idnits 2.17.1 draft-ietf-karp-ops-model-00.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack an IANA Considerations section. (See Section 2.2 of https://www.ietf.org/id-info/checklist for how to handle the case when there are no actions for IANA.) Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (April 6, 2011) is 4769 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Outdated reference: A later version (-10) exists of draft-ietf-karp-design-guide-02 -- Obsolete informational reference (is this intentional?): RFC 4572 (Obsoleted by RFC 8122) Summary: 1 error (**), 0 flaws (~~), 2 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group S. Hartman 3 Internet-Draft Painless Security 4 Intended status: Informational D. Zhang 5 Expires: October 8, 2011 Huawei 6 April 6, 2011 8 Operations Model for Router Keying 9 draft-ietf-karp-ops-model-00.txt 11 Abstract 13 Developing an operational and management model for routing protocol 14 security that works across protocols will be critical to the success 15 of routing protocol security efforts. This document discusses issues 16 and begins to consider development of these models. 18 Status of this Memo 20 This Internet-Draft is submitted in full conformance with the 21 provisions of BCP 78 and BCP 79. 23 Internet-Drafts are working documents of the Internet Engineering 24 Task Force (IETF). Note that other groups may also distribute 25 working documents as Internet-Drafts. The list of current Internet- 26 Drafts is at http://datatracker.ietf.org/drafts/current/. 28 Internet-Drafts are draft documents valid for a maximum of six months 29 and may be updated, replaced, or obsoleted by other documents at any 30 time. It is inappropriate to use Internet-Drafts as reference 31 material or to cite them other than as "work in progress." 33 This Internet-Draft will expire on October 8, 2011. 35 Copyright Notice 37 Copyright (c) 2011 IETF Trust and the persons identified as the 38 document authors. All rights reserved. 40 This document is subject to BCP 78 and the IETF Trust's Legal 41 Provisions Relating to IETF Documents 42 (http://trustee.ietf.org/license-info) in effect on the date of 43 publication of this document. Please review these documents 44 carefully, as they describe your rights and restrictions with respect 45 to this document. Code Components extracted from this document must 46 include Simplified BSD License text as described in Section 4.e of 47 the Trust Legal Provisions and are provided without warranty as 48 described in the Simplified BSD License. 50 Table of Contents 52 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 53 2. Requirements notation . . . . . . . . . . . . . . . . . . . . 4 54 3. Breakdown of KARP configuration . . . . . . . . . . . . . . . 5 55 3.1. Integrity of the Key Table . . . . . . . . . . . . . . . . 6 56 3.2. Management of Key Table . . . . . . . . . . . . . . . . . 6 57 3.3. Protocol Limitations from the Key Table . . . . . . . . . 7 58 3.4. VRFs . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 59 4. Credentials and Authorization . . . . . . . . . . . . . . . . 8 60 4.1. Preshared Keys . . . . . . . . . . . . . . . . . . . . . . 9 61 4.2. Asymmetric Keys . . . . . . . . . . . . . . . . . . . . . 11 62 4.3. Public Key Infrastructure . . . . . . . . . . . . . . . . 11 63 4.4. The role of Central Servers . . . . . . . . . . . . . . . 12 64 5. Grouping Peers Together . . . . . . . . . . . . . . . . . . . 13 65 6. Administrator Involvement . . . . . . . . . . . . . . . . . . 15 66 6.1. Enrollment . . . . . . . . . . . . . . . . . . . . . . . . 15 67 6.2. Handling Faults . . . . . . . . . . . . . . . . . . . . . 15 68 7. Upgrade Considerations . . . . . . . . . . . . . . . . . . . . 17 69 8. Related Work . . . . . . . . . . . . . . . . . . . . . . . . . 18 70 9. Security Considerations . . . . . . . . . . . . . . . . . . . 19 71 10. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 20 72 11. References . . . . . . . . . . . . . . . . . . . . . . . . . . 21 73 11.1. Normative References . . . . . . . . . . . . . . . . . . . 21 74 11.2. Informative References . . . . . . . . . . . . . . . . . . 21 75 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 22 77 1. Introduction 79 The KARP working group is designing improvements to the cryptographic 80 authentication of IETF routing protocols. These improvements include 81 improvements to how integrity functions are handled within each 82 protocol as well as designing an automated key management solution. 84 This document discusses issues to consider when thinking about the 85 operational and management model for KARP. Each implementation will 86 take its own approach to management; this is one area for vendor 87 differentiation. However, it is desirable to have a common baseline 88 for the management objects allowing administrators, security 89 architects and protocol designers to understand what management 90 capabilities they can depend on in heterogeneous environments. 91 Similarly, designing and deploying the protocol will be easier with 92 thought paid to a common operational model. This will also help with 93 the design of NetConf schemas or MIBs later. 