idnits 2.17.1 draft-ietf-6man-stable-privacy-addresses-10.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 : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (June 12, 2013) is 3971 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) ** Obsolete normative reference: RFC 2460 (Obsoleted by RFC 8200) ** Downref: Normative reference to an Informational RFC: RFC 3493 ** Downref: Normative reference to an Informational RFC: RFC 3542 ** Obsolete normative reference: RFC 4941 (Obsoleted by RFC 8981) ** Downref: Normative reference to an Informational RFC: RFC 6105 -- Obsolete informational reference (is this intentional?): RFC 1948 (Obsoleted by RFC 6528) == Outdated reference: A later version (-08) exists of draft-ietf-opsec-ipv6-host-scanning-01 Summary: 5 errors (**), 0 flaws (~~), 2 warnings (==), 2 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 IPv6 maintenance Working Group (6man) F. Gont 3 Internet-Draft SI6 Networks / UTN-FRH 4 Intended status: Standards Track June 12, 2013 5 Expires: December 14, 2013 7 A method for Generating Stable Privacy-Enhanced Addresses with IPv6 8 Stateless Address Autoconfiguration (SLAAC) 9 draft-ietf-6man-stable-privacy-addresses-10 11 Abstract 13 This document specifies a method for generating IPv6 Interface 14 Identifiers to be used with IPv6 Stateless Address Autoconfiguration 15 (SLAAC), such that addresses configured using this method are stable 16 within each subnet, but the Interface Identifier changes when hosts 17 move from one network to another. This method is meant to be an 18 alternative to generating Interface Identifiers based on hardware 19 address (e.g., using IEEE identifiers), such that the benefits of 20 stable addresses can be achieved without sacrificing the privacy of 21 users. The method specified in this document applies to all prefixes 22 a host may be employing, including link-local, global, and unique- 23 local addresses. 25 Status of this Memo 27 This Internet-Draft is submitted in full conformance with the 28 provisions of BCP 78 and BCP 79. 30 Internet-Drafts are working documents of the Internet Engineering 31 Task Force (IETF). Note that other groups may also distribute 32 working documents as Internet-Drafts. The list of current Internet- 33 Drafts is at http://datatracker.ietf.org/drafts/current/. 35 Internet-Drafts are draft documents valid for a maximum of six months 36 and may be updated, replaced, or obsoleted by other documents at any 37 time. It is inappropriate to use Internet-Drafts as reference 38 material or to cite them other than as "work in progress." 40 This Internet-Draft will expire on December 14, 2013. 42 Copyright Notice 44 Copyright (c) 2013 IETF Trust and the persons identified as the 45 document authors. All rights reserved. 47 This document is subject to BCP 78 and the IETF Trust's Legal 48 Provisions Relating to IETF Documents 49 (http://trustee.ietf.org/license-info) in effect on the date of 50 publication of this document. Please review these documents 51 carefully, as they describe your rights and restrictions with respect 52 to this document. Code Components extracted from this document must 53 include Simplified BSD License text as described in Section 4.e of 54 the Trust Legal Provisions and are provided without warranty as 55 described in the Simplified BSD License. 57 Table of Contents 59 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 60 2. Design goals . . . . . . . . . . . . . . . . . . . . . . . . . 7 61 3. Algorithm specification . . . . . . . . . . . . . . . . . . . 8 62 4. Resolving Duplicate Address Detection (DAD) conflicts . . . . 13 63 5. Specified Constants . . . . . . . . . . . . . . . . . . . . . 14 64 6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 15 65 7. Security Considerations . . . . . . . . . . . . . . . . . . . 16 66 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 18 67 9. References . . . . . . . . . . . . . . . . . . . . . . . . . . 19 68 9.1. Normative References . . . . . . . . . . . . . . . . . . . 19 69 9.2. Informative References . . . . . . . . . . . . . . . . . . 20 70 Appendix A. Possible sources for the Net_Iface parameter . . . . 22 71 A.1. Interface Index . . . . . . . . . . . . . . . . . . . . . 22 72 A.2. Interface Name . . . . . . . . . . . . . . . . . . . . . . 22 73 A.3. Link-layer Addresses . . . . . . . . . . . . . . . . . . . 22 74 A.4. Logical Network Service Identity . . . . . . . . . . . . . 23 75 Appendix B. Security/privacy issues with traditional SLAAC 76 addresses . . . . . . . . . . . . . . . . . . . . . . 24 77 B.1. Correlation of node activities within the same network . . 24 78 B.2. Correlation of node activities across networks (host 79 tracking) . . . . . . . . . . . . . . . . . . . . . . . . 24 80 B.3. Address-scanning attacks . . . . . . . . . . . . . . . . . 25 81 B.4. Exploitation of device-specific information . . . . . . . 25 82 Appendix C. Privacy issues still present when temporary 83 addresses are employed . . . . . . . . . . . . . . . 26 84 C.1. Host tracking . . . . . . . . . . . . . . . . . . . . . . 26 85 C.1.1. Tracking hosts across networks #1 . . . . . . . . . . 26 86 C.1.2. Tracking hosts across networks #2 . . . . . . . . . . 27 87 C.1.3. Revealing the identity of devices performing 88 server-like functions . . . . . . . . . . . . . . . . 27 89 C.2. Address-scanning attacks . . . . . . . . . . . . . . . . . 27 90 C.3. Information Leakage . . . . . . . . . . . . . . . . . . . 28 91 C.4. Correlation of node activities within a network . . . . . 28 92 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 29 94 1. Introduction 96 [RFC4862] specifies Stateless Address Autoconfiguration (SLAAC) for 97 IPv6 [RFC2460], which typically results in hosts configuring one or 98 more "stable" addresses composed of a network prefix advertised by a 99 local router, and an Interface Identifier (IID) that typically embeds 100 a hardware address (e.g., using IEEE identifiers) [RFC4291]. 102 Cryptographically Generated Addresses (CGAs) [RFC3972] are yet 103 another method for generating Interface Identifiers, which bind a 104 public signature key to an IPv6 address in the SEcure Neighbor 105 Discovery (SEND) [RFC3971] protocol. 107 Generally, the traditional SLAAC addresses are thought to simplify 108 network management, since they simplify Access Control Lists (ACLs) 109 and logging. However, they have a number of drawbacks: 111 o since the resulting Interface Identifiers do not vary over time, 112 they allow correlation of node activities within the same network, 113 thus negatively affecting the privacy of users. 