 5), defined in [3], which aims at providing One-Time Password support similar to [22] and [39]. o GTC (EAP-Request/Response type 6), defined in [3], which aims at providing Generic Token Card Support. Unfortunately, all three methods are deprecated for security reasons that are explained in part in [3]. Myriads of EAP methods have, however, been otherwise proposed: o One as an experimental RFC (EAP-TLS [11]), which therefore is not a standard (see [25]). o Some as individual Internet-Draft submissions (e.g., [42] or this document). o And some even undocumented (e.g., Rob EAP, which has EAP-Request/ Response type 31). However, no secure and mature Pre-Shared Key EAP method is yet easily and widely available, which is all the more regrettable because Pre- Shared Key methods are the most basic ones! The existing proposals for a future Pre-Shared Key EAP method are briefly reviewed hereafter (please refer to [16] for a more thorough synthesis of EAP methods). Bersani & Tschofenig Experimental [Page 9] RFC 4764 EAP-PSK January 2007 Among these proposals, there are some that: o Are broken from a security point of view, e.g.: * LEAP, which is specified in [38] and whose vulnerabilities are discussed in [49]. * EAP-MSCHAPv2, which is specified in [34] and whose vulnerabilities are indirectly discussed in [43]. o Essentially require additional infrastructure, e.g., EAP-SIM [24], EAP-AKA [12], or OTP/token card methods like [31]. o Are not shared key methods but are often confused with them, namely, the password methods, e.g., EAP-SRP [18] or SPEKE [30], whose wide adoption very unfortunately seems to be hindered by Intellectual Property Rights issues. o Are generic tunneling methods, which do not essentially rely on Pre-Shared Keys as they require a public-key certificate for the server and allow the peer to authenticate with whatever EAP method or even other non-EAP authentication mechanisms, namely, [32] and [21]. o Are abandoned but have provided the basis for EAP-PSK, namely, EAP-Archie [47]. o Are possible alternatives to EAP-PSK (i.e., claimed to be secure and subject of active work): * EAP-FAST [42]. * EAP-IKEv2 [46]. * EAP-TLS (when shared key/password support is added to TLS; see [50]). EAP-PSK differs from the aforementioned methods on the following points: o No attacks on EAP-PSK within its threat model have yet been found. o EAP-PSK was not designed to leverage a pre-existing infrastructure. Thus, it does not inherit potential limitations of such an infrastructure and it should be easier to deploy "from scratch". o EAP-PSK wished to avoid IPR blockages. Bersani & Tschofenig Experimental [Page 10] RFC 4764 EAP-PSK January 2007 o EAP-PSK does not have any dependencies on protocols other than EAP. o EAP-PSK was restricted to simply proposing a Pre-Shared Key method with symmetric cryptography * To remain simple to understand and implement * To avoid potentially complex configurations and negotiations o EAP-PSK was designed with efficiency in mind. Bersani & Tschofenig Experimental [Page 11] RFC 4764 EAP-PSK January 2007 2. Protocol Overview Figure 1 presents an overview of the EAP-PSK key hierarchy. +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-++ ---+ | | ^ | EAP-PSK Protocol: a Pre-Shared Key EAP Method | | | | | | +----------+ | | | | PSK | | | | |(16 bytes)| | | | +----------+ | | | | | | | v | | | *********************** | | | *Modified Counter Mode* | | | *********************** | | | | | | | | v v | | | +----------+ +----------+ +----------------+ | | | | AK | | KDK | | RAND_P | | | | |(16 bytes)| |(16 bytes)| | (16 bytes) | | | | +----------+ +----------+ +----------------+ | | | | | | | | | | | | | +-----------+ | | | | | +--------+|Plain Text | | | | | |+-------+|Header H||Var.Length | | | | | ||Nonce N||22 bytes|+-----------+ v v Local | ||4 bytes|+--------+ | *********************** to EAP | |+-------+ | +--------+ +------*Modified Counter Mode* Method | | | v v v *********************** | | | | ******* +--------+ |64 |64 | | | | * EAX *-------|TEK | |bytes |bytes | | | +-->******* |16 bytes| | | | | | | +--------+ | | | | | +-----+----+ | | | | | v v | | | | |+--------+ +-------------------+ | | | | ||Tag | |Cipher Text Payload| | | | | ||16 bytes| | Variable length L | | | | | |+--------+ +-------------------+ | | | V +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-++ ---+ | | ^ +-+-+-+-+-++ +-+-+-+-+-++ | |MSK | |EMSK | | | | | | Exported | +-+-+-+-+-++ +-+-+-+-+-++ by EAP | Bersani & Tschofenig Experimental [Page 12] RFC 4764 EAP-PSK January 2007 | | Method | | | | V V | ************************* V * AAA Key Derivation * ---+ * Naming & Binding * ************************* Figure 1: EAP-PSK Key Hierarchy Overview 2.1. EAP-PSK Key Hierarchy This section presents the key hierarchy used by EAP-PSK. This hierarchy is inspired by the EAP key hierarchy described in [9]. 2.1.1. The PSK The PSK is shared between the EAP peer and the EAP server. EAP-PSK assumes that the PSK is known only to the EAP peer and EAP server. The security properties of the protocol are compromised if it has wider distribution. Please note that EAP-PSK shares this property with all other symmetric key methods (including all password-based methods). EAP-PSK also assumes the EAP server and EAP peer identify the correct PSK to use with each other thanks to their respective NAIs. This means that there MUST only be at most one PSK shared between an EAP server using a given server NAI and an EAP peer using a given peer NAI. This PSK is used, as shown in Figure 2, to derive two 16-byte static long-lived subkeys, respectively called the Authentication Key (AK) and the Key-Derivation Key (KDK). This derivation should only be done once: it is called the key setup. See Section 3.1 for an explanation of why PSK is not used as a static long-lived key, but only as the initial keying material for deriving the static long- lived keys, AK and KDK, which are actually used by the protocol EAP- PSK. Bersani & Tschofenig Experimental [Page 13] RFC 4764 EAP-PSK January 2007 +---------------------------+ | PSK | | (16 bytes) | +---------------------------+ | | v v +---------------------------+ +---------------------------+ | AK | | KDK | | (16 bytes) | | (16 bytes) | +---------------------------+ +---------------------------+ Figure 2: Derivation of AK and KDK from the PSK 2.1.2. AK EAP-PSK uses AK to mutually authenticate the EAP peer and the EAP server. AK is a static long-lived key derived from the PSK; see Section 3.1. AK is not a session key. The EAP server and EAP peer identify the correct AK to use with each other thanks to their respective NAIs. This means that there MUST only be at most one AK shared between an EAP server using a given server NAI and an EAP peer using a given peer NAI. This is the case when there is at most one PSK shared between an EAP server using a given server NAI and an EAP peer using a given peer NAI; see Section 2.1.1. The EAP peer chooses the AK to use based on the EAP server NAI that has been sent by the EAP server in the first EAP-PSK message (namely, ID_S; see Section 4.1) and the EAP peer NAI it chooses to include in the second EAP-PSK message (namely, ID_P; see Section 4.1). 2.1.3. KDK EAP-PSK uses KDK to derive session keys shared by the EAP peer and the EAP server (namely, the TEK, MSK, and EMSK). KDK is a static long-lived key derived from the PSK; see Section 3.1. KDK is not a session key. The EAP server and EAP peer identify the correct AK to use with each other thanks to their respective NAIs. This means that there MUST only be at most one AK shared between an EAP server using a given server NAI and an EAP peer using a given peer NAI. This is the case Bersani & Tschofenig Experimental [Page 14] RFC 4764 EAP-PSK January 2007 when there is at most one PSK shared between an EAP server using a given server NAI and an EAP peer using a given peer NAI; see Section 2.1.1. The EAP peer chooses the AK to use based on the EAP server NAI that has been sent by the EAP server in the first EAP-PSK message (namely, ID_S; see Section 4.1) and the EAP peer NAI it chooses to include in the second EAP-PSK message (namely, ID_P; see Section 4.1). 2.2. The TEK EAP-PSK derives a 16-byte TEK thanks to a random number exchanged during authentication (RAND_P; see Section 5.1) and KDK. This TEK is used to implement a protected channel for both mutually authenticated parties to communicate over securely. 2.3. The MSK EAP-PSK derives a MSK thanks to a random number exchanged during authentication (RAND_P; see Section 5.1) and the KDK. The MSK is 64 bytes long, which complies with [3]. 2.4. The EMSK EAP-PSK derives an EMSK thanks to a random number exchanged during authentication (RAND_P; see Section 5.1) and the KDK. The EMSK is 64 bytes long, which complies with [3]. 2.5. The IV EAP-PSK does not derive any IV, which complies with [9]. 3. Cryptographic Design of EAP-PSK EAP-PSK relies on a single cryptographic primitive, a block cipher, which is instantiated with AES-128. AES-128 takes a 16-byte Pre- Shared Key and a 16-byte Plain Text block as inputs. It outputs a 16-byte Cipher Text block. For a detailed description of AES-128, please refer to [7]. AES-128 has been chosen because: o It is standardized and implementations are widely available. Bersani & Tschofenig Experimental [Page 15] RFC 4764 EAP-PSK January 2007 o It has been carefully reviewed by the cryptographic community and is believed to be secure. Other block ciphers could easily be proposed for EAP-PSK, as EAP-PSK does not intrinsically depend on AES-128. The only parameters of AES-128 that EAP-PSK depends on are the AES-128 block and key size (16 bytes). For the sake of simplicity, EAP-PSK has, however, been chosen to restrict to a single mandatory block cipher and not allow the negotiation of other block ciphers. In the case that AES-128 is deprecated for security reasons, EAP-PSK should also be deprecated and a cut-and-paste EAP-PSK' should be defined with another block cipher. This EAP-PSK' should not be backward compatible with EAP-PSK because of the security issues with AES-128. EAP-PSK' should therefore use a different EAP-Request/Response Type number. With the EAP-Request/Response Type number space structure defined in [3], this should not be a problem. The use of a different EAP-Request/Response Type number for EAP-PSK' will prevent this new method from being vulnerable to chosen protocol attacks. EAP-PSK uses three cryptographic parts: o A key setup to derive AK and KDK from the PSK. o An authenticated key exchange protocol to mutually authenticate the communicating parties and derive session keys. o A protected channel protocol for both mutually authenticated parties to communicate over. Each part is discussed in more detail in the subsequent paragraphs. 