95 2. Requirements notation 97 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 98 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 99 document are to be interpreted as described in [RFC2119]. 101 3. Breakdown of KARP configuration 103 There are multiple ways of structuring configuration information. 104 One factor to consider is the scope of the configuration information. 105 Several protocols are peer-to-peer routing protocols where a 106 different key could potentially be used for each neighbor. Other 107 protocols require the same group key to be used for all nodes in an 108 administrative domain or routing area. In other cases, the same 109 group key needs to be used for all routers on an interface, but 110 different group keys can be used for each interface. 112 Within situations where a per-interface, per-area or per-peer key can 113 be used for manually configured long-term keys, that flexibility may 114 not be desirable from an operational standpoint. For example 115 consider OSPF [RFC2328]. Each OSPF link needs to use the same 116 authentication configuration, including the set of keys used for 117 reception and the set of keys used for transmission, but may use 118 different keys for different links. The most general management 119 model would be to configure keys per link. However for deployments 120 where the area uses the same key it would be strongly desirable to 121 configure the key as a property of the area. If the keys are 122 configured per-link, they can get out of sync. In order to support 123 generality of configuration and common operational situations, it 124 would be desirable to have some sort of inheritance where default 125 configurations are made per-area unless overridden per-interface. 127 As described in [I-D.housley-saag-crypto-key-table], the 128 cryptographic keys are separated from the interface configuration 129 into their own configuration store. This document should specify how 130 key selection interacts with the key table. One possible approach 131 would be to assume that all keys that permit use on a given interface 132 would be used on that interface. This model would need to be 133 expanded in cases where keys are configured per-area or per-domain. 134 It's not clear why "all" is permitted as an interface specification 135 in this model; it seems unlikely that it would be desirable to use 136 the same set of keys for two different instances of an IGP or across 137 autonomous system boundaries. 139 Another model is that the interface specification in the key table is 140 a restriction. Then a set of keys from the key table is attached to 141 an interface, area or routing domain using an additional 142 configuration step. This avoids the previous problems at the expense 143 of significant complexity of configuration. 145 Operational Requirements: KARP MUST support configuration of keys at 146 the most general scope for the underlying protocol; protocols 147 supporting per-peer keys MUST permit configuration of per-peer keys, 148 protocols supporting per-interface keys MUST support configuration of 149 per-interface keys, and so on. KARP MUST NOT permit configuration of 150 an inappropriate key scope. For example, configuration of separate 151 keys per interface MUST NOT be supported for a protocol requiring 152 per-area keys. 154 3.1. Integrity of the Key Table 156 The routing key table [I-D.housley-saag-crypto-key-table] provides a 157 very general mechanism to abstract the storage of keys for routing 158 protocols. To avoid misconfiguration and simplify problem 159 determination, the router MUST verify the internal consistency of 160 entries added to the table. At a minimum, the router MUST verify: 162 o The cryptographic algorithms are valid for the protocol. 164 o The key derivation function is valid for the protocol. 166 o The direction is valid for the protocol; for example protocols 167 that require the same session key be used in both directions MUST 168 have a direction of both. 170 o The peer and interface specification is consistent with the 171 protocol. 173 Other checks are possible. For example the router could verify that 174 if a key is associated with a peer, that peer is a configured peer 175 for the specified protocol. However, this may be undesirable. It 176 may be desirable to load a key table when some peers have not yet 177 been configured. Also, it may be desirable to share portions of a 178 key table across devices even when their current configuration does 179 not require an adjacency with a particular peer in the interest of 180 uniform configuration or preparing for fail-over. 182 3.2. Management of Key Table 184 Several management operations will be quite common. For service 185 provider deployments the configuration management system can simply 186 update the key table. However, for smaller deployments, efficient 187 management operations are important. 189 As part of adding a new key it is typically desirable to set an 190 expiration time for an old key. The management interface SHOULD 191 provide a mechanism to easily update the expiration time for a 192 current key used with a given peer or interface. Also when adding a 193 key it is desirable to push the key out to nodes that will need it, 194 allowing use for receiving packets then later enabling transmit. 