115 o since the resulting Interface Identifiers are constant across 116 networks, the resulting IPv6 addresses can be leveraged to track 117 and correlate the activity of a node across multiple networks 118 (e.g. track and correlate the activities of a typical client 119 connecting to the public Internet from different locations), thus 120 negatively affecting the privacy of users. 122 o since embedding the underlying link-layer address in the Interface 123 Identifier will result in specific address patterns, such patterns 124 may be leveraged by attackers to reduce the search space when 125 performing address scanning attacks. For example, the IPv6 126 addresses of all nodes manufactured by the same vendor (at a given 127 time frame) will likely contain the same IEEE Organizationally 128 Unique Identifier (OUI) in the Interface Identifier. 130 o embedding the underlying link-layer address in the Interface 131 Identifier leaks device-specific information that could be 132 leveraged to launch device-specific attacks. 134 o embedding the underlying link-layer address in the Interface 135 Identifier means that replacement of the underlying interface 136 hardware will result in a change of the IPv6 address(es) assigned 137 to that interface. 139 Appendix B provides additional details regarding how these 140 vulnerabilities could be exploited, and the extent to which the 141 method discussed in this document mitigates them. 143 The "Privacy Extensions for Stateless Address Autoconfiguration in 144 IPv6" [RFC4941] (henceforth referred to as "temporary addresses") 145 were introduced to complicate the task of eavesdroppers and other 146 information collectors (e.g. IPv6 addresses in web server logs or 147 email headers, etc.) to correlate the activities of a node, and 148 basically result in temporary (and random) Interface Identifiers. 149 These temporary addresses are generated in addition to the 150 traditional IPv6 addresses based on IEEE identifiers, with the 151 "temporary addresses" being employed for "outgoing communications", 152 and the traditional SLAAC addresses being employed for "server" 153 functions (i.e., receiving incoming connections). 155 It should be noted that temporary addresses can be challenging in 156 a number of areas. For example, from a network-management point 157 of view, they tend to increase the complexity of event logging, 158 trouble-shooting, enforcement of access controls and quality of 159 service, etc. As a result, some organizations disable the use of 160 temporary addresses even at the expense of reduced privacy 161 [Broersma]. Temporary addresses may also result in increased 162 implementation complexity, which might not be possible or 163 desirable in some implementations (e.g., some embedded devices). 165 In scenarios in which temporary addresses are deliberately not 166 used (possibly for any of the aforementioned reasons), all a host 167 is left with is the stable addresses that have been generated 168 using e.g. IEEE identifiers. In such scenarios, it may still be 169 desirable to have addresses that mitigate address scanning 170 attacks, and that at the very least do not reveal the node's 171 identity when roaming from one network to another -- without 172 complicating the operation of the corresponding networks. 174 However, even with "temporary addresses" in place, a number of issues 175 remain to be mitigated. Namely, 177 o since "temporary addresses" [RFC4941] do not eliminate the use of 178 fixed identifiers for server-like functions, they only partially 179 mitigate host-tracking and activity correlation across networks 180 (see Appendix C.1 for some example attacks that are still possible 181 with temporary addresses). 183 o since "temporary addresses" [RFC4941] do not replace the 184 traditional SLAAC addresses, an attacker can still leverage 185 patterns in those addresses to greatly reduce the search space for 186 "alive" nodes [Gont-DEEPSEC2011] [CPNI-IPv6] 187 [I-D.ietf-opsec-ipv6-host-scanning]. 189 Hence, there is a motivation to improve the properties of "stable" 190 addresses regardless of whether temporary addresses are employed or 191 not. 193 We note that attackers can employ a plethora of probing techniques 194 [I-D.ietf-opsec-ipv6-host-scanning] to exploit the aforementioned 195 issues. Some of them (such as the use of ICMPv6 Echo Request and 196 ICMPv6 Echo Response packets) could mitigated by a personal firewall 197 at the target host. For other vectors, such listening to ICMPv6 198 "Destination Unreachable, Address Unreachable" (Type 1, Code 3) error 199 messages referring to the target addresses 200 [I-D.ietf-opsec-ipv6-host-scanning], there is nothing a host can do 201 (e.g., a personal firewall at the target host would not be able to 202 mitigate this probing technique). 204 This document specifies a method to generate Interface Identifiers 205 that are stable/constant for each network interface within each 206 subnet, but that change as hosts move from one network to another, 207 thus keeping the "stability" properties of the Interface Identifiers 208 specified in [RFC4291], while still mitigating address-scanning 209 attacks and preventing correlation of the activities of a node as it 210 moves from one network to another. 212 The method specified in this document is a orthogonal to the use of 213 "temporary" addresses [RFC4941], since it is meant to improve the 214 security and privacy properties of the stable addresses that are 215 employed along with the aforementioned "temporary" addresses. In 216 scenarios in which "temporary addresses" are employed, implementation 217 of the mechanism described in this document (in replacement of stable 218 addresses based on e.g. IEEE identifiers) would mitigate address- 219 scanning attacks and also mitigate the remaining vectors for 220 correlating host activities based on the node's constant (i.e. stable 221 across networks) Interface Identifiers. On the other hand, for nodes 222 that currently disable "temporary addresses" [RFC4941] for some of 223 the reasons described earlier in this document, implementation of 224 this mechanism will result in stable privacy-enhanced addresses which 225 address some of the concerns related to addresses that embed IEEE 226 identifiers [RFC4291], and which mitigate IPv6 address-scanning 227 attacks. 229 We note that this method is incrementally deployable, since it does 230 not pose any interoperability implications when deployed on networks 231 where other nodes do not implement or employ it. Additionally, we 232 note that this document does not update or modify IPv6 StateLess 233 Address Auto-Configuration (SLAAC) [RFC4862] itself, but rather only 234 specifies an alternative algorithm to generate Interface Identifiers. 