3.1. The Key Setup EAP-PSK needs two cryptographically separated 16-byte subkeys for mutual authentication and session key derivation. Indeed, it is a rule of thumb in cryptography to use different keys for different applications. It could have implemented these two subkeys either by specifying a 32-byte PSK that would then be split in two 16-byte subkeys, or by specifying a 16-byte PSK that would then be cryptographically expanded to two 16-byte subkeys. Because provisioning a 32-byte long-term credential is more cumbersome than a 16-byte one, and the strength of the derived session keys is 16 bytes either way, the latter option was chosen. Bersani & Tschofenig Experimental [Page 16] RFC 4764 EAP-PSK January 2007 Hence, the PSK is only used by EAP-PSK to derive AK and KDK. This derivation should be done only once, immediately after the PSK has been provisioned. As soon as AK and KDK have been derived, the PSK should be deleted. If the PSK is deleted, it should be done so securely (see, for instance, [19] for guidance on secure deletion of the PSK). Derivation of AK and KDK from the PSK is called the key setup: o The input to the key setup is the PSK. o The outputs of the key setup are AK and KDK. AK and KDK are derived from the PSK using the modified counter mode of operation of AES-128. The modified counter mode is a length increasing function, i.e., it expands one AES-128 input block into a longer t-block output, where t>=2. This mode was chosen for the key setup because it had already been chosen for the derivation of the session keys (see Section 3.2). The details of the derivation of AK and KDK from the PSK are shown in Figure 3. Bersani & Tschofenig Experimental [Page 17] RFC 4764 EAP-PSK January 2007 +--------------------------+ | "0" | | Input Block (16 bytes) | +--------------------------+ | v +----------------+ | | | AES-128(PSK,.) | | | +----------------+ | | +----------------------------+ | | v v +--------+ +---+ +--------+ +---+ | c1="1" |->|XOR| | c2="2" |->|XOR| |16 bytes| +---+ |16 bytes| +---+ +--------+ | +--------+ | | | +----------------+ +----------------+ | | | | | AES-128(PSK,.) | | AES-128(PSK,.) | | | | | +----------------+ +----------------+ | | | | v v +------------------------+ +------------------------+ | AK | | KDK | | (16 bytes) | | (16 bytes) | +------------------------+ +------------------------+ Figure 3: Derivation of AK and KDK from the PSK in Details The input block is "0". For the sake of simplicity, this input block has been chosen constant: it could have been set to a value depending on the peer and the server (for instance, the XOR of their respective NAIs appropriately truncated or zero-padded), but this did not seem to add much security to the scheme, whereas it added complexity. Any 16-byte constant could have been chosen, as the security is not supposed to depend on the particular value taken by the constant. "0" was arbitrarily chosen. Bersani & Tschofenig Experimental [Page 18] RFC 4764 EAP-PSK January 2007 3.2. The Authenticated Key Exchange The authentication protocol used by EAP-PSK is inspired by AKEP2, which is described in [14]. AKEP2 consists of a one-and-a-half round-trip exchange, as shown in Figure 4, which is inspired by Figure 5 of [14]. Bob Alice | RA | |<---------------------------------------------------------| | | | [B||A||RA||RB] | |--------------------------------------------------------->| | | | [A||RB] | |<---------------------------------------------------------| Figure 4: Overview of AKEP2 It is also worth noting that [14] focuses on cryptography and not on designing a real-life protocol. Thus, as noted in subsection "Out- Of-Band-Data" of [14], Alice has to send A, its identity, to Bob so that Bob may select the appropriate credential for the sequel to the conversation. This leads to a slightly complemented version of AKEP2 for EAP-PSK as depicted in Figure 5. Bob Alice | A||RA | |<---------------------------------------------------------| | | | [B||A||RA||RB] | |--------------------------------------------------------->| | | | [A||RB] | |<---------------------------------------------------------| Figure 5: Overview of AKEP2 In AKEP2, o RA and RB are random numbers chosen respectively by Alice and Bob. o A and B are Alice's and Bob's respective identities. They allow Alice and Bob to retrieve the key that they have to use to run an authenticated key exchange between each other. They are also included in the protocol for cryptographic reasons. Bersani & Tschofenig Experimental [Page 19] RFC 4764 EAP-PSK January 2007 o The MACs (see Section 1.3 for the notation "[]") are calculated using a dedicated key. EAP-PSK instantiates this protocol with: o The server as Alice and the peer as Bob. o RA and RB as 16-byte random numbers, using Section 4.1 notations; this means RA=RAND_S and RB=RAND_P. o A and B as Alice's and Bob's respective NAIs, using Section 4.1 notations; this means A=ID_S and B=ID_P. o The MAC algorithm as CMAC with AES-128 using AK and producing a tag length of 16 bytes. o The modified counter mode of operation of AES-128 using KDK, to derive session keys as a result of this exchange. CMAC was chosen as the MAC algorithm because it is capable of handling arbitrary length messages, and its design is simple. It also enjoys up-to-date review by the cryptographic community, especially using provable security concepts. It has been recommended by the NIST. For a detailed description of CMAC, please refer to [8]. In AKEP2, the key exchange is "implicit": the session keys are derived from RB. In EAP-PSK, the session keys are thus derived from RAND_P by using KDK and the modified counter mode of operation of AES-128 described in [5]. This mode was chosen because it is a simple key derivation scheme that relies on a block cipher and has a proof of its security. It is a length increasing function, i.e., it expands one AES-128 input block into a longer t-block output, where t>=2. The derivation of the session keys is shown in Figure 6. Bersani & Tschofenig Experimental [Page 20] RFC 4764 EAP-PSK January 2007 +--------------------------+ +-------------------------------+ | RAND_P | | KDK | | Input Block (16 bytes) | | Key Derivation Key (16 bytes) | +--------------------------+ +-------------------------------+ | | v v +-----------------------------------------------------------------+ | | | Modified Counter Mode | | | +-----------------------------------------------------------------+ | | | v v v +------------+ +----------------------+ +----------------------+ | TEK | | MSK | | EMSK | | (16 bytes) | | (64 bytes) | | (64 bytes) | +------------+ +----------------------+ +----------------------+ Figure 6: Derivation of the Session Keys The input to the derivation of the session keys is RAND_P. The outputs of the derivation of the session keys are: o The 16-byte TEK (the first output block). o The 64-byte MSK (the concatenation of the second to fifth output blocks). o The 64-byte EMSK (the concatenation of the sixth to ninth output blocks). The details of the derivation of the session keys are shown in Figure 7. Bersani & Tschofenig Experimental [Page 21] RFC 4764 EAP-PSK January 2007 +--------------------------+ | RAND_P | | Input Block (16 bytes) | +--------------------------+ | v +----------------+ | | | AES-128(KDK,.) | | | +----------------+ | | +---------------------+-- - - - - - - - - - --+ | | | v v v +--------+ +---+ +--------+ +---+ +--------+ +---+ | c1="1" |->|XOR| | c2="2" |->|XOR|.......| c9="9" |->|XOR| |16 bytes| +---+ |16 bytes| +---+ |16 bytes| +---+ +--------+ | +--------+ | +--------+ | | | | +----------------+ +----------------+ +----------------+ | | | | | | | AES-128(KDK,.) | | AES-128(KDK,.) |......| AES-128(KDK,.) | | | | | | | +----------------+ +----------------+ +----------------+ | | | | | | v v v +-----------------+ +-----------------+ +------------------+ | Output Block #1 | | Output Block #2 | | Output Block #9 | | (16 bytes) | | (16 bytes) |.....| (16 bytes) | | TEK | | MSK (block 1/4) | | EMSK (block 4/4) | +-----------------+ +-----------------+ +------------------+ Figure 7: Derivation of the Session Keys in Details The counter values are set respectively to the first t integers (that is, ci="i", with i=1 to 9). Keying material is sensitive information and should be handled accordingly (see Section 8.10 for further discussion). Bersani & Tschofenig Experimental [Page 22] RFC 4764 EAP-PSK January 2007 3.3. The Protected Channel EAP-PSK provides a protected channel for both parties to communicate over, in case of a successful authentication. This protected channel is currently used to exchange protected result indications and may be used in the future to implement extensions. EAP-PSK uses the EAX mode of operation to provide this protected channel. For a detailed description of EAX, please refer to [4]. Figure 8 shows how EAX is used to implement EAP-PSK protected channel. +-----------+ +----------------+ +---------------------+ +----------+ | Nonce N | | Header H | | Plain Text Payload | | TEK | | 4 bytes | | 22 bytes | | Variable length L | | 16 bytes | +-----------+ +----------------+ +---------------------+ +----------+ | | | | v v v v +-------------------------------------------------------------------+ | | | EAX | | | +-------------------------------------------------------------------+ | | v v +---------------------+ +----------+ | Cipher Text Payload | | Tag | | Variable length L | | 16 bytes | +---------------------+ +----------+ Figure 8: The Protected Channel This protected channel: o Provides replay protection. o Encrypts and authenticates a Plain Text Payload that becomes an Encrypted Payload. The Plain Text Payload must not exceed 960 bytes; see Sections 5.3, 5.4, and 8.11. o Only authenticates a Header that is thus sent in clear. EAX is instantiated with AES-128 as the underlying block cipher. AES-128 is keyed with the TEK. Bersani & Tschofenig Experimental [Page 23] RFC 4764 EAP-PSK January 2007 The nonce N is used to provide cryptographic security to the encryption and data origin authentication as well as protection replay. Indeed, N is a 4-byte sequence number starting from <0> that is monotonically incremented at each EAP-PSK message within one EAP- PSK dialog, except retransmissions, of course. N was taken to be 4 bytes to avoid 16-byte arithmetic. Since EAX uses a 16-byte nonce, N is padded with 96 zero bits for its high- order bits. For cryptographic reasons, N is not allowed to wrap around. In the unlikely, yet possible, event of the server sending an EAP-PSK message with N set to <2**32-2>, it must not send any further message on this protected channel, which would cause to reusing the value 0. Either the conversation is finished after the server receives the EAP-PSK answer from the peer with N set to <2**32-1> and the server proceeds (typically by sending an EAP-Success or Failure), or the