195 This can be accomplished automatically by providing a delay between 196 when a key becomes valid for reception and transmission. However, 197 some environments may not be able to predict when all the necessary 198 changes will be made. In these cases having a mechanism to enable a 199 key for sending is desirable. 201 3.3. Protocol Limitations from the Key Table 203 The format of the key table imposes a few limitations on routing 204 protocols. The first is that the key ID is 16 bits; some routing 205 protocols have 32-bit key identifiers. A key mapping table as 206 discussed in 4.1 of [I-D.polk-saag-rtg-auth-keytable] could be used 207 to map to the larger key identifier. However it's probably desirable 208 to either decide that only 16 bits of the key ID space is to be used 209 or to expand the identifier space in the key table. From a 210 management standpoint we need to make concrete requirements around 211 whether a key ID is per-protocol or whether subspaces in the key ID 212 space are reserved for each protocol. This is necessary so that 213 implementations from different vendors can be managed consistently. 215 The second requirement that the key table places is that the key ID 216 is scoped fairly broadly. At least within some protocols such as 217 OSPF, the key ID might only need to be unique per-link or per-peer. 218 That is, packets sent on two different interfaces could use key ID 32 219 even if the keys were different for these interfaces. An 220 implementation could use the interface and the key ID as a lookup to 221 find the right key. However, the key table draft requires that a key 222 ID be sufficient to look up a key, meaning that the key ID is a 223 globally scoped identifier. There is nothing wrong with this 224 restriction, but it does need to be noted when assigning key IDs for 225 a domain. 227 Consideration is required for how an automated key management 228 protocol will assign key IDs for group keys. All members of the 229 group may need to use the same key ID. This requires careful 230 coordination of global key IDs. Interactions with the peer key ID 231 field may make this easier; this requires additional study. 233 3.4. VRFs 235 Many core and enterprise routers support multiple routing instances. 236 For example a router serving multiple VPNs is likely to have a 237 forwarding/routing instance for each of these VPNs. We need to 238 decide how the key table and other configuration information for KARP 239 interacts with this. The obvious first-order answer is that each 240 routing instance gets its own key table. However, we need to 241 consider how these instances interact with each other and confirm 242 this makes sense. 244 4. Credentials and Authorization 246 Several methods for authentication have been proposed for KARP. The 247 simplest is preshared keys used directly as traffic keys. In this 248 mode, the traffic integrity keys are directly configured. This is 249 the mode supported by today's routing protocols. 251 As discussed in [I-D.polk-saag-rtg-auth-keytable], preshared keys can 252 be used as the input to a key derivation function (KDF) to generate 253 traffic keys. For example the TCP Authentication Option (TCP-AO) 254 [RFC5925] derives keys based on the initial TCP session state. 255 Typically a KDF will combine a long-term key with public inputs 256 exchanged as part of the protocol to form fresh session keys. a KDF 257 could potentially be used with some inputs that are configured along 258 with the long-term key. Also, it's possible that inputs to a KDF 259 will be private and exchanged as part of the protocol, although this 260 will be uncommon in KARP's uses of KDFs. 262 Preshared keys could also be used by an automated key management 263 protocol. In this mode, preshared keys would be used for 264 authentication. However traffic keys would be generated by some key 265 agreement mechanism or transported in a key encryption key derived 266 from the preshared key. This mode may provide better replay 267 protection. Also, in the absence of active attackers, key agreement 268 strategies such as Diffie-Hellman can be used to produce high-quality 269 traffic keys even from relatively weak preshared keys. 271 Public keys can be used for authentication. The design guide 272 [I-D.ietf-karp-design-guide] describes a mode in which routers have 273 the hashes of peer routers' public keys. In this mode, a traditional 274 public-key infrastructure is not required. The advantage of this 275 mode is that a router only contains its own keying material, limiting 276 the scope of a compromise. The disadvantage is that when a router is 277 added or deleted from the set of authorized routers, all routers that 278 peer need to be updated. Note that self-signed certificates are a 279 common way of communicating public-keys in this style of 280 authentication. 282 Certificates signed by a certification authority or some other PKI 283 could be used. The advantage of this approach is that routers may 284 not need to be directly updated when peers are added or removed. The 285 disadvantage is that more complexity and cost is required. 