235 Therefore, the usual address lifetime properties (as specified in the 236 corresponding Prefix Information Options) apply when IPv6 addresses 237 are generated as a result of employing the algorithm specified in 238 this document with SLAAC [RFC4862]. Additionally, from the point of 239 view of renumbering, we note that these addresses behave like the 240 traditional IPv6 addresses (that embed a hardware address) resulting 241 from SLAAC [RFC4862]. 243 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 244 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 245 document are to be interpreted as described in RFC 2119 [RFC2119]. 247 2. Design goals 249 This document specifies a method for selecting Interface Identifiers 250 to be used with IPv6 SLAAC, with the following goals: 252 o The resulting Interface Identifiers remain constant/stable for 253 each prefix used with SLAAC within each subnet. That is, the 254 algorithm generates the same Interface Identifier when configuring 255 an address (for the same interface) belonging to the same prefix 256 within the same subnet. 258 o The resulting Interface Identifiers do change when addresses are 259 configured for different prefixes. That is, if different 260 autoconfiguration prefixes are used to configure addresses for the 261 same network interface card, the resulting Interface Identifiers 262 must be (statistically) different. This means that, given two 263 addresses produced by the method specified in this document, it 264 must be difficult for an attacker tell whether the addresses have 265 been generated/used by the same node. 267 o It must be difficult for an outsider to predict the Interface 268 Identifiers that will be generated by the algorithm, even with 269 knowledge of the Interface Identifiers generated for configuring 270 other addresses. 272 o Depending on the specific implementation approach (see Section 3 273 and Appendix A), the resulting Interface Identifiers may be 274 independent of the underlying hardware (e.g. link-layer address). 275 This means that e.g. replacing a Network Interface Card (NIC) will 276 not have the (generally undesirable) effect of changing the IPv6 277 addresses used for that network interface. 279 o The method specified in this document is meant to be an 280 alternative to producing IPv6 addresses based on e.g. IEEE 281 identifiers (as specified in [RFC2464]). It is meant to be 282 employed for all of the stable (i.e. non-temporary) IPv6 addresses 283 configured with SLAAC for a given interface, including global, 284 link-local, and unique-local IPv6 addresses. 286 We note that of use of the algorithm specified in this document is 287 (to a large extent) orthogonal to the use of "temporary addresses" 288 [RFC4941]. When employed along with "temporary addresses", the 289 method specified in this document will mitigate address-scanning 290 attacks and correlation of node activities across networks (see 291 Appendix C and [IAB-PRIVACY]). On the other hand, hosts that do not 292 implement/use "temporary addresses" but employ the method specified 293 in this document would, at the very least, mitigate the host-tracking 294 and address scanning issues discussed in the previous section. 296 3. Algorithm specification 298 IPv6 implementations conforming to this specification MUST generate 299 Interface Identifiers using the algorithm specified in this section 300 in replacement of any other algorithms used for generating "stable" 301 addresses (such as those specified in [RFC2464]). However, 302 implementations conforming to this specification MAY employ the 303 algorithm specified in [RFC4941] to generate temporary addresses in 304 addition to the addresses generated with the algorithm specified in 305 this document. The method specified in this document MUST be 306 employed for generating the Interface Identifiers for all the stable 307 addresses of a given interface, including IPv6 global, link-local, 308 and unique-local addresses. 310 This means that this document does not formally obsolete or 311 deprecate any of the existing algorithms to generate Interface 312 Identifiers (e.g. such as that specified in [RFC2464]). However, 313 those IPv6 implementations that employ this specification MUST 314 generate all of their "stable" addresses as specified in this 315 document. 317 Implementations conforming to this specification SHOULD provide the 318 means for a system administrator to enable or disable the use of this 319 algorithm for generating Interface Identifiers. 321 Unless otherwise noted, all of the parameters included in the 322 expression below MUST be included when generating an Interface 323 Identifier. 325 1. Compute a random (but stable) identifier with the expression: 327 RID = F(Prefix, Net_Iface, Network_ID, DAD_Counter, secret_key) 329 Where: 331 RID: 332 Random (but stable) Interface Identifier 334 F(): 335 A pseudorandom function (PRF) that is not computable from the 336 outside (without knowledge of the secret key), which should 337 produce an output of at least 64 bits.The PRF could be 338 implemented as a cryptographic hash of the concatenation of 339 each of the function parameters. 341 Prefix: 342 The prefix to be used for SLAAC, as learned from an ICMPv6 343 Router Advertisement message, or the link-local IPv6 unicast 344 prefix. 346 Net_Iface: 347 An implementation-dependent stable identifier associated with 348 the network interface for which the RID is being generated. 349 An implementation MAY provide a configuration option to select 350 the source of the identifier to be used for the Net_Iface 351 parameter. A discussion of possible sources for this value 352 (along with the corresponding trade-offs) can be found in 353 Appendix A. 355 Network_ID: 356 Some network specific data that identifies the subnet to which 357 this interface is attached. For example the IEEE 802.11 358 Service Set Identifier (SSID) corresponding to the network to 359 which this interface is associated. This parameter is 360 OPTIONAL. 362 DAD_Counter: 363 A counter that is employed to resolve Duplicate Address 364 Detection (DAD) conflicts. It MUST be initialized to 0, and 365 incremented by 1 for each new tentative address that is 366 configured as a result of a DAD conflict. Implementations 367 that record DAD_Counter in non-volatile memory for each 368 {Prefix, Net_Iface, Network_ID} tuple MUST initialize 369 DAD_Counter to the recorded value if such an entry exists in 370 non-volatile memory). See Section 4 for additional details. 