conversation is not finished and must then be aborted (a new EAP-PSK dialog may subsequently be started to try again to authenticate). Thus, the maximum number of messages that can be exchanged over the same protected channel is 2**32 (which should not be a limitation in practice, as this is approximately equal to 4 billion). The Header H consists of the first 22 bytes of the EAP Request or Response packet (i.e., the EAP Code, Identifier, Length, and Type fields followed by the EAP-PSK Flags and RAND_S fields). Although it may appear unorthodox that an upper layer (EAP-PSK) protects some information of the lower layer (EAP), this was chosen to comply with EAP recommendation (see Section 7.5. of [3]) and seems to be existing practice at IETF (see, for instance, [35]). The Plain Text Payload is the payload that is to be encrypted and integrity protected. The Cipher Text Payload is the result of the encryption of the Plain Text. The Tag is a MAC that protects both the Header and the Plain Text Payload. The verification of the Tag must only be done after a successful verification of the Nonce for replay protection. If the verification of the Tag succeeds, then the Encrypted Payload is decrypted to recover the Plain Text Payload. If the verification of the Tag fails, then no decryption is performed and this MAC failure should be logged. The tag length is chosen to be 16 bytes for EAX within EAP-PSK. This length is considered appropriate by the cryptographic community. Bersani & Tschofenig Experimental [Page 24] RFC 4764 EAP-PSK January 2007 EAX was mainly chosen because: o It strongly relies on OMAC in its design and OMAC1, a variant of OMAC, had already been chosen in EAP-PSK for the authentication part (please remember that OMAC1 and CMAC are analogous). o Its design is simple. o It enjoys a security proof. o It is free of any Intellectual Property Rights claims. 4. EAP-PSK Message Flows EAP-PSK may consist of two different types of message flows: o The "standard authentication", which is: * Mandatory to implement. * Fully specified in this document. * The simpler type of message flow, which is expected to be used most frequently. o The "extended authentication", which is: * Optional to implement (i.e., there are no mandatory extensions). * Partly specified in this document since it depends on extensions and none are currently specified, let alone in this document. * The type of message flow that should be used when extensions of EAP-PSK are needed by more sophisticated usage scenarios and are available. EAP-PSK introduces the concept of a session to facilitate its analysis and provide a cleaner interface to other layers. A session is a particular instance of an EAP-PSK dialog between two parties. This session is identified by a session identifier. In the first EAP-PSK message, the EAP server asserts its identity. Given that the EAP-Request/Identity and EAP-Response/Identity may not be assumed to have occurred prior to this sending and that the response included in EAP-Response/Identity (if this EAP Identity exchange takes place) may not contain the actual NAI the peer shall Bersani & Tschofenig Experimental [Page 25] RFC 4764 EAP-PSK January 2007 use with EAP-PSK, this means that an EAP server implementing EAP-PSK must use the same EAP server NAI for all EAP-PSK dialogs with any EAP peer implementing EAP-PSK. 4.1. EAP-PSK Standard Authentication EAP-PSK standard authentication is comprised of four messages, i.e., two round-trips; see Figure 9. peer server | Flags||RAND_S||ID_S | |<---------------------------------------------------------| | | | Flags||RAND_S||RAND_P||MAC_P||ID_P | |--------------------------------------------------------->| | | | Flags||RAND_S||MAC_S||PCHANNEL_S_0 | |<---------------------------------------------------------| | | | Flags||RAND_S||PCHANNEL_P_1 | |--------------------------------------------------------->| | | Figure 9: EAP-PSK Standard Authentication o The first message is sent by the server to the peer to: * Send a 16-byte random challenge (RAND_S). RAND_S was called RA in Section 3.2 * State its identity (ID_S). ID_S was denoted by A in Section 3.2. o The second message is sent by the peer to the server to: * Send another 16-byte random challenge (RAND_P). RAND_P was called RB in Section 3.2 * State its identity (ID_P). ID_P was denoted by B in Section 3.2. * Authenticate to the server by proving that it is able to compute a particular MAC (MAC_P), which is a function of the two challenges and AK: MAC_P = CMAC-AES-128(AK, ID_P||ID_S||RAND_S||RAND_P) Bersani & Tschofenig Experimental [Page 26] RFC 4764 EAP-PSK January 2007 o The third message is sent by the server to the peer to: * Authenticate to the peer by proving that it is able to compute another MAC (MAC_S), which is a function of the peer's challenge and AK: MAC_S = CMAC-AES-128(AK, ID_S||RAND_P) * Set up the protected channel (P_CHANNEL_S_0) to: + Confirm that it has derived session keys (at least the TEK). + Give a protected result indication of the authentication. o The fourth message is sent by the peer to the server to finish the setup of the protected channel (P_CHANNEL_P_1) to: * Confirm that it has derived session keys (at least the TEK). * Give a protected result indication of the authentication. The PCHANNEL_S_0 and PCHANNEL_P_1 fields of the third and fourth EAP- PSK messages contain a MAC-computed thanks to TEK that protects the integrity of the messages. For a detailed list of the fields of the messages that are integrity protected, please refer to Section 3.3. All EAP-PSK messages include a sort of header, which is comprised of two fields: o Flags, a 1-byte field that is currently only used to number EAP- PSK messages. o RAND_S, a 16-byte challenge sent by the server that is used as a session identifier. This standard message flow could be comprised of only three messages, like AKEP2, were it not the request/response nature of EAP that prevents the third message to be the last one. Since the fourth message is mandatory, EAP-PSK chose to take advantage of this and set up a protected channel. The standard message flow also includes a statement by the peer of its identity, in addition to the EAP-Response/Identity it may have sent. This behavior follows Section 5.1 of [3], which recommends that the EAP-Response/Identity be used primarily for routing purposes and selecting which EAP method to use, and therefore that EAP methods include a method-specific mechanism for obtaining the identity, so that they do not have to rely on the Identity Response. Bersani & Tschofenig Experimental [Page 27] RFC 4764 EAP-PSK January 2007 When a party receives an EAP-PSK message, it checks that the message is syntactically valid in accordance with the message formats defined in Section 5. If the message is syntactically incorrect, then it is silently discarded. Then it checks the cryptographic validity of this message, i.e., it checks the MAC(s) as follows: o If the received message is the first EAP-PSK message, there is no MAC to check as none is included in message 1. o If the received message is the second EAP-PSK message, the validity of MAC_P is checked. o If the received message is the third EAP-PSK message, the validity of MAC_S is checked and then the validity of the Tag included in P_CHANNEL_S_0 is checked. The validity checks must be done in this order to avoid unnecessarily deriving TEK, MSK, and EMSK in case MAC_S is invalid, meaning that mutual authentication has failed. Indeed, TEK is used to verify the validity of the Tag included in P_CHANNEL_S_0. o If the received message is the fourth EAP-PSK message, the validity of the Tag included in P_CHANNEL_P_1 is checked. If a validity check fails, the message is silently discarded. There can be a counter to track the number of silently discarded messages Section 8.8. If there is an encrypted payload in the message (namely, in the PCHANNEL attribute), then the encrypted payload is decrypted. Then, if the decrypted payload is syntactically incorrect, the message is silently discarded. 4.2. EAP-PSK Extended Authentication To remain simple and yet be extensible to meet future requirements, EAP-PSK provides an extension mechanism within its protected channel: the payload of the protected channel may contain an optional extension field (EXT). Figure 10 presents the message sequence for EAP-PSK extended authentication. Extended authentication MUST be supported, i.e., any EAP-PSK implementation MUST support sending and reception of an EXT attribute according to rules of operation described in Section 6. Yet, although support of the EXT field is mandatory, there is no mandatory extension type to support. This means that if a server engages in EAP-PSK extended authentication, as only the server can start extended authentication per Section 6, a peer will recognize the attempt to start extended authentication through its EXT support. If Bersani & Tschofenig Experimental [Page 28] RFC 4764 EAP-PSK January 2007 the peer does not support the particular extension type used by the server, the peer will still be able to conclude the EAP-PSK dialog. The mandatory support of the EXT field is dictated: o To guarantee a robust behavior in the future where some peers might support some extensions and others not. All peers will thus be able to understand that an extended authentication is being attempted and indicate whether or not they support the extension that is tried. o To ensure that all implementations will indeed be extensible. No extension is currently defined. At most, one extension may be run within a single EAP-PSK dialog: there can neither be sequences of extensions nor interleaved extensions. However, extensions may take a variable number of round- trips to complete. Only the server can start an extension and, if it does so, it must start it in the first payload it sends over the protected channel. peer server | Flags||RAND_S||ID_S | |<---------------------------------------------------------| | | | Flags||RAND_S||RAND_P||MAC_P||ID_P | |--------------------------------------------------------->| | | | Flags||RAND_S||MAC_S||PCHANNEL_S_0(EXT) | |<---------------------------------------------------------| | | | Flags||RAND_S||PCHANNEL_P_1(EXT) | |--------------------------------------------------------->| | | . . . . . . | Flags||RAND_S||PCHANNEL_S_2i(EXT) | |<---------------------------------------------------------| | | | Flags||RAND_S||PCHANNEL_P_2i+1(EXT) | |--------------------------------------------------------->| | | Figure 10: EAP-PSK Extended Authentication Bersani & Tschofenig Experimental [Page 29] RFC 4764 EAP-PSK January 2007 Please refer to Section 6 for more details on how extended authentication works. The PCHANNEL_S_2j and PCHANNEL_P_2j+1 fields of the EAP-PSK messages (where j varies from 0 to i) contain a MAC-computed thanks