287 Each of these approaches has a different set of management and 288 operational requirements. Key differences include how authorization 289 is handled and how identity works. This section discusses these 290 differences. 292 4.1. Preshared Keys 294 In the protocol, manual preshared keys are either unnamed or named by 295 a small integer (typically 16 or 32 bits) key ID. Implementations 296 that support multiple keys for protocols that have no names for keys 297 need to try all possible keys before deciding a packet cannot be 298 validated [RFC4808]. Typically key IDs are names used by one group 299 or peer. 301 Manual preshared keys are often known by a group of peers rather than 302 just one peer. This is an interesting security property: it is 303 impossible to identify the peer sending a message cryptographically; 304 it is only possible to identify a group of peers using cryptographic 305 means. Within the routing threat model the peer sending a message 306 can be identified only because peers are trusted and thus can be 307 assumed to correctly label the packets they send. This contrasts 308 with a protocol where cryptographic means such as digital signatures 309 are used to verify the origin of a message. As a consequence, 310 authorization is typically based on knowing the preshared key rather 311 than on being a particular peer. Note that once an authorization 312 decision is made, the peer can assert its identity; this identity is 313 trusted just as the routing information from the peer is trusted. 314 However, for the process of authorization, it would be more 315 complicated to identify peers this way and would not gain a security 316 benefit in most deployments. 318 Preshared keys used with key derivation function similarly to manual 319 preshared keys. However to form the actual traffic keys, session or 320 peer specific information is combined with the key. From an 321 authorization standpoint, the derivation key works the same as a 322 manual key. An additional routing protocol step or transport step 323 forms the key that is actually used. 325 Preshared keys that are used via automatic key management have not 326 been specified. Their naming and authorization may differ. In 327 particular, such keys may end up being known only by two peers. 328 Alternatively they may also be known by a group of peers. 329 Authorization could potentially be based on peer identity, although 330 it is likely that knowing the right key will be sufficient. There 331 does not appear to be a compelling reason to decouple the 332 authorization of a key for some purpose from authorization of peers 333 holding that key to perform the authorized function. 335 Care needs to be taken when symmetric keys are used for multiple 336 purposes. Consider the implications of using the same preshared key 337 for two interfaces: it becomes impossible to distinguish a router on 338 one interface from a router on another interface. So, a router that 339 is trusted to participate in a routing protocol on one interface 340 becomes implicitly trusted for the other interfaces that share the 341 key. For many cases, such as link-state routers in the same routing 342 area, there is no significant advantage that an attacker could gain 343 from this trust within the KARP threat model. However, distance- 344 vector protocols, such as BGP and RIP, permit routes to be filtered 345 across a trust boundary. For these protocols, participation in one 346 interface might be more advantageous than another. Operationally, 347 when this trust distinction is important to a deployment, different 348 keys need to be used on each side of the trust boundary. Key 349 derivation can help prevent this problem in cases of accidental 350 misconfiguration. However, key derivation cannot protect against a 351 situation where a system was incorrectly trusted to have the key used 352 to perform the derivation. To the extent that there are multiple 353 zones of trust and a routing protocol is determining whether a 354 particular router is within a certain zone, the question of untrusted 355 actors is within the scope of the routing threat model. 357 Key derivation can be part of a management solution to a desire to 358 have multiple keys for different zones of trust. A master key could 359 be combined with peer, link or area identifiers to form a router- 360 specific preshared key that is loaded onto routers. Provider that 361 the master key lives only on the management server and not the 362 individual routers, trust is preserved. However in many cases, 363 generating independent keys for the routers and storing the result is 364 more practical. If the master key were somehow compromised, all the 365 resulting keys would need to be changed. However if independent keys 366 are used, the scope of a compromise may be more limited. 