372 secret_key: 373 A secret key that is not known by the attacker. The secret 374 key MUST be initialized at operating system installation time 375 to a pseudo-random number (see [RFC4086] for randomness 376 requirements for security). An implementation MAY provide the 377 means for the the system administrator to change or display 378 the secret key. 380 2. The Interface Identifier is finally obtained by taking as many 381 bits from the RID value (computed in the previous step) as 382 necessary, starting from the least significant bit. 384 We note that [RFC4291] requires that, the Interface IDs of all 385 unicast addresses (except those that start with the binary 386 value 000) be 64-bit long. However, the method discussed in 387 this document could be employed for generating Interface IDs 388 of any arbitrary length, albeit at the expense of reduced 389 entropy (when employing Interface IDs smaller than 64 bits). 391 The resulting Interface Identifier should be compared against the 392 Subnet-Router Anycast [RFC4291] and the Reserved Subnet Anycast 393 Addresses [RFC2526], and against those Interface Identifiers 394 already employed in an address of the same network interface and 395 the same network prefix. In the event that an unacceptable 396 identifier has been generated, this situation should be handled 397 in the same way as the case of duplicate addresses (see 398 Section 4). 400 This document does not require the use of any specific PRF for the 401 function F() above, since the choice of such PRF is usually a trade- 402 off between a number of properties (processing requirements, ease of 403 implementation, possible intellectual property rights, etc.), and 404 since the best possible choice for F() might be different for 405 different types of devices (e.g. embedded systems vs. regular 406 servers) and might possibly change over time. 408 Note that the result of F() in the algorithm above is no more secure 409 than the secret key. If an attacker is aware of the PRF that is 410 being used by the victim (which we should expect), and the attacker 411 can obtain enough material (i.e. addresses configured by the victim), 412 the attacker may simply search the entire secret-key space to find 413 matches. To protect against this, the secret key should be of a 414 reasonable length. Key lengths of at least 128 bits should be 415 adequate. The secret key is initialized at system installation time 416 to a pseudo-random number, thus allowing this mechanism to be 417 enabled/used automatically, without user intervention. 419 Including the SLAAC prefix in the PRF computation causes the 420 Interface Identifier to vary across each prefix (link-local, global, 421 etc.) employed by the node and, as consequently, also across 422 networks. This mitigates the correlation of activities of multi- 423 homed nodes (since each of the corresponding addresses will employ a 424 different Interface ID), host-tracking (since the network prefix will 425 change as the node moves from one network to another), and any other 426 attacks that benefit from predictable Interface Identifiers (such as 427 address scanning attacks). 429 The Net_Iface is a value that identifies the network interface for 430 which an IPv6 address is being generated. The following properties 431 are required for the Net_Iface parameter: 433 o it MUST be constant across system bootstrap sequences and other 434 network events (e.g., bringing another interface up or down) 436 o it MUST be different for each network interface simultaneously in 437 use 439 Since the stability of the addresses generated with this method 440 relies on the stability of all arguments of F(), it is key that the 441 Net_Iface be constant across system bootstrap sequences and other 442 network events. Additionally, the Net_Iface must uniquely identify 443 an interface within the node, such that two interfaces connecting to 444 the same network do not result in duplicate addresses. Different 445 types of operating systems might benefit from different stability 446 properties of the Net_Iface parameter. For example, a client- 447 oriented operating system might want to employ Net_Iface identifiers 448 that are attached to the underlying network interface card (NIC), 449 such that a removable NIC always gets the same IPv6 address, 450 irrespective of the system communications port to which it is 451 attached. On the other hand, a server-oriented operating system 452 might prefer Net_Iface identifiers that are attached to system slots/ 453 ports, such that replacement of a network interface card does not 454 result in an IPv6 address change. Appendix A discusses possible 455 sources for the Net_Iface, along with their pros and cons. 457 Including the optional Network_ID parameter when computing the RID 458 value above would cause the algorithm to produce a different 459 Interface Identifier when connecting to different networks, even when 460 configuring addresses belonging to the same prefix. This means that 461 a host would employ a different Interface Identifier as it moves from 462 one network to another even for IPv6 link-local addresses or Unique 463 Local Addresses (ULAs). In those scenarios where the Network_ID is 464 unknown to the attacker, including this parameter might help mitigate 465 attacks where a victim node connects to the same subnet as the 466 attacker, and the attacker tries to learn the Interface Identifier 467 used by the victim node for a remote network (see Section 7 for 468 further details). 470 The DAD_Counter parameter provides the means to intentionally cause 471 this algorithm produce a different IPv6 addresses (all other 472 parameters being the same). This could be necessary to resolve 473 Duplicate Address Detection (DAD) conflicts, as discussed in detail 474 in Section 4. 476 Finally, we note that all of the bits in the resulting Interface IDs 477 are treated as "opaque" bits. For example, the universal/local bit 478 of Modified EUI-64 format identifiers is treated as any other bit of 479 such identifier. In theory, this might result in Duplicate Address 480 Detection (DAD) failures that would otherwise not be encountered. 481 However, this is not deemed as a real issue, because of the following 482 considerations: 484 o The interface IDs of all addresses (except those of addresses that 485 that start with the binary value 000) are 64-bit long. Since the 486 method specified in this document results in random Interface IDs, 487 the probability of DAD failures is very small. 