to TEK that protects the integrity of the messages. For a detailed list of the fields of the messages that are integrity protected, please refer to Section 3.3. When a party receives an EAP-PSK message, it checks that the message is syntactically valid in accordance with the message formats defined in Section 5. If the message is syntactically incorrect, then it is silently discarded. Then it checks the cryptographic validity of this message, i.e., it checks the MAC(s) as follows: o If the received message is the first EAP-PSK message, there is no MAC to check as none is included in message 1. o If the received message is the second EAP-PSK message, the validity of MAC_P is checked. o If the received message is the third EAP-PSK message, the validity of MAC_S is checked and then the validity of the Tag included in P_CHANNEL_S_0 is checked. The validity checks must be done in this order to avoid unnecessarily deriving TEK, MSK, and EMSK in case MAC_S is invalid, meaning that mutual authentication has failed. Indeed, TEK is used to verify the validity of the Tag included in P_CHANNEL_S_0. o If the received message is the fourth EAP-PSK message, the validity of the Tag included in P_CHANNEL_P_1 is checked. o If the received message is an EAP-PSK message different from the first four ones, then validity of the Tag included in P_CHANNEL is checked. If a validity check fails, the message is silently discarded. There can be a counter to track the number of silently discarded messages Section 8.8. If there is an encrypted payload in the message (namely in the PCHANNEL attribute), then the encrypted payload is decrypted. Then, if the decrypted payload is syntactically incorrect, the message is silently discarded. Bersani & Tschofenig Experimental [Page 30] RFC 4764 EAP-PSK January 2007 5. EAP-PSK Message Format For the sake of simplicity, EAP-PSK uses a fixed message format. There are four different types of EAP-PSK messages: o The first EAP-PSK message, which is sent by the server to the peer. o The second EAP-PSK message, which is sent by the peer to the server. o The third EAP-PSK message, which is sent by the server to the peer. o The fourth EAP-PSK message, which is sent by the peer to the server. This is also the type of message that the peer further sends to the server in case of an extended authentication. This is also essentially the type of message that the server further sends to the peer in case of an extended authentication: the only slight modification that occurs in this last case is the setting of the EAP Code to 1 instead of 2 in the other cases. For the sake of clarity, the whole EAP packet that encapsulates the EAP-PSK message (i.e., the EAP-PSK message plus its EAP headers) is depicted in Figures 11, 13, 14, and 18. Bersani & Tschofenig Experimental [Page 31] RFC 4764 EAP-PSK January 2007 5.1. EAP-PSK First Message The first EAP-PSK message is sent by the server to the peer. It has the format presented in Figure 11. 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Code=1 | Identifier | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type EAP-PSK | Flags | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | + + | RAND_S | + + | | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + : : : ID_S : : : + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 11: EAP-PSK First Message Since IANA has allocated EAP method type 47 for EAP-PSK, Type EAP-PSK for the first EAP-PSK message as well as any other EAP-PSK message MUST be 47. The first EAP-PSK message consists of: o A 1-byte Flags field o A 16-byte random number: RAND_S o A variable length field that conveys the server's NAI: ID_S. The length of this field is deduced from the EAP length field. The length of this NAI must not exceed 966 bytes. This restriction aims at avoiding fragmentation issues (see Section 8.11). The Flags field has the format presented in Figure 12. Bersani & Tschofenig Experimental [Page 32] RFC 4764 EAP-PSK January 2007 0 0 1 2 3 4 5 6 7 8 +-+-+-+-+-+-+-+-+ | T | Reserved | +-+-+-+-+-+-+-+-+ Figure 12: EAP-PSK Flags Field The Flags field is comprised of two subfields: o A 2-bit T subfield, which indicates the type of EAP-PSK message: * T=0 for the first EAP-PSK message presented in Section 5.1. * T=1 for the second EAP-PSK message presented in Section 5.2. * T=2 for the third EAP-PSK message presented in Section 5.3. * T=3 for the fourth EAP-PSK message presented in Section 5.4 and the subsequent EAP-PSK messages that may be exchanged during extended authentication. o A 6-bit Reserved subfield that is set to zero on transmission and ignored on reception. The PCHANNEL Nonce field N (see Section 5.3) is used to distinguish between the different EAP-PSK messages that may be exchanged during extended authentication that all have T set to 3, i.e., the fourth EAP-PSK message and possibly the next ones. Bersani & Tschofenig Experimental [Page 33] RFC 4764 EAP-PSK January 2007 5.2. EAP-PSK Second Message The second EAP-PSK message is sent by the peer to the server. It has the format presented in Figure 13. 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Code=2 | Identifier | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type EAP-PSK | Flags | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | + + | RAND_S | + + | | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | + + | RAND_P | + + | | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | +-+-+-+-+-+-+-+-+ + | | + + | MAC_P | + + | | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | +-+-+-+-+-+-+-+-+ + : ID_P : : : + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 13: EAP-PSK Second Message Bersani & Tschofenig Experimental [Page 34] RFC 4764 EAP-PSK January 2007 It consists of: o A 1-byte Flags field o The 16-byte random number sent by the server in the first EAP-PSK message (RAND_S) that serves as a session identifier o A 16-byte random number: RAND_P o A 16-byte MAC: MAC_P o A variable length field that conveys the peer's NAI: ID_P. The length of this field is deduced from the EAP length field. The length of this NAI must not exceed 966 bytes. This restriction aims at avoiding fragmentation issues (see Section 8.11). The Flags field format is presented in Figure 12. Bersani & Tschofenig Experimental [Page 35] RFC 4764 EAP-PSK January 2007 5.3. EAP-PSK Third Message The third EAP-PSK message is sent by the server to the peer. It has the format presented in Figure 14. 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Code=1 | Identifier | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type EAP-PSK | Flags | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | + + | RAND_S | + + | | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | + + | MAC_S | + + | | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | +-+-+-+-+-+-+-+-+ + : PCHANNEL : : : : : + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 14: EAP-PSK Third Message It consists of: o A 1-byte Flags field o The 16-byte random number sent by the server in the first EAP-PSK message (RAND_S) that is used as a session identifier o A 16-byte MAC: MAC_S o A variable length field that constitutes the protected channel: PCHANNEL Bersani & Tschofenig Experimental [Page 36] RFC 4764 EAP-PSK January 2007 The Flags field format is presented in Figure 12. If there is no extension, i.e., if the authentication is standard, the PCHANNEL field consists of: o A 4-byte Nonce N (see Section 3.3). o A 16-byte Tag (see Section 3.3). o A 2-bit result indication flag R. o A 1-bit extension flag E, which is set to 0. o A 5-bit Reserved field, which is set to zero on emission and ignored on reception. R, E, and Reserved are sent encrypted by the protected channel (see Section 3.3). If there is no extension, PCHANNEL has the format presented in Figure 15 (where R, E, and Reserved are presented in the clear for the sake of clarity, although in reality they are sent encrypted). 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Nonce | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | + + | Tag | + + | | + + | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | R |0| Reserved| +-+-+-+-+-+-+-+-+ Figure 15: The PCHANNEL Field with E=0 If there is an extension, i.e., if the authentication is extended, the PCHANNEL field consists of: o A 4-byte Nonce N (see Section 3.3). o A 16-byte Tag (see Section 3.3). Bersani & Tschofenig Experimental [Page 37] RFC 4764 EAP-PSK January 2007 o A 2-bit result indication flag R. o A 1-bit extension flag E, which is set to 1. o A 5-bit Reserved field, which is set to zero on emission and ignored on reception. o A variable length EXT field. R, E, Reserved, and EXT are sent encrypted by the protected channel (see Section 3.3). If there is an extension, PCHANNEL has the format presented in Figure 16 where R, E, Reserved and EXT are presented in the clear for the sake of clarity, although in reality they are sent encrypted). 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Nonce | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | + + | Tag | + + | | + + | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | R |1| Reserved| | +-+-+-+-+-+-+-+-+ + : EXT : : : + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 16: The PCHANNEL Field with E=1 This EXT field is split in two subfields: o The EXT_Type subfield, which indicates the type of the extension o The EXT_Payload subfield, which consists of the payload of the extension. The EXT_Payload length is derived from the EAP Length field. EXT_Payload must have a bit length that is a multiple of 8 bits and must not exceed 960 bytes. The latter restriction aims Bersani & Tschofenig Experimental [Page 38] RFC 4764 EAP-PSK January 2007 at avoiding fragmentation issues (see Section 8.11), whereas the former comes from the EAP length being specified in bytes. The format of the EXT field is presented in Figure 17. 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | EXT_Type | | +-+-+-+-+-+-+-+-+ + : EXT_Payload : : : + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 17: The EXT Field 5.4. EAP-PSK Fourth Message The fourth EAP-PSK message is sent by the peer to the server. It has the format presented in Figure 18. 