368 More subtle problems with key separation can appear in protocol 369 design. Two protocols that use the same traffic keys may work 370 together in unintended ways permitting one protocol to be used to 371 attack the other. Consider two hypothetical protocols. Protocol A 372 starts its messages with a set of extensions that are ignored if not 373 understood. Protocol B has a fixed header at the beginning of its 374 messages but ends messages with extension information. It may be 375 that the same message is valid both as part of protocol A and 376 protocol B. An attacker may be able to gain an advantage by getting a 377 router to generate this message with one protocol under situations 378 where the other protocol would not generate the message. This 379 hypothetical example is overly simplistic; real-world attacks 380 exploiting key separation weaknesses tend to be complicated and 381 involve specific properties of the cryptographic functions involved. 382 The key point is that whenever the same key is used in multiple 383 protocols, attacks may be possible. All the involved protocols need 384 to be analyzed to understand the scope of potential attacks. 386 Key separation attacks interact with the KARP operational model in a 387 number of ways. Administrators need to be aware of situations where 388 using the same manual traffic key with two different protocols (or 389 the same protocol in different contexts) creates attack 390 opportunities. Design teams should consider how their protocol might 391 interact with other routing protocols and describe any attacks 392 discovered so that administrators can understand the operational 393 implications. When designing automated key management or new 394 cryptographic authentication within routing protocols, we need to be 395 aware that administrators expect to be able to use the same preshared 396 keys in multiple contexts. As a result, we should use appropriate 397 key derivation functions so that different cryptographic keys are 398 used even when the same initial input key is used. 400 4.2. Asymmetric Keys 402 Outside of a PKI, public keys are expected to be known by the hash of 403 a key or (potentially self-signed) certificate. The Session 404 Description Protocol provides a standardized mechanism for naming 405 keys (in that case certificates) based on hashes (section 5 406 [RFC4572]). KARP SHOULD adopt this approach or another approach 407 already standardized within the IETF rather than inventing a new 408 mechanism for naming public keys. 410 A public key is typically expected to belong to one peer. As a peer 411 generates new keys and retires old keys, its public key may change. 412 For this reason, from a management standpoint, peers should be 413 thought of as associated with multiple public keys rather than as 414 containing a single public key hash as an attribute of the peer 415 object. 417 Authorization of public keys could be done either by key hash or by 418 peer identity. Performing authorizations by peer identity should 419 make it easier to update the key of a peer without risk of losing 420 authorizations for that peer. However management interfaces need to 421 be carefully designed to avoid making this extra level of indirection 422 complicated for operators. 424 4.3. Public Key Infrastructure 426 When a PKI is used, certificates are used. The certificate binds a 427 key to a name of a peer. The key management protocol is responsible 428 for exchanging certificates and validating them to a trust anchor. 430 Authorization needs to be done in terms of peer identities not in 431 terms of keys. One reason for this is that when a peer changes its 432 key, the new certificate needs to be sufficient for authentication to 433 continue functioning even though the key has never been seen before. 435 Potentially authorization could be performed in terms of groups of 436 peers rather than single peers. An advantage of this is that it may 437 be possible to add a new router with no authentication related 438 configuration of the peers of that router. For example, a domain 439 could decide that any router with a particular keyPurposeID signed by 440 the organization's certificate authority is permitted to join the 441 IGP. Just as in configurations where cryptographic authentication is 442 not used, automatic discovery of this router can establish 443 appropriate adjacencies. 445 Assuming that potentially self-signed certificates are used by 446 routers that wish to use public keys but that do not need a PKI, then 447 PKI and the infrastructureless mode of public-key operation described 448 in the previous section can work well together. One router could 449 identify its peers based on names and use certificate validation. 450 Another router could use hashes of certificates. This could be very 451 useful for border routers between two organizations. Smaller 452 organizations could use public keys and larger organizations could 453 use PKI. 455 4.4. The role of Central Servers 457 An area to explore is the role of central servers like RADIUS or 458 directories. As discussed in the design-guide, a system where keys 459 are pushed by a central management system is undesirable as an end 460 result for KARP. However central servers may play a role in 461 authorization and key rollover. For example a node could send a hash 462 of a public key to a RADIUS server. 464 If central servers do play a role it will be critical to make sure 465 that they are not required during routine operation or a cold-start 466 of a network. They are more likely to play a role in enrollment of 467 new peers or key migration/compromise. 