489 o Real world data indicates that MAC address reuse is far more 490 common than assumed [HDMoore]. This means that even IPv6 491 addresses that employ (allegedly) unique identifiers (such as IEEE 492 identifiers) might result in DAD failures, and hence 493 implementations should be prepared to gracefully handle such 494 occurrences. 496 o Since some popular and widely-deployed operating systems (such as 497 Microsoft Windows) do not employ unique hardware identifiers for 498 the Interface IDs of their stable addresses, reliance on such 499 unique identifiers is more reduced in the deployed world (fewer 500 deployed systems rely on them for the avoidance of address 501 collisions). 503 4. Resolving Duplicate Address Detection (DAD) conflicts 505 If as a result of performing Duplicate Address Detection (DAD) 506 [RFC4862] a host finds that the tentative address generated with the 507 algorithm specified in Section 3 is a duplicate address, it SHOULD 508 resolve the address conflict by trying a new tentative address as 509 follows: 511 o DAD_Counter is incremented by 1. 513 o A new Interface Identifier is generated with the algorithm 514 specified in Section 3, using the incremented DAD_Counter value. 516 This procedure may be repeated a number of times until the address 517 conflict is resolved. Hosts SHOULD try at least IDGEN_RETRIES (see 518 Section 5) tentative addresses if DAD fails for successive generated 519 addresses, in the hopes of resolving the address conflict. We also 520 note that hosts MUST limit the number of tentative addresses that are 521 tried (rather than indefinitely try a new tentative address until the 522 conflict is resolved). 524 In those (unlikely) scenarios in which duplicate addresses are 525 detected and in which the order in which the conflicting nodes 526 configure their addresses may vary (e.g., because they may be 527 bootstrapped in different order), the algorithm specified in this 528 section for resolving DAD conflicts could lead to addresses that are 529 not stable within the same subnet. In order to mitigate this 530 potential problem, nodes MAY record the DAD_Counter value employed 531 for a specific {Prefix, Net_Iface, Network_ID} tuple in non-volatile 532 memory, such that the same DAD_Counter value is employed when 533 configuring an address for the same Prefix and subnet at any other 534 point in time. 536 In the event that a DAD conflict cannot be solved (possibly after 537 trying a number of different addresses), address configuration would 538 fail. In those scenarios, nodes MUST NOT automatically fall back to 539 employing other algorithms for generating Interface Identifiers. 541 5. Specified Constants 543 This document specifies the following constant: 545 IDGEN_RETRIES: 546 defaults to 3. 548 6. IANA Considerations 550 There are no IANA registries within this document. The RFC-Editor 551 can remove this section before publication of this document as an 552 RFC. 554 7. Security Considerations 556 This document specifies an algorithm for generating Interface 557 Identifiers to be used with IPv6 Stateless Address Autoconfiguration 558 (SLAAC), as an alternative to e.g. Interface Identifiers that embed 559 IEEE identifiers (such as those specified in [RFC2464]). When 560 compared to such identifiers, the identifiers specified in this 561 document have a number of advantages: 563 o They prevent trivial host-tracking, since when a host moves from 564 one network to another the network prefix used for 565 autoconfiguration and/or the Network ID (e.g., IEEE 802.11 SSID) 566 will typically change, and hence the resulting Interface 567 Identifier will also change (see Appendix C.1). 569 o They mitigate address-scanning techniques which leverage 570 predictable Interface Identifiers (e.g., known Organizationally 571 Unique Identifiers) [I-D.ietf-opsec-ipv6-host-scanning]. 573 o They may result in IPv6 addresses that are independent of the 574 underlying hardware (i.e. the resulting IPv6 addresses do not 575 change if a network interface card is replaced) if an appropriate 576 source for Net_Iface (Section 3) is employed. 578 o They prevent the information leakage produced by embedding 579 hardware addresses in the Interface Identifier (which could be 580 exploited to launch device-specific attacks). 582 o Since the method specified in this document will result in 583 different Interface Identifiers for each configured address, 584 knowledge/leakage of the Interface Identifier employed for one 585 stable address of will not negatively affect the security/privacy 586 of other stable addresses configured for other prefixes (whether 587 at the same time or at some other point in time). 589 In scenarios in which an attacker can connect to the same subnet as a 590 victim node, the attacker might be able to learn the Interface 591 Identifier employed by the victim node for an arbitrary prefix, by 592 simply sending a forged Router Advertisement [RFC4861] for that 593 prefix, and subsequently learning the corresponding address 594 configured by the victim node (either listening to the Duplicate 595 Address Detection packets, or to any other traffic that employs the 596 newly configured address). We note that a number of factors might 597 limit the ability of an attacker to successfully perform such an 598 attack: 600 o First-Hop security mechanisms such as RA-Guard [RFC6105] 601 [I-D.ietf-v6ops-ra-guard-implementation] could prevent the forged 602 Router Advertisement from reaching the victim node 604 o If the victim implementation includes the (optional) Network_ID 605 parameter for computing F() (see Section 3), and the Network_ID 606 employed by the victim for a remote network is unknown to the 607 attacker, the Interface Identifier learned by the attacker would 608 differ from the one used by the victim when connecting to the 609 legitimate network. 611 In any case, we note that at the point in which this kind of attack 612 becomes a concern, a host should consider employing Secure Neighbor 613 Discovery (SEND) [RFC3971] to prevent an attacker from illegitimately 614 claiming authority for a network prefix. 616 We note that this algorithm is meant to be an alternative to 617 Interface Identifiers such as those specified in [RFC2464], but is 618 not meant as an alternative to temporary Interface Identifiers (such 619 as those specified in [RFC4941]). Clearly, temporary addresses may 620 help to mitigate the correlation of activities of a node within the 621 same network, and may also reduce the attack exposure window (since 622 temporary addresses are short-lived when compared to the addresses 623 generated with the method specified in this document). We note that 624 implementation of this algorithm would still benefit those hosts 625 employing "temporary addresses", since it would mitigate host- 626 tracking vectors still present when such addresses are used (see 627 Appendix C.1), and would also mitigate address-scanning techniques 628 that leverage patterns in IPv6 addresses that embed IEEE identifiers. 