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Code=2 | Identifier | Length | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Type EAP-PSK | Flags | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + | | + + | RAND_S | + + | | + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ + : : : PCHANNEL : : : : : + +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 18: EAP-PSK Fourth Message Bersani & Tschofenig Experimental [Page 39] RFC 4764 EAP-PSK January 2007 It consists of: o A 1-byte Flags field o The 16-byte random number sent by the server in the first EAP-PSK message (RAND_S) that is used as a session identifier o A variable length field that constitutes the protected channel: PCHANNEL The Flags field format is presented in Figure 12. The PCHANNEL field has the following structure, which was already described in Section 5.3. If there is no extension, i.e., if the authentication is standard, the PCHANNEL field consists of: o A 4-byte Nonce N (see Section 3.3). o A 16-byte Tag (see Section 3.3). o A 2-bit result indication flag R. o A 1-bit extension flag E, which is set to 0. o A 5-bit Reserved field, which is set to zero on emission and ignored on reception. R, E, and Reserved are sent encrypted by the protected channel (see Section 3.3). If there is no extension, PCHANNEL has the format presented in Figure 15. If there is an extension, i.e., if the authentication is extended, the PCHANNEL field consists of: o A 4-byte Nonce N (see Section 3.3). o A 16-byte Tag (see Section 3.3). o A 2-bit result indication flag R. o A 1-bit extension flag E, which is set to 1. o A 5-bit Reserved field, which is set to zero on emission and ignored on reception. Bersani & Tschofenig Experimental [Page 40] RFC 4764 EAP-PSK January 2007 o A variable length EXT field. R, E, Reserved, and EXT are sent encrypted by the protected channel (see Section 3.3). If there is an extension, PCHANNEL has the format presented in Figure 16. This EXT field is split in two subfields: o The EXT_Type subfield, which indicates the type of the extension o The EXT_Payload subfield, which consists of the payload of the extension. The EXT_Payload length is derived from the EAP Length field. EXT_Payload must have a bit length that is a multiple of 8 bits and must not exceed 960 bytes. The latter restriction aims at avoiding fragmentation issues (see Section 8.11). The format of the EXT field is presented in Figure 17. 6. Rules of Operation for the EAP-PSK Protected Channel In this section, the rules of operation of the EAP-PSK protected channel are presented: o How protected result indications are implemented. o How an extended authentication works in details. 6.1. Protected Result Indications The R flag of the PCHANNEL field in the third and fourth types of EAP-PSK messages is used to provide result indications. Since this 2-bit flag is communicated over the protected channel, it is: o Encrypted so that only the peer and the server can know its value. o Integrity-protected so that it cannot be modified by an attacker without the peer or the server detecting this modification. o Protected against replays. This 2-bit R flag can take the following values: o 01 to mean CONT Bersani & Tschofenig Experimental [Page 41] RFC 4764 EAP-PSK January 2007 o 10 to mean DONE_SUCCESS o 11 to mean DONE_FAILURE The peer and the server each remember some information about both the values of R that they have sent and the values of R they have received. It is the conjunction of both sent and received R values that indicate the success or the failure of the EAP-PSK dialog. In the case of a standard authentication, the following values of R should be exchanged: o Either the server sends a DONE_SUCCESS in the PCHANNEL of the third EAP-PSK message, to which the peer replies with a DONE_SUCCESS in the PCHANNEL of the fourth EAP-PSK message, which successfully ends the EAP-PSK dialog. o Or the server sends a DONE_FAILURE in the PCHANNEL of the third EAP-PSK message, to which the peer replies with a DONE_FAILURE in the PCHANNEL of the fourth EAP-PSK message, which unsuccessfully ends the EAP-PSK dialog. In the case of an extended authentication, more complex exchanges may occur, which is why the CONT value was introduced. The rules of operation for each value that R may take are detailed below. 6.1.1. CONT The server and the peer each initialize the values of R they intend to send and receive as CONT. Here CONT stands for "Continue". It indicates that the EAP-PSK dialog is not yet successful and that the party sending it wants to continue the dialog to try and reach success. Indeed, although the peer and the server must have successfully authenticated each other, thanks to MAC_P and MAC_S, before they start communicating over the protected channel, the EAP-PSK dialog may not yet be deemed successful after this mutual authentication because of authorization issues. For instance, a prepaid customer of a wireless Hot-Spot might have successfully authenticated but has to refill its account, e.g., with a credit card transaction over the protected channel, before it is authorized. Bersani & Tschofenig Experimental [Page 42] RFC 4764 EAP-PSK January 2007 6.1.2. DONE_SUCCESS DONE_SUCCESS indicates that the party that sent it deems the EAP-PSK dialog successful and therefore proposes to end this dialog. Once the server has sent a DONE_SUCCESS, it must keep sending this value for R. The peer must first receive a DONE_SUCCESS from the server before it is allowed to send a DONE_SUCCESS. After the peer has received a DONE_SUCCESS from the server, it may: o Send a CONT to the server if it has not reached success on its side. The server that receives a CONT should continue the EAP-PSK dialog (see Section 8.2 for some discussion on the security implications of this). o Send a DONE_SUCCESS to the server, which will end the EAP-PSK dialog with success. o Send a DONE_FAILURE to the server, which will end the EAP-PSK dialog with failure. 6.1.3. DONE_FAILURE DONE_FAILURE indicates that the party that sent it deems the EAP-PSK dialog unsuccessful and proposes to end this dialog because nothing will make it change its mind. If the server is the first to send a DONE_FAILURE, then the peer that receives this DONE_FAILURE must reply with a DONE_FAILURE and fail, which ends the EAP-PSK dialog. If the peer is the first to send a DONE_FAILURE, then the server that receives this DONE_FAILURE must immediately end this EAP-PSK dialog without sending any further EAP-PSK message, and fail. 6.2. Extended Authentication An extended authentication can only be started by the server. Exactly one extension (identified by the EXT_Type subfield of the EXT field) must be run during an EAP-PSK extended authentication dialog. The extension is run over the protected channel: it can assume confidentiality, integrity, and replay protection. Bersani & Tschofenig Experimental [Page 43] RFC 4764 EAP-PSK January 2007 To start an extended authentication, the server sets the PCHANNEL E flag to 1 and includes the EXT_Payload of the extension it has chosen. Since EAP-PSK does not provide fragmentation, the extension must not send an EXT_Payload larger than 960 bytes, which corresponds to the 1020-byte EAP MTU that may minimally be assumed (see [3]). Moreover, an extension must not send an empty EXT_Payload (because this has a particular meaning for EAP-PSK; see below). When the peer receives the third EAP-PSK message with the E flag set to 1, it checks whether it is able to process the proposed extension. If the peer is not able to process the proposed extension, i.e., it does not recognize the EXT_Type of the proposed extension, it sets E=1 in its reply (the fourth EAP-PSK message) and include an EXT field of the same EXT_Type but with an empty EXT_Payload. Depending on the values taken by the R flags, the EAP-PSK dialog may: o End * If the peer's policy mandates that it fails in the case of an unrecognized extension, it sends a DONE_FAILURE in the fourth EAP-PSK message. * If the server has sent a DONE_SUCCESS in the third EAP-PSK message, and the peer's policy authorizes it to succeed even if the extension is not recognized, the peer sends a DONE_SUCCESS. o Continue for exactly one round-trip; namely, in case the server has sent a CONT in the third EAP-PSK message and the peer's policy authorizes it to succeed even if the extension is not recognized, the peer replies with a CONT in the fourth EAP-PSK message. The server must then, depending on its policy, send either a DONE_SUCCESS or a DONE_FAILURE to the peer in the fifth EAP-PSK message. If the server sent a DONE_SUCCESS in the fifth EAP-PSK message, the peer must send a DONE_SUCCESS in the sixth EAP-PSK message. All these messages must have the E flag set to 1 with an EXT field with the EXT_Type of the extension that was proposed and an empty EXT_Payload (this behavior was chosen to simplify implementations). If the peer is able to process the proposed extension, then it does so. In this case, the extension must be aware of the R values sent and received and able to propose to update them. All the subsequent messages exchanged between the peer and the server must have the E Bersani & Tschofenig Experimental [Page 44] RFC 4764 EAP-PSK January 2007 flag set to 1 with an EXT field of the EXT_Type of the extension that was proposed and a non-empty EXT_Payload. 7. IANA Considerations This section provides guidance to the IANA regarding registration of values related to the EAP-PSK protocol, in accordance with [6]. The following terms are used here with the meanings defined in [6]: "name space" and "registration". The following policies are used here with the meanings defined in [6]: "Expert Review" and "Specification Required". This document introduces one new Internet Assigned Numbers Authority (IANA) consideration: there is one name space in EAP-PSK that requires registration: the EXT_Type values (see Section 5.3 and Section 5.4). For registration requests where a Designated Expert should be consulted, the responsible IETF Area Director should appoint the Designated Expert. The intention is that any allocation will be accompanied by a published RFC. But in order to allow for the allocation of values prior to the RFC being approved for publication, the Designated Expert can approve allocations once it seems clear that an RFC will be published. The Designated Expert will post a request to the EAP WG mailing list (or a successor designated by the Area Director) for comment and review, including an Internet-Draft. Before a period of 30 days has passed, the Designated Expert will either approve or deny the registration request and publish a notice of the decision to the EAP WG mailing list or its successor, as well as informing IANA. A denial notice must be justified by an explanation and, in the cases where it is possible, concrete suggestions on how the request can be modified so as to become acceptable. 7.1. Allocation of an EAP-Request/Response Type for EAP-PSK IANA allocated a new EAP Type for EAP-PSK. 7.2. Allocation of EXT Type Numbers EAP-PSK is not intended as a general-purpose protocol, and allocations of EXT_Type should not be made for purposes unrelated to authentication, authorization, and accounting. EXT_Type numbers have a range from 1 to 255. Bersani & Tschofenig Experimental [Page 45] RFC 4764 EAP-PSK January 2007 EXT_Type 255 has been allocated for Experimental use. EXT_Type 1-254 may be allocated on the advice of a Designated Expert, with Specification Required. 