469 Another area where central servers may play a role is for group key 470 agreement. As an example, [I-D.liu-ospfv3-automated-keying-req] 471 discusses the potential need for key agreement servers in OSPF. 472 Other routing protocols that use multicast or broadcast such as IS-IS 473 are likely to need a similar approach. 475 5. Grouping Peers Together 477 One significant management consideration will be the grouping of 478 management objects necessary to determine who is authorized to act as 479 a peer for a given routing action. As discussed previously, the 480 following objects are potentially required: 482 o Key objects are required. Symmetric keys may be preshared. 483 Asymmetric public keys may be used directly for authorization as 484 well. During key transitions more than one key may refer to a 485 given peer. Group preshared keys may refer to multiple peers. 487 o A peer is a router that this router might wish to communicate 488 with. Peers may be identified by names or keys. 490 o Groups of peers may be authorized for a given routing protocol. 492 Establishing a management model is difficult because of the complex 493 relationships between each set of objects. As discussed there may be 494 more than one key for a peer. However in the preshared key case, 495 there may be more than one peer for a key. This is true both for 496 group security association protocols such as an IGP or one-to-one 497 protocols where the same key is used administratively. In some of 498 these situations, it may be undesirable to explicitly enumerate the 499 peers in the configuration; for example IGP peers are auto-discovered 500 for broadcast links but not for non-broadcast multi-access links. 502 Peers may be identified either by name or key. If peers are 503 identified by key it is probably strongly desirable from an 504 operational standpoint to consider any peer identifiers or name to be 505 a local matter and not require the names or identifiers to be 506 synchronized. Obviously if peers are identified by names (for 507 example with certificates in a PKI), identifiers need to be 508 synchronized between the authorized peer and the peer making the 509 authorization decision. 511 In many cases peers will explicitly be identified. In these cases it 512 is possible to attach the authorization information (keys or 513 identifiers) to the peer's configuration object. Two cases do not 514 involve enumerating peers. The first is the case where preshared 515 keys are shared among a group of peers. It is likely that this case 516 can be treated from a management standpoint as a single peer 517 representing all the peers that share the keys. The other case is 518 one where certificates in a PKI are used to introduce peers to a 519 router. In this case, rather than configuring peers, , the router 520 needs to be configured with information on what certificates 521 represent acceptable peers. 523 Another consideration is what routing protocols share peers. For 524 example it may be common for LDP peers to also be peers of some other 525 routing protocol. Also, RSVP-TE may be associated with some TE-based 526 IGP. In some of these cases it would be desirable to use the same 527 authorization information for both routing protocols. 529 In order to develop a management model for authorization, the working 530 group needs to consider several questions. What protocols support 531 auto-discovery of peers? What protocols require more configuration 532 of a peer than simply the peer's authorization information and 533 network address? What management operations are going to be common 534 as security information for peers is configured and updated? What 535 operations will be common while performing key transitions or while 536 migrating to new security technologies? 538 6. Administrator Involvement 540 One key operational question is what areas will administrator 541 involvement be required. Likely areas where involvement may be 542 useful includes enrollment of new peers. Fault recovery should also 543 be considered. 545 6.1. Enrollment 547 One area where the management of routing security needs to be 548 optimized is the deployment of a new router. In some cases a new 549 router may be deployed on an existing network where routing to 550 management servers is already available. In other cases, routers may 551 be deployed as part of connecting or creating a site. Here, the 552 router and infrastructure may not be available until the router has 553 securely authenticated. This problem is similar to the problem of 554 getting initial configuration of routing instances onto the router. 555 However, especially in cases where asymmetric keys or per-peer 556 preshared keys are used, the configuration of other routers needs to 557 be modified to bring up the security association. Also, there has 558 been discussion of generating keys on routers and not allowing them 559 to leave devices. This also impacts what strategies are possible. 