630 Finally, we note that the method described in this document addresses 631 some of the privacy concerns arising from the use of IPv6 addresses 632 that embed IEEE identifiers, without the use of temporary addresses, 633 thus possibly offering an interesting trade-off for those scenarios 634 in which the use of temporary addresses is not feasible. 636 8. Acknowledgements 638 The algorithm specified in this document has been inspired by Steven 639 Bellovin's work ([RFC1948]) in the area of TCP sequence numbers. 641 The author would like to thank (in alphabetical order) Ran Atkinson, 642 Karl Auer, Steven Bellovin, Matthias Bethke, Ben Campbell, Brian 643 Carpenter, Tassos Chatzithomaoglou, Tim Chown, Alissa Cooper, Dominik 644 Elsbroek, Brian Haberman, Bob Hinden, Christian Huitema, Ray Hunter, 645 Jouni Korhonen, Eliot Lear, Jong-Hyouk Lee, Andrew McGregor, Tom 646 Petch, Michael Richardson, Mark Smith, Dave Thaler, Ole Troan, James 647 Woodyatt, and He Xuan, for providing valuable comments on earlier 648 versions of this document. 650 This document is based on the technical report "Security Assessment 651 of the Internet Protocol version 6 (IPv6)" [CPNI-IPv6] authored by 652 Fernando Gont on behalf of the UK Centre for the Protection of 653 National Infrastructure (CPNI). 655 Fernando Gont would like to thank CPNI (http://www.cpni.gov.uk) for 656 their continued support. 658 9. References 660 9.1. Normative References 662 [RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 663 (IPv6) Specification", RFC 2460, December 1998. 665 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 666 Requirement Levels", BCP 14, RFC 2119, March 1997. 668 [RFC2526] Johnson, D. and S. Deering, "Reserved IPv6 Subnet Anycast 669 Addresses", RFC 2526, March 1999. 671 [RFC3493] Gilligan, R., Thomson, S., Bound, J., McCann, J., and W. 672 Stevens, "Basic Socket Interface Extensions for IPv6", 673 RFC 3493, February 2003. 675 [RFC3542] Stevens, W., Thomas, M., Nordmark, E., and T. Jinmei, 676 "Advanced Sockets Application Program Interface (API) for 677 IPv6", RFC 3542, May 2003. 679 [RFC3971] Arkko, J., Kempf, J., Zill, B., and P. Nikander, "SEcure 680 Neighbor Discovery (SEND)", RFC 3971, March 2005. 682 [RFC3972] Aura, T., "Cryptographically Generated Addresses (CGA)", 683 RFC 3972, March 2005. 685 [RFC4086] Eastlake, D., Schiller, J., and S. Crocker, "Randomness 686 Requirements for Security", BCP 106, RFC 4086, June 2005. 688 [RFC4122] Leach, P., Mealling, M., and R. Salz, "A Universally 689 Unique IDentifier (UUID) URN Namespace", RFC 4122, 690 July 2005. 692 [RFC4291] Hinden, R. and S. Deering, "IP Version 6 Addressing 693 Architecture", RFC 4291, February 2006. 695 [RFC4861] Narten, T., Nordmark, E., Simpson, W., and H. Soliman, 696 "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861, 697 September 2007. 699 [RFC4862] Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless 700 Address Autoconfiguration", RFC 4862, September 2007. 702 [RFC4941] Narten, T., Draves, R., and S. Krishnan, "Privacy 703 Extensions for Stateless Address Autoconfiguration in 704 IPv6", RFC 4941, September 2007. 706 [RFC6105] Levy-Abegnoli, E., Van de Velde, G., Popoviciu, C., and J. 707 Mohacsi, "IPv6 Router Advertisement Guard", RFC 6105, 708 February 2011. 710 9.2. Informative References 712 [RFC1948] Bellovin, S., "Defending Against Sequence Number Attacks", 713 RFC 1948, May 1996. 715 [RFC2464] Crawford, M., "Transmission of IPv6 Packets over Ethernet 716 Networks", RFC 2464, December 1998. 718 [I-D.ietf-opsec-ipv6-host-scanning] 719 Gont, F. and T. Chown, "Network Reconnaissance in IPv6 720 Networks", draft-ietf-opsec-ipv6-host-scanning-01 (work in 721 progress), April 2013. 723 [I-D.ietf-v6ops-ra-guard-implementation] 724 Gont, F., "Implementation Advice for IPv6 Router 725 Advertisement Guard (RA-Guard)", 726 draft-ietf-v6ops-ra-guard-implementation-07 (work in 727 progress), November 2012. 729 [HDMoore] HD Moore, "The Wild West", Louisville, Kentucky, U.S.A. 730 September 25-29, 2012, 731 . 733 [Gont-DEEPSEC2011] 734 Gont, "Results of a Security Assessment of the Internet 735 Protocol version 6 (IPv6)", DEEPSEC 2011 Conference, 736 Vienna, Austria, November 2011, . 740 [Gont-BRUCON2012] 741 Gont, "Recent Advances in IPv6 Security", BRUCON 2012 742 Conference, Ghent, Belgium, September 2012, . 746 [Broersma] 747 Broersma, R., "IPv6 Everywhere: Living with a Fully IPv6- 748 enabled environment", Australian IPv6 Summit 2010, 749 Melbourne, VIC Australia, October 2010, . 752 [IAB-PRIVACY] 753 IAB, "Privacy and IPv6 Addresses", July 2011, . 757 [CPNI-IPv6] 758 Gont, F., "Security Assessment of the Internet Protocol 759 version 6 (IPv6)", UK Centre for the Protection of 760 National Infrastructure, (available on request). 762 Appendix A. Possible sources for the Net_Iface parameter 764 The following subsections describe a number of possible sources for 765 the Net_Iface parameter employed by the F() function in Section 3. 766 The choice of a specific source for this value represents a number of 767 trade-offs, which may vary from one implementation to another. 769 A.1. Interface Index 771 The Interface Index [RFC3493] [RFC3542] of an interface uniquely 772 identifies an interface within a node. However, these identifiers 773 might or might not have the stability properties required for the 774 Net_Iface value employed by this method. For example, the Interface 775 Index might change upon removal or installation of a network 776 interface (typically one with a smaller value for the Interface 777 Index, when such a naming scheme is used), or when network interfaces 778 happen to be initialized in a different order. We note that some 779 implementations are known to provide configuration knobs to set the 780 Interface Index for a given interface. Such configuration knobs 781 could be employed to prevent the Interface Index from changing (e.g. 782 as a result of the removal of a network interface). 