8. Security Considerations [3] highlights several attacks that are possible against EAP, as EAP does not provide any robust security mechanism. This section discusses the claimed security properties of EAP-PSK as well as vulnerabilities and security recommendations in the threat model of [3]. 8.1. Mutual Authentication EAP-PSK provides mutual authentication. The server believes that the peer is authentic because it can calculate a valid MAC and the peer believes that the server is authentic because it can calculate another valid MAC. The authentication protocol that inspired EAP-PSK, AKEP2, enjoys a security proof in the provable security paradigm; see [14]. The MAC algorithm used in the instantiation of AKEP2 within EAP-PSK, CMAC, also enjoys a security proof in the provable security paradigm; see [29]. A tag length of 16 bytes for CMAC is currently deemed appropriate by the cryptographic community for entity authentication. The underlying block cipher used, AES-128, is widely believed to be a secure block cipher. Finally, the key used for mutual authentication, AK, is only used for that purpose, which makes this part cryptographically independent of the other parts of the protocol. EAP-PSK provides mutual authentication if it is based on a pairwise PSK of sufficient strength. If the PSK is not pairwise or not sufficiently strong, then it does not provide authentication. In this way, EAP-PSK is no different than other authentication protocols based on Pre-Shared Keys. Bersani & Tschofenig Experimental [Page 46] RFC 4764 EAP-PSK January 2007 8.2. Protected Result Indications EAP-PSK provides protected result indications thanks to its 2-bit R flag (see Section 6.1). This 2-bit R flag is protected because it is encrypted and integrity protected by the EAX mode of operation; see Section 3.3. Care may be taken against Byzantine failures, that is to say, for instance, when a peer tries to force a server to engage in a never- ending conversation. This could, for example, be done by a peer that keeps sending a CONT after it has received a DONE_SUCCESS from the server. A policy may limit the number of rounds in an EAP-PSK extended authentication to mitigate this threat, which is outside our threat model. It should also be noted that the cryptographic protection of the result indications does not prevent message deletion. For instance, let us consider a scenario in which: o A server sends a DONE_SUCCESS to a peer. o The peer replies with a DONE_SUCCESS. In the case that the last message from the peer is intercepted, and an EAP Success is sent to the peer before any retransmission from the server reaches it, or the retransmissions from the server are also deleted, the peer will believe that it has successfully authenticated to the server while the server will fail. This behavior is well known (see, e.g., [23]) and in a sense unavoidable. There is a trade-off between efficiency and the "level" of information sharing that is attainable. EAP-PSK specified a single round-trip of DONE_SUCCESS because it is believed that: o If there is an adversary capable of disrupting the communication channel, it can do so whenever it wants (be it after 1 or 10 round-trips or even during data communication). o Other layers/applications will generally start by doing a specific key exchange and confirmation procedure using the keys derived by EAP-PSK. This is typically done by IEEE 802.11i "four-way handshake". In case the error is not detected by EAP-PSK, it should be detected then (please note, however, that it is bad practice to rely on an external mechanism to ensure synchronization, unless this is an explicit property of the external mechanism). Bersani & Tschofenig Experimental [Page 47] RFC 4764 EAP-PSK January 2007 8.3. Integrity Protection EAP-PSK provides integrity protection thanks to the Tag of its protected channel (see Section 3.3). EAP-PSK provides integrity protection if it is based on a pairwise PSK of sufficient strength. If the PSK is not pairwise or not sufficiently strong, then it does not provide authentication. In this way, it is no different than other authentication protocols based on Pre-Shared Keys. 8.4. Replay Protection EAP-PSK provides replay protection of its mutual authentication part thanks to the use of random numbers RAND_S and RAND_P. Since RAND_S is 128 bits long, one expects to have to record 2**64 (i.e., approximately 1.84*10**19) EAP-PSK successful authentications before an authentication can be replayed. Hence, EAP-PSK provides replay protection of its mutual authentication part as long as RAND_S and RAND_P are chosen at random; randomness is critical for security. EAP-PSK provides replay protection during the conversation of the protected channel thanks to the Nonce N of its protected channel (see Section 3.3). This nonce is initialized to 0 by the server and monotonically incremented by one by the party that receives a valid EAP-PSK message. For instance, after receiving from the server a valid EAP-PSK message with Nonce set to x, the peer will answer with an EAP-PSK message with Nonce set to x+1 and wait for an EAP-PSK message with Nonce set to x+2. A retransmission of the server's message with Nonce set to x would cause the peer EAP layer to resend the message in which Nonce was set to x+1, which would be transparent to the EAP-PSK layer. The EAP peer must check that the Nonce is indeed initialized to 0 by the server. 8.5. Reflection Attacks EAP-PSK provides protection against reflection attacks in case of an extended authentication because: o It integrity protects the EAP header (which contains the indication Request/Response. o It includes two separate spaces for the Nonces: the EAP server only receives messages with odd nonces, whereas the EAP peer only receives messages with even nonces. Bersani & Tschofenig Experimental [Page 48] RFC 4764 EAP-PSK January 2007 8.6. Dictionary Attacks Because EAP-PSK is not a password protocol, it is not vulnerable to dictionary attacks. Indeed, the PSK used by EAP-PSK must not be derived from a password. Derivation of the PSK from a password may lead to dictionary attacks. However, using a 16-byte PSK has: o Ergonomic impacts: some people may find it cumbersome to manually provision a 16-byte PSK. o Deployment impacts: some people may want to reuse existing credential databases that contain passwords and not PSKs. Because people will probably not heed the warning not to use passwords, guidance to derive a PSK from a password is provided in Appendix A. The method proposed in Appendix A only tries to make dictionary attacks harder. It does not eliminate them. However, it does not cause a fatal error if passwords are used instead of PSKs: people rarely use password-derived certificates, so why should they do so for shared keys? 8.7. Key Derivation EAP-PSK supports key derivation. The key hierarchy is specified in Section 2.1. The mechanism used for key derivation is the modified counter mode. The instantiation of the modified counter in EAP-PSK complies with the conditions stated in [5] so that the security proof for this mode holds. The underlying block cipher used, AES-128, is widely believed to be a secure block cipher. A first key derivation occurs to calculate AK and KDK from the PSK: it is called the key setup (see Section 3.1). It uses the PSK as the key to the modified counter mode. Thus, AK and KDK are believed to be cryptographically separated and computable only to those who have knowledge of the PSK. Bersani & Tschofenig Experimental [Page 49] RFC 4764 EAP-PSK January 2007 A second key derivation occurs to derive session keys, namely, the TEK, MSK, and EMSK (see Section 3.2). It uses KDK as the key to the modified counter mode. The protocol design explicitly assumes that neither AK nor KDK are shared beyond the two parties utilizing them. AK loses its efficacy to mutually authenticate the peer and server with each other when it is shared. Similarly, the derived TEK, MSK, and EMSK lose their value when KDK is shared with a third party. It should be emphasized that the peer has control of the session keys derived by EAP-PSK. In particular, it can easily choose the random number it sends in EAP-PSK so that one of the nine derived 16-byte key blocks (see Section 2.1) takes a pre-specified value. It was chosen not to prevent this control of the session keys by the peer because: o Preventing it would have added some complexity to the protocol (typically, the inclusion of a one-way mode of operation of AES in the key derivation part). o It is believed that the peer won't try to force the server to use some pre-specified value for the session keys. Such an attack is outside the threat model and seems to have little value compared to a peer sharing its PSK. However, this is not the behavior recommended by EAP in Section 7.10 of [3]. Since deriving the session keys requires some cryptographic computations, it is recommended that the session keys be derived only once authentication has succeeded (i.e., once the server has successfully verified MAC_P for the server side, and once the peer has successfully verified MAC_S for the peer side). It is recommended to take great care in implementations, so that derived keys are not made available if the EAP-PSK dialog fails (e.g., ends with DONE_FAILURE). The TEK must not be made available to anyone except to the current EAP-PSK dialog. Bersani & Tschofenig Experimental [Page 50] RFC 4764 EAP-PSK January 2007 8.8. Denial-of-Service Resistance Denial of Service (DoS) resistance has not been a design goal for EAP-PSK. It is, however, believed that EAP-PSK does not provide any obvious and avoidable venue for such attacks. It is worth noting that the server has to do a cryptographic calculation and maintain some state when it engages in an EAP-PSK conversation, namely, generate and remember the 16-byte RAND_S. However, this should not lead to resource exhaustion as this state and the associated computation are fairly lightweight. Please note that both the peer and the server must commit to their RAND_S and RAND_P to protect their partners from flooding attacks. It is recommended that EAP-PSK not allow EAP notifications to be interleaved in its dialog to prevent potential DoS attacks. Indeed, since EAP notifications are not integrity protected, they can easily be spoofed by an attacker. Such an attacker could force a peer that allows EAP notifications to engage in a discussion that would delay his or her authentication or result in the peer taking unexpected actions (e.g., in case a notification is used to prompt the peer to do some "bad" action). It is up to the implementation of EAP-PSK or to the peer and the server to specify the maximum number of failed cryptographic checks that are allowed. For instance, does the reception of a bogus MAC_P in the second EAP-PSK message cause a fatal error or is it discarded to continue waiting for the valid response of the valid peer? There is a trade-off between possibly allowing multiple tentative forgeries and allowing a direct DoS (in case the first error is fatal). For the sake of simplicity and denial-of-service resilience, EAP-PSK has chosen not to include any error messages. Hence, an "invalid" EAP-PSK message is silently discarded. Although this makes interoperability testing and debugging harder, this leads to simpler implementations and does not open any venue for denial-of-service attacks. 