560 For example this might mean that routers need to be booted in a 561 secure environment where keys can be generated, and public keys 562 copied to a management server to push out the new public key to 563 potential peers. Then, the router needs to be packaged, moved to 564 where it will be deployed and set up.Alternatives are possible; it is 565 critical that we understand how what we propose impacts operators. 567 We need to work through examples with operators familiar with 568 specific real-world deployment practices and understand how proposed 569 security mechanisms will interact with these practices. 571 6.2. Handling Faults 573 Faults may interact with operational practice in at least two ways. 574 First, security solutions may introduce faults. For example if 575 certificates expire in a PKI, previous adjacencies may no longer 576 form. Operational practice will require a way of repairing these 577 errors. This may end up being very similar to deploying a router 578 that is connecting a new site as the security fault may have 579 partitioned the network. However, unlike a new deployment, the event 580 is unplanned. Strategies such as configuring a router and shipping 581 it to a site may not be appropriate for recovering a fault even 582 though they may be more useful for new deployments. 584 Monitoring will play a critical role in avoiding security faults such 585 as certificate expiration. However, the protocols MUST still have 586 adequate operational mechanisms to recover from these situations. 587 Also, some faults, such as those resulting from a compromise or 588 actual attack on a facility are inherent and may not be prevented. 590 A second class of faults is equipment faults that impact security. 591 For example if keys are stored on a router and never moved from that 592 device, failure of a router implies a need to update security 593 provisioning on the replacement router and its peers. 595 To address these operational considerations, we should identify 596 circumstances surrounding recovery from today's faults and understand 597 how protocols will impact mechanisms used today. 599 7. Upgrade Considerations 601 It needs to be possible to deploy automated key management in an 602 organization without either having to disable existing security or 603 disrupting routing. As a result, it needs to be possible to perform 604 a phased upgrade from manual keying to automated key management. 606 For peer-to-peer protocols such as BGP, this is likely to be 607 relatively easy. First, code that supports automated key management 608 needs to be loaded on both peers. Then the adjacency can be 609 upgraded. The configuration can be updated to switch to automated 610 key management when the second router reboots. 612 The situation is more complicated for multicast protocols. It's 613 probably not reasonable to bring down an entire link to reconfigure 614 it as using automated key management. Two approaches should be 615 considered. One is to support key table rows from the automated key 616 management and manually configured for the same link at the same 617 time. Coordinating this may be tricky. Another possibility is for 618 the automated key management protocol to actually select the same 619 traffic key that is being used manually 621 8. Related Work 623 Discuss draft-housley-saag-*, draft-polk-saag-*, the discussions in 624 the KARP framework, etc. 626 9. Security Considerations 628 This document does not define a protocol. It does discuss the 629 operational and management implications of several security 630 technologies. 632 10. Acknowledgments 634 Funding for Sam Hartman's work on this memo is provided by Huawei. 636 The authors would like to thank Gregory Lebovitz, Russ White and Bill 637 Atwood for valuable reviews. 639 11. References 641 11.1. Normative References 643 [I-D.housley-saag-crypto-key-table] 644 Housley, R. and T. Polk, "Database of Long-Lived Symmetric 645 Cryptographic Keys", 646 draft-housley-saag-crypto-key-table-04 (work in progress), 647 October 2010. 649 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 650 Requirement Levels", BCP 14, RFC 2119, March 1997. 652 11.2. Informative References 654 [I-D.ietf-karp-design-guide] 655 Lebovitz, G. and M. Bhatia, "Keying and Authentication for 656 Routing Protocols (KARP) Design Guidelines", 657 draft-ietf-karp-design-guide-02 (work in progress), 658 March 2011. 660 [I-D.liu-ospfv3-automated-keying-req] 661 Liu, Y., "OSPFv3 Automated Group Keying Requirements", 662 draft-liu-ospfv3-automated-keying-req-01 (work in 663 progress), July 2007. 665 [I-D.polk-saag-rtg-auth-keytable] 666 Polk, T. and R. Housley, "Routing Authentication Using A 667 Database of Long-Lived Cryptographic Keys", 668 draft-polk-saag-rtg-auth-keytable-05 (work in progress), 669 November 2010. 671 [RFC2328] Moy, J., "OSPF Version 2", STD 54, RFC 2328, April 1998. 673 [RFC4572] Lennox, J., "Connection-Oriented Media Transport over the 674 Transport Layer Security (TLS) Protocol in the Session 675 Description Protocol (SDP)", RFC 4572, July 2006. 677 [RFC4808] Bellovin, S., "Key Change Strategies for TCP-MD5", 678 RFC 4808, March 2007. 680 [RFC5925] Touch, J., Mankin, A., and R. Bonica, "The TCP 681 Authentication Option", RFC 5925, June 2010. 683 Authors' Addresses 685 Sam Hartman 686 Painless Security 688 Email: hartmans-ietf@mit.edu 690 Dacheng Zhang 691 Huawei 693 Email: zhangdacheng@huawei.com