784 A.2. Interface Name 786 The Interface Name (e.g., "eth0", "em0", etc) tends to be more stable 787 than the underlying Interface Index, since such stability is 788 required/desired when interface names are employed in network 789 configuration (firewall rules, etc.). The stability properties of 790 Interface Names depend on implementation details, such as what is the 791 namespace used for Interface Names. For example, "generic" interface 792 names such as "eth0" or "wlan0" will generally be invariant with 793 respect to network interface card replacements. On the other hand, 794 vendor-dependent interface names such as "rtk0" or the like will 795 generally change when a network interface card is replaced with one 796 from a different vendor. 798 We note that Interface Names might still change when network 799 interfaces are added or removed once the system has been bootstrapped 800 (for example, consider Universal Serial Bus-based network interface 801 cards which might be added or removed once the system has been 802 bootstrapped). 804 A.3. Link-layer Addresses 806 Link-layer addresses typically provide for unique identifiers for 807 network interfaces; although, for obvious reasons, they generally 808 change when a network interface card is replaced. In scenarios where 809 neither Interface Indexes nor Interface Names have the stability 810 properties specified in Section 3 for Net_Iface, an implementation 811 might want to employ the link-layer address of the interface for the 812 Net_Iface parameter, albeit at the expense of making the 813 corresponding IPv6 addresses dependent on the underlying network 814 interface card (i.e., the corresponding IPv6 address would typically 815 change upon replacement of the underlying network interface card). 817 A.4. Logical Network Service Identity 819 Host operating systems with a conception of logical network service 820 identity, distinct from network interface identity or index, may keep 821 a Universally Unique Identifier (UUID) [RFC4122] or similar 822 identifier with the stability properties appropriate for use as the 823 Net_Iface parameter. 825 Appendix B. Security/privacy issues with traditional SLAAC addresses 827 This section provides additional details regarding security/privacy 828 issues arising from traditional SLAAC addresses- Namely, it provides 829 additional details regarding how those issues could be exploited, and 830 the extent to which the method specified in this document mitigates 831 such issues. 833 B.1. Correlation of node activities within the same network 835 Since traditional SLAAC addresses employ Interface Identifiers that 836 are constant within the same network, such identifiers can be 837 leveraged to correlate the activities of a node within the same 838 network. One sample scenario is that in which a client repeatedly 839 connects to a server over a period of time, and hence, based on the 840 stable Interface Identifier, the server can correlate all 841 communication instances as being initiated by the same node. 843 The method specified in this document does not mitigate this attack 844 vector, since it produces Interface Identifiers that are constant 845 within a given network. 847 This attack vector could only be mitigated by employing "temporary 848 addresses" [RFC4941]. However, as noted in Appendix C.4, in 849 scenarios in which there is a reduced number of nodes in a given 850 network, mitigation of this vector might be difficult (if at all 851 possible) -- even with "temporary addresses" [RFC4941] in place. 853 B.2. Correlation of node activities across networks (host tracking) 855 Since traditional SLAAC addresses employ Interface Identifiers that 856 are constant across networks, such identifiers can be leveraged to 857 correlate the activities of a node across networks. 859 A passive version of this attack would be that in which a client 860 repeatedly connects to an attacker-operated server over a period of 861 time, and hence, based on the client's stable Interface Identifier, 862 the server can correlate all communication instances as being 863 initiated by the same node. 865 An attacker could also launch an active version of this attack. For 866 example, let us assume that the attacker knows the Interface 867 Identifier employed by the target node (e.g., the target and the 868 attacker were simultaneously connected to the same subnetwork at some 869 point in time). If the attacker knows the possible networks the 870 target might connect to, he could probe whether there is an address 871 with the target's Interface Identifier in each of those networks. If 872 such address is found to be "alive", then the attacker could infer 873 that the target node has connected to the corresponding network. 875 This vector is discussed in detail in Appendix C.1.2. 877 Since the method specified in this document results in Interface 878 Identifiers that are not constant across networks, both the passive 879 and active versions of this attack vector are mitigated. 881 B.3. Address-scanning attacks 883 Since traditional SLAAC addresses typically embed the underlying 884 link-layer address, the aforementioned addresses follow specific 885 patterns that can be leveraged to reduce the search space when 886 performing IPv6 address-scanning attacks (this is discussed in detail 887 in [I-D.ietf-opsec-ipv6-host-scanning]). The method specified in 888 this document produces random (but table within each subnet) 889 Interface Identifiers, thus mitigating this attack vector. 891 B.4. Exploitation of device-specific information 893 Since traditional SLAAC addresses typically embed the underlying 894 link-layer address, the aforementioned addresses leaks device- 895 specific information that might be leveraged to launch device- 896 specific attacks. For example, an attacker with knowledge about a 897 specific vulnerability in devices manufactured by some vendor might 898 easily identify potential targets by looking at the Interface 899 Identifier of a list of IPv6 addresses. The method specified in this 900 document produces random (but table within each subnet) Interface 901 Identifiers, thus mitigating this attack vector. 903 Appendix C. Privacy issues still present when temporary addresses are 904 employed 906 It is not unusual for people to assume or expect that all the 907 security/privacy implications of traditional SLAAC addresses are 908 mitigated when "temporary addresses" [RFC4941] are employed. 909 However, as noted in Section 1 of this document and [IAB-PRIVACY], 910 since temporary addresses are employed in addition to (rather than in 911 replacement of) traditional SLAAC addresses, many of the security/ 912 privacy implications of traditional SLAAC addresses are not mitigated 913 by the use of temporary addresses. 