8.9. Session Independence Thanks to its key derivation mechanisms, EAP-PSK provides session independence: passive attacks (such as capture of the EAP conversation) or active attacks (including compromise of the MSK or EMSK) do not enable compromise of subsequent or prior MSKs or EMSKs. Bersani & Tschofenig Experimental [Page 51] RFC 4764 EAP-PSK January 2007 The assumption that RAND_P and RAND_S are random is central for the security of EAP-PSK in general and session independence in particular. 8.10. Exposition of the PSK EAP-PSK does not provide Perfect Forward Secrecy. Compromise of the PSK leads to compromise of recorded past sessions. Compromise of the PSK enables the attacker to impersonate the peer and the server: compromise of the PSK leads to "full" compromise of future sessions. EAP-PSK provides no protection against a legitimate peer sharing its PSK with a third party. Such protection may be provided by appropriate repositories for the PSK, whose choice is outside the scope of this document. The PSK used by EAP-PSK must only be shared between two parties: the peer and the server. In particular, this PSK must not be shared by a group of peers communicating with the same server. The PSK used by EAP-PSK must be cryptographically separated from keys used by other protocols, otherwise the security of EAP-PSK may be compromised. It is a rule of thumb in cryptography to use different keys for different applications. 8.11. Fragmentation EAP-PSK does not support fragmentation and reassembly. Indeed, the largest EAP-PSK frame is at most 1015 bytes long, because: o The maximum length for the peer NAI identity used in EAP-PSK is 966 bytes (see Section 5.2). This should not be a limitation in practice (see Section 2.2 of [2] for more considerations on NAI length). o The maximum length for the EXT_Payload field used in EAP-PSK is 960 bytes (see Section 5.3 and Section 5.4). Per Section 3.1 of [3], the lower layers over which EAP may be run are assumed to have an EAP MTU of 1020 bytes or greater. Since the EAP header is 5 bytes long, supporting fragmentation for EAP-PSK is unnecessary. Extensions that require sending a payload larger than 960 bytes should provide their own fragmentation and reassembly mechanism. Bersani & Tschofenig Experimental [Page 52] RFC 4764 EAP-PSK January 2007 8.12. Channel Binding EAP-PSK does not provide channel binding as this feature is still very much a work in progress (see [13]). However, it should be easy to add it to EAP-PSK as an extension (see Section 4.2). 8.13. Fast Reconnect EAP-PSK does not provide any fast reconnect capability. Indeed, as noted, for instance, in [15], mutual authentication (without counters or timestamps) requires three exchanges, thus four exchanges in EAP since any EAP-Request must be answered to by an EAP- Response. Since this minimum bound is already reached in EAP-PSK standard authentication, there is no way the number of round-trips used within EAP-PSK can be reduced without using timestamps or counters. Timestamps and counters were deliberately avoided for the sake of simplicity and security (e.g., synchronization issues). 8.14. Identity Protection Since it was chosen to restrict to a single cryptographic primitive from symmetric cryptography, namely, the block cipher AES-128, it appears that it is not possible to provide "reasonable" identity protection without failing to meet the simplicity goal. Hereafter is an informal discussion of what is meant by identity protection and the rationale behind the requirement of identity protection. For some complementary discussion, refer to [37]. Identity protection basically means preventing the disclosure of the identities of the communicating parties over the network, which is quite contradictory to authentication. There are two levels of identity protection: protection against passive attackers and protection against active eavesdroppers. As explained in [37], "a common example [for identity protection] is the case of mobile devices wishing to prevent an attacker from correlating their (changing) location with the logical identity of the device (or user)". If only symmetric cryptography is used, only a weak form of identity protection may be offered, namely, pseudonym management. In other words, the peer and the server agree on pseudonyms that they use to Bersani & Tschofenig Experimental [Page 53] RFC 4764 EAP-PSK January 2007 identify each other and usually change them periodically, possibly in a protected way so that an attacker cannot learn new pseudonyms before they are used. With pseudonym management, there is a trade-off between allowing for pseudonym resynchronization (thanks to a permanent identity) and being vulnerable to active attacks (in which the attacker forges messages simulating a pseudonym desynchronization). Indeed, a protocol using time-varying pseudonyms may want to anticipate "desynchronization" situations such as, for instance, when the peer believes that its current pseudonym is "pseudo1@bigco.com" whereas the server believes this peer will use the pseudonym "pseudo2@bigco.com" (which is the pseudonym the server has sent to update "pseudo1@bigco.com"). Because pseudonym management adds complexity to the protocol and implies this unsatisfactory trade-off, it was decided not to include this feature in EAP-PSK. However, EAP-PSK may trivially provide some protection when the concern is to avoid the "real-life" identity of the user being "discovered". For instance, let us take the example of user John Doe that roams and connects to a Hot-Spot owned and operated by Wireless Internet Service Provider (WISP) BAD. Suppose this user authenticates to his home WISP (WISP GOOD) with an EAP method under an identity (e.g., "john.doe@wispgood.com") that allows WISP BAD (or an attacker) to recover his "real-life" identity, i.e., John Doe. An example drawback of this is when a competitor of John Doe's WISP wants to win John Doe as a new customer by sending him some special targeted advertisement. EAP-PSK can very simply thwart this attack, merely by avoiding to provide John Doe with an NAI that allows easy recovery of his real- life identity. It is believed that when an NAI that is not correlated to a real-life identity is used, no valuable information leaks because of the EAP method. Indeed, the identity of the WISP used by a peer has to be disclosed anyway in the realm portion of its NAI to allow AAA routing. Moreover, the Medium Access Control Address of the peer's Network Interface Card can generally be used to track the peer as efficiently as a fixed NAI. Bersani & Tschofenig Experimental [Page 54] RFC 4764 EAP-PSK January 2007 It is worth noting that the server systematically discloses its identity, which may allow probing attacks. This may not be a problem as the identity of the server is not supposed to remain secret. On the contrary, users tend to want to know to whom they will be talking in order to choose the right network to attach to. 8.15. Protected Ciphersuite Negotiation EAP-PSK does not allow negotiating ciphersuites. Hence, it is not vulnerable to negotiation attacks and does not implement protected ciphersuite negotiation. 8.16. Confidentiality Although EAP-PSK provides confidentiality in its protected channel, it cannot claim to do so as per Section 7.2.1 of [3]: "A method making this claim must support identity protection". 8.17. Cryptographic Binding Since EAP-PSK is not intended to be tunneled within another protocol that omits peer authentication, it does not implement cryptographic binding. 8.18. Implementation of EAP-PSK To really provide security, not only must a protocol be well thought- out and correctly specified, but its implementation must take special care. For instance, implementing cryptographic algorithms requires special skills since cryptographic software is vulnerable not only to classical attacks (e.g., buffer overflow or missing checks) but also to some special cryptographic attacks (e.g., side channels attacks like timing ones; see [36]). In particular, care must be taken to avoid such attacks in EAX implementation; please refer to [4] for a note on this point. An EAP-PSK implementation should use a good source of randomness to generate the random numbers required in the protocol. Please refer to [20] for more information on generating random numbers for security applications. Handling sensitive material (namely, keying material such as the PSK, AK, KDK, etc.) should be done in a secure way (see, for instance, [19] for guidance on secure deletion). Bersani & Tschofenig Experimental [Page 55] RFC 4764 EAP-PSK January 2007 The specification of a repository for the PSK that EAP-PSK uses is outside the scope of this document. In particular, nothing prevents one from storing this PSK on a tamper-resistant device such as a smart card rather than having it memorized or written down on a sheet of paper. The choice of the PSK repository may have important security impacts. 9. Security Claims This section provides the security claims required by [3]. [a] Mechanism. EAP-PSK is based on symmetric cryptography (AES-128) and uses a 16-byte Pre-Shared Key (PSK). [b] Security claims. EAP-PSK provides: * Mutual authentication (see Section 8.1) * Integrity protection (see Section 8.3) * Replay protection (see Section 8.4) * Key derivation (see Section 8.7) * Dictionary attack resistance (see Section 8.6) * Session independence (see Section 8.9) [c] Key strength. EAP-PSK provides a 16-byte effective key strength. [d] Description of key hierarchy. Please see Section 2.1. [e] Indication of vulnerabilities. EAP-PSK does not provide: * Identity protection (see Section 8.14) * Confidentiality (see Section 8.16) * Fast reconnect (see Section 8.13) * Fragmentation (see Section 8.11) * Cryptographic binding (see Section 8.17) * Protected ciphersuite negotiation (see Section 8.15) * Perfect Forward Secrecy (see Section 8.10) Bersani & Tschofenig Experimental [Page 56] RFC 4764 EAP-PSK January 2007 * Key agreement: the session key is chosen by the peer (see Section 8.7) * Channel binding (see Section 8.12) 10. Acknowledgments This EAP method has been inspired by EAP-Archie and EAP-SIM. Many thanks to their respective authors: Jesse Walker (extra thanks to Jesse Walker for his thorough and challenging expert review of EAP- PSK), Russ Housley, Henry Haverinen, and Joseph Salowey. Thanks to o Henri Gilbert for some interesting discussions on the cryptographic parts of EAP-PSK. o Aurelien Magniez for his valuable feedback on network aspects of EAP-PSK, his curiosity and rigor that led to numerous improvements, and his help in the first implementation of EAP-PSK under Microsoft Windows and Freeradius. o Thomas Otto for his valuable feedback on EAP-PSK and the implementation of the first version of EAP-PSK under Xsupplicant. o Nancy Cam-Winget for some exchanges on EAP-PSK. o Jari Arkko and Bernard Aboba, the beloved EAP WG chairs, for the work they stimulate. Finally, thanks to Vir Z., who has brought a permanent and outstanding though discreet contribution to this protocol. 11. References 11.1. Normative References [1] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [2] Aboba, B., Beadles, M., Arkko, J., and P. Eronen, "The Network Access Identifier", RFC 4282, December 2005. [3] Aboba, B., Blunk, L., Vollbrecht, J., Carlson, J., and H. Levkowetz, "Extensible Authentication Protocol (EAP)", RFC 3748, June 2004. Bersani & Tschofenig Experimental [Page 57] RFC 4764 EAP-PSK January 2007 [4] Bellare, M., Rogaway, P., and D. Wagner, "The EAX mode of operation", FSE 04, Springer-Verlag LNCS 3017, 2004. [5] Gilbert, H., "The Security of One-Block-to-Many Modes of Operation", FSE 03, Springer-Verlag LNCS 2287, 2003. [6] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 2434, October 1998. [7] National Institute of Standards and Technology, "Specification for the Advanced Encryption Standard (AES)", Federal Information Processing Standards (FIPS) 197, November 2001. [8] National Institute of Standards and Technology, "Recommendation for Block Cipher Modes of Operation: The CMAC Mode for Authentication", Special Publication (SP) 800-38B, May 2005. 11.2. Informative References [9] Aboba, B., Simon, D., Eronen, P., and H. Levkowetz,"Extensible Authentication Protocol (EAP) Key Management Framework", Work in Progress, October 2006. [10] Aboba, B., Calhoun, P., Glass, S., Hiller, T., McCann, P., Shiino, H., Zorn, G., Dommety, G., Perkins, C., Patil, B., Mitton, D., Manning, S., Beadles, M., Walsh, P., Chen, X., Sivalingham, S., Hameed, A., Munson, M., Jacobs, S., Lim, B., Hirschman, B., Hsu, R., Xu, Y., Campell, E., Baba, S., and E. Jaques, "Criteria for Evaluating AAA Protocols for work Access", RFC 2989, November 2000. [11] Aboba, B. and D. Simon, "PPP EAP TLS Authentication Protocol", RFC 2716, October 1999. [12] Arkko, J. and H. Haverinen, "Extensible Authentication Protocol Method for 3rd Generation Authentication and Key Agreement (EAP-AKA)", RFC 4187, January 2006. [13] Arkko, J. and P. Eronen, "Authenticated Service Information for the Extensible Authentication Protocol (EAP)", Work in Progress, October 2005. [14] Bellare, M. and P. Rogaway, "Entity Authentication and Key Distribution", CRYPTO 93, Springer-Verlag LNCS 773, 1994. Bersani & Tschofenig Experimental [Page 58] RFC 4764 EAP-PSK January 2007 [15] Bellare, M., Pointcheval, D., and P. Rogaway, "Authenticated Key Exchange Secure Against Dictionary attacks", EUROCRYPT 00, Springer-Verlag LNCS 1807, 2000. [16] Bersani, F., "EAP shared key methods: a tentative synthesis of those proposed so far", Work in Progress, April 2004. [17] Bradner, S., "The Internet Standards Process -- Revision 3", BCP 9, RFC 2026, October 1996. [18] Carlson, J., Aboba, B., and H. Haverinen, "EAP SRP-SHA1 Authentication Protocol", Work in Progress, July 2001. [19] Department of Defense of the United States, "National Industrial Security Program Operating Manual", DoD 5220-22M, January 1995. [20] Eastlake, D., Schiller, J., and S. Crocker, "Randomness Requirements for Security", BCP 106, RFC 4086, June 2005. [21] Funk, P. and S. Blake-Wilson, "EAP Tunneled TLS Authentication Protocol (EAP-TTLS)", Work in Progress, July 2004. [22] Haller, N., Metz, C., Nesser, P., and M. Straw, "A One-Time Password System", RFC 2289, February 1998. [23] Halpern, J. and Y. Moses, "Knowledge and common knowledge in a distributed environment", Journal of the ACM 37:3, 1990. [24] Haverinen, H. and J. Salowey, "Extensible Authentication Protocol Method for Global System for Mobile Communications (GSM) Subscriber Identity Modules (EAP-SIM)", RFC 4186, January 2006. [25] Huitema, C., Postel, J., and S. Crocker, "Not All RFCs are Standards", RFC 1796, April 1995. [26] Institute of Electrical and Electronics Engineers, "Local and Metropolitan Area Networks: Port-Based Network Access Control", IEEE Standard 802.1X, September 2001. [27] Institute of Electrical and Electronics Engineers, "Approved Draft Supplement to Standard for Telecommunications and Information Exchange Between Systems-LAN/MAN Specific Requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications: Specification for Enhanced Security", IEEE 802.11i-2004, 2004. Bersani & Tschofenig Experimental [Page 59] RFC 4764 EAP-PSK January 2007 [28] Institute of Electrical and Electronics Engineers, "Standard for Telecommunications and Information Exchange Between Systems - LAN/MAN Specific Requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications", IEEE Standard 802.11, 1999. [29] Iwata, T. and K. Kurosawa, "OMAC: One-Key CBC MAC", FSE 03, Springer-Verlag LNCS 2887, 2003. [30] Jablon, D., "The SPEKE Password-Based Key Agreement Methods", Work in Progress, November 2002. [31] Josefsson, S., "The EAP SecurID(r) Mechanism", Work in Progress, February 2002. [32] Josefsson, S., Palekar, A., Simon, D., and G. Zorn, "Protected EAP Protocol (PEAP) Version 2", Work in Progress, October 2004. [33] Kaliski, B., "PKCS #5: Password-Based Cryptography Specification Version 2.0", RFC 2898, September 2000. [34] Kamath, V. and A. Palekar, "Microsoft EAP CHAP Extensions", Work in Progress, April 2004. [35] Kent, S., "IP Authentication Header", RFC 4302, December 2005 [36] Kocher, P., "Timing Attacks on Implementations of Diffie- Hellman, RSA, DSS, and Other Systems", CRYPTO 96, Springer- Verlag LNCS 1109, 1996. [37] Krawczyk, H., "SIGMA: the `SIGn-and-MAc' Approach to Authenticated Diffie-Hellman and its Use in the IKE Protocols", CRYPTO 03, Springer-Verlag LNCS 2729, June 2003. [38] MacNally, C., "Cisco LEAP protocol description", September 2001, available from . [39] Metz, C., "OTP Extended Responses", RFC 2243, November 1997. [40] Menezes, A., van Oorschot, P., and S. Vanstone, "Handbook of Applied Cryptography", CRC Press , 1996. [41] National Institute of Standards and Technology, "Password Usage", Federal Information Processing Standards (FIPS) 112, May 1985. Bersani & Tschofenig Experimental [Page 60] RFC 4764 EAP-PSK January 2007 [42] Cam-Winget, N., McGrew, D., Salowey, J., and H. Zhou, "The Flexible Authentication via Secure Tunneling Extensible Authentication Protocol Method (EAP-FAST)", Work in Progress, October 2006. [43] Schneier, B., Mudge, and D. Wagner, "Cryptanalysis of Microsoft's PPTP Authentication Extensions (MS-CHAPv2)", CQRE 99, Springer-Verlag LNCS 1740, October 1999. [44] Simpson, W., "The Point-to-Point Protocol (PPP)", STD 51, RFC 1661, July 1994. [45] Simpson, W., "PPP Challenge Handshake Authentication Protocol (CHAP)", RFC 1994, August 1996. [46] Tschofenig, H., Kroeselberg, D., Pashalidis, A., Ohba, Y., and F. Bersani, "EAP IKEv2 Method", Work in Progress, October 2006. [47] Walker, J. and R. Housley, "The EAP Archie Protocol", Work in Progress, June 2003. [48] Wi-Fi Alliance, "Wi-Fi Protected Access, version 2.0", April 2003. [49] Wright, J., "Weaknesses in LEAP Challenge/Response", Defcon 03, August 2003. [50] Eronen, P. and H. Tschofenig, "Pre-Shared Key Ciphersuites for Transport Layer Security (TLS)", RFC 4279, December 2005. Bersani & Tschofenig Experimental [Page 61] RFC 4764 EAP-PSK January 2007 Appendix A. Generation of the PSK from a Password - Discouraged It is formally discouraged to use a password to generate the PSK, since this opens the door to exhaustive search or dictionary attacks, two attacks that would not otherwise be possible. EAP-PSK only provides a 16-byte key strength when a 16-byte PSK is drawn at random from the set of all possible 16-byte strings. However, as people will probably do this anyway, guidance is provided hereafter to generate the PSK from a password. For some hints on how passwords should be selected, please refer to [41]. The technique presented herein is drawn from [33]. It is intended to try to mitigate the risks associated with password usage in cryptography, typically dictionary attacks. If the binary representation of the password is strictly fewer than 16 bytes long (which by the way means that the chosen password is probably weak because it is too short), then it is padded to 16 bytes with zeroes as its high-order bits. If the binary representation of the password is strictly more than 16 bytes long, then it is hashed down to exactly 16 bytes using the Matyas-Meyer-Oseas hash (please refer to [40] for a description of this hash. Using the notation of Figure 9.3 of [40], g is the identity function and E is AES-128 in our construction.) with IV=0x0123456789ABCDEFFEDCBA9876543210 (this value has been arbitrarily selected). We now assume that we have a 16-byte number derived from the initial password (that can be the password itself if its binary representation is exactly 16 bytes long). We shall call this number P16. Following the notations used in [33], the PSK is derived thanks to PBKDF2 instantiated with: o P16 as P o The first 96 bits of the XOR of the peer and server NAIs as Salt (zero-padded in the high-order bits if necessary). o 5000 as c o 16 as dkLen Bersani & Tschofenig Experimental [Page 62] RFC 4764 EAP-PSK January 2007 Although this gives better protection than nothing, this derivation does not stricto sensu protect against dictionary attacks. It only makes dictionary precomputation harder. Authors' Addresses Florent Bersani France Telecom R&D 38, rue du General Leclerc Issy-Les-Moulineaux 92794 Cedex 9 FR EMail: bersani_florent@yahoo.fr Hannes Tschofenig Siemens Networks GmbH & Co KG Otto-Hahn-Ring 6 Munich 81739 GE EMail: Hannes.Tschofenig@siemens.com Bersani & Tschofenig Experimental [Page 63] RFC 4764 EAP-PSK January 2007 Full Copyright Statement Copyright (C) The IETF Trust (2007). This document is subject to the rights, licenses and restrictions contained in BCP 78 and at www.rfc-editor.org/copyright.html, and except as set forth therein, the authors retain all their rights. This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM 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. Intellectual Property The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org. Acknowledgement Funding for the RFC Editor function is currently provided by the Internet Society. Bersani & Tschofenig Experimental [Page 64]