915 This section discusses a (non-exhaustive) number of scenarios in 916 which host security/privacy is still negatively affected as a result 917 of employing Interface Identifiers that are constant across networks 918 (e.g., those resulting from embedding IEEE identifiers), even when 919 temporary addresses [RFC4941] are employed. It aims to clarify the 920 motivation of employing the method specified in this document in 921 replacement of the traditional SLAAC addresses even when privacy/ 922 temporary addresses [RFC4941] are employed. 924 C.1. Host tracking 926 This section describes two attack scenarios which illustrate that 927 host-tracking may still be possible when privacy/temporary addresses 928 [RFC4941] are employed. These examples should remind us that one 929 should not disclose more than it is really needed for achieving a 930 specific goal (and an Interface Identifier that is constant across 931 different networks does exactly that: it discloses more information 932 than is needed for providing a stable address). 934 C.1.1. Tracking hosts across networks #1 936 A host configures its stable addresses with the constant Interface 937 Identifier, and runs any application that needs to perform a server- 938 like function (e.g. a peer-to-peer application). As a result of 939 that, an attacker/user participating in the same application (e.g., 940 P2P) would learn the constant Interface Identifier used by the host 941 for that network interface. 943 Some time later, the same host moves to a completely different 944 network, and employs the same P2P application. The attacker now 945 interacts with the same host again, and hence can learn its newly- 946 configured stable address. Since the Interface Identifier is the 947 same as the one used before, the attacker can infer that it is 948 communicating with the same device as before. 950 C.1.2. Tracking hosts across networks #2 952 Once an attacker learns the constant Interface Identifier employed by 953 the victim host for its stable address, the attacker is able to 954 "probe" a network for the presence of such host at any given network. 956 See Appendix C.1.1 for just one example of how an attacker could 957 learn such value. Other examples include being able to share the 958 same network segment at some point in time (e.g. a conference 959 network or any public network), etc. 961 For example, if an attacker learns that in one network the victim 962 used the Interface Identifier 1111:2222:3333:4444 for its stable 963 addresses, then he could subsequently probe for the presence of such 964 device in the network 2011:db8::/64 by sending a probe packet (ICMPv6 965 Echo Request, or any other probe packet) to the address 2001:db8:: 966 1111:2222:3333:4444. 968 C.1.3. Revealing the identity of devices performing server-like 969 functions 971 Some devices, such as storage devices, may typically perform server- 972 like functions and may be usually moved from one network to another. 973 Such devices are likely to simply disable (or not even implement) 974 privacy/temporary addresses [RFC4941]. If the aforementioned devices 975 employ Interface Identifiers that are constant across networks, it 976 would be trivial for an attacker to tell whether the same device is 977 being used across networks by simply looking at the Interface 978 Identifier. Clearly, performing server-like functions should not 979 imply that a device discloses its identity (i.e., that the attacker 980 can tell whether it is the same device providing some function in two 981 different networks, at two different points in time). 983 The scheme proposed in this document prevents such information 984 leakage by causing nodes to generate different Interface Identifiers 985 when moving from one network to another, thus mitigating this kind of 986 privacy attack. 988 C.2. Address-scanning attacks 990 While it is usually assumed that IPv6 address-scanning attacks are 991 unfeasible, an attacker can leverage address patterns in IPv6 992 addresses to greatly reduce the search space 993 [I-D.ietf-opsec-ipv6-host-scanning] [Gont-BRUCON2012]. Addresses 994 that embed IEEE identifiers result in one of such patterns that could 995 be leveraged to reduce the search space when other nodes employ the 996 same IEEE OUI (Organizationally Unique Identifier). 998 As noted earlier in this document, temporary addresses [RFC4941] do 999 not replace/eliminate the use of IPv6 addresses that embed IEEE 1000 identifiers (they are employed in addition to those), and hence hosts 1001 implementing [RFC4941] would still be vulnerable to address-scanning 1002 attacks. The method specified in this document is meant as an 1003 alternative to addresses that embed IEEE identifiers, and hence 1004 eliminates such patterns (thus mitigating the aforementioned address- 1005 scanning attacks). 1007 C.3. Information Leakage 1009 IPv6 addresses embedding IEEE identifiers leak information about the 1010 device (Network Interface Card vendor, or even Operating System 1011 and/or software type), which could be leveraged by an attacker with 1012 device/software-specific vulnerabilities knowledge to quickly find 1013 possible targets. Since temporary addresses do not replace the 1014 traditional SLAAC addresses that typically embed IEEE identifiers, 1015 employing temporary addresses does not eliminate this possible 1016 information leakage. 1018 C.4. Correlation of node activities within a network 1020 In scenarios in which the number of nodes connected to a subnetwork 1021 is small, preventing the correlation of the activities of those nodes 1022 within such network might be difficult (if at all possible) to 1023 achieve, even with temporary addresses [RFC4941] in place. As a 1024 trivial example, consider a scenario where there is a single node (or 1025 a reduced number of nodes) connected to a specific network. An 1026 attacker could detect new addresses in use at that network along with 1027 addresses that are no longer in use, and infer which addresses are 1028 being employed by which hosts. This task is made particularly easier 1029 by the fact that use of "temporary addresses" can be easily inferred 1030 (since they follow different patterns from that of traditional SLAAC 1031 addresses), and since they are re-generated periodically (i.e., after 1032 a specific amount of time has elapsed). 1034 Author's Address 1036 Fernando Gont 1037 SI6 Networks / UTN-FRH 1038 Evaristo Carriego 2644 1039 Haedo, Provincia de Buenos Aires 1706 1040 Argentina 1042 Phone: +54 11 4650 8472 1043 Email: fgont@si6networks.com 1044 URI: http://www.si6networks.com