idnits 2.17.1 draft-ietf-rtcweb-security-06.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 seems to contain a disclaimer for pre-RFC5378 work, but was first submitted on or after 10 November 2008. The disclaimer is usually necessary only for documents that revise or obsolete older RFCs, and that take significant amounts of text from those RFCs. If you can contact all authors of the source material and they are willing to grant the BCP78 rights to the IETF Trust, you can and should remove the disclaimer. Otherwise, the disclaimer is needed and you can ignore this comment. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (January 22, 2014) is 3718 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) == Unused Reference: 'I-D.kaufman-rtcweb-security-ui' is defined on line 980, but no explicit reference was found in the text == Unused Reference: 'RFC2818' is defined on line 991, but no explicit reference was found in the text == Unused Reference: 'RFC6454' is defined on line 1042, but no explicit reference was found in the text == Outdated reference: A later version (-19) exists of draft-ietf-rtcweb-overview-08 == Outdated reference: A later version (-20) exists of draft-ietf-rtcweb-security-arch-07 -- Obsolete informational reference (is this intentional?): RFC 2818 (Obsoleted by RFC 9110) -- Obsolete informational reference (is this intentional?): RFC 4347 (Obsoleted by RFC 6347) -- Obsolete informational reference (is this intentional?): RFC 5245 (Obsoleted by RFC 8445, RFC 8839) -- Obsolete informational reference (is this intentional?): RFC 6222 (Obsoleted by RFC 7022) Summary: 1 error (**), 0 flaws (~~), 7 warnings (==), 5 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 RTC-Web E. Rescorla 3 Internet-Draft RTFM, Inc. 4 Intended status: Standards Track January 22, 2014 5 Expires: July 26, 2014 7 Security Considerations for WebRTC 8 draft-ietf-rtcweb-security-06 10 Abstract 12 The Real-Time Communications on the Web (RTCWEB) working group is 13 tasked with standardizing protocols for real-time communications 14 between Web browsers, generally called "WebRTC". The major use cases 15 for WebRTC technology are real-time audio and/or video calls, Web 16 conferencing, and direct data transfer. Unlike most conventional 17 real-time systems (e.g., SIP-based soft phones) WebRTC communications 18 are directly controlled by a Web server, which poses new security 19 challenges. For instance, a Web browser might expose a JavaScript 20 API which allows a server to place a video call. Unrestricted access 21 to such an API would allow any site which a user visited to "bug" a 22 user's computer, capturing any activity which passed in front of 23 their camera. This document defines the WebRTC threat model and 24 analyzes the security threats of WebRTC in that model. 26 Status of this Memo 28 This Internet-Draft is submitted in full conformance with the 29 provisions of BCP 78 and BCP 79. 31 Internet-Drafts are working documents of the Internet Engineering 32 Task Force (IETF). Note that other groups may also distribute 33 working documents as Internet-Drafts. The list of current Internet- 34 Drafts is at http://datatracker.ietf.org/drafts/current/. 36 Internet-Drafts are draft documents valid for a maximum of six months 37 and may be updated, replaced, or obsoleted by other documents at any 38 time. It is inappropriate to use Internet-Drafts as reference 39 material or to cite them other than as "work in progress." 41 This Internet-Draft will expire on July 26, 2014. 43 Copyright Notice 45 Copyright (c) 2014 IETF Trust and the persons identified as the 46 document authors. All rights reserved. 48 This document is subject to BCP 78 and the IETF Trust's Legal 49 Provisions Relating to IETF Documents 50 (http://trustee.ietf.org/license-info) in effect on the date of 51 publication of this document. Please review these documents 52 carefully, as they describe your rights and restrictions with respect 53 to this document. Code Components extracted from this document must 54 include Simplified BSD License text as described in Section 4.e of 55 the Trust Legal Provisions and are provided without warranty as 56 described in the Simplified BSD License. 58 This document may contain material from IETF Documents or IETF 59 Contributions published or made publicly available before November 60 10, 2008. The person(s) controlling the copyright in some of this 61 material may not have granted the IETF Trust the right to allow 62 modifications of such material outside the IETF Standards Process. 63 Without obtaining an adequate license from the person(s) controlling 64 the copyright in such materials, this document may not be modified 65 outside the IETF Standards Process, and derivative works of it may 66 not be created outside the IETF Standards Process, except to format 67 it for publication as an RFC or to translate it into languages other 68 than English. 70 Table of Contents 72 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 73 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 5 74 3. The Browser Threat Model . . . . . . . . . . . . . . . . . . . 5 75 3.1. Access to Local Resources . . . . . . . . . . . . . . . . 6 76 3.2. Same Origin Policy . . . . . . . . . . . . . . . . . . . . 6 77 3.3. Bypassing SOP: CORS, WebSockets, and consent to 78 communicate . . . . . . . . . . . . . . . . . . . . . . . 7 79 4. Security for WebRTC Applications . . . . . . . . . . . . . . . 7 80 4.1. Access to Local Devices . . . . . . . . . . . . . . . . . 8 81 4.1.1. Threats from Screen Sharing . . . . . . . . . . . . . 9 82 4.1.2. Calling Scenarios and User Expectations . . . . . . . 9 83 4.1.2.1. Dedicated Calling Services . . . . . . . . . . . . 9 84 4.1.2.2. Calling the Site You're On . . . . . . . . . . . . 10 85 4.1.3. Origin-Based Security . . . . . . . . . . . . . . . . 10 86 4.1.4. Security Properties of the Calling Page . . . . . . . 12 87 4.2. Communications Consent Verification . . . . . . . . . . . 13 88 4.2.1. ICE . . . . . . . . . . . . . . . . . . . . . . . . . 13 89 4.2.2. Masking . . . . . . . . . . . . . . . . . . . . . . . 13 90 4.2.3. Backward Compatibility . . . . . . . . . . . . . . . . 14 91 4.2.4. IP Location Privacy . . . . . . . . . . . . . . . . . 15 92 4.3. Communications Security . . . . . . . . . . . . . . . . . 15 93 4.3.1. Protecting Against Retrospective Compromise . . . . . 16 94 4.3.2. Protecting Against During-Call Attack . . . . . . . . 17 95 4.3.2.1. Key Continuity . . . . . . . . . . . . . . . . . . 17 96 4.3.2.2. Short Authentication Strings . . . . . . . . . . . 18 97 4.3.2.3. Third Party Identity . . . . . . . . . . . . . . . 19 98 4.3.2.4. Page Access to Media . . . . . . . . . . . . . . . 19 99 4.3.3. Malicious Peers . . . . . . . . . . . . . . . . . . . 20 100 4.4. Privacy Considerations . . . . . . . . . . . . . . . . . . 20 101 4.4.1. Correlation of Anonymous Calls . . . . . . . . . . . . 20 102 4.4.2. Browser Fingerprinting . . . . . . . . . . . . . . . . 21 103 5. Security Considerations . . . . . . . . . . . . . . . . . . . 21 104 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 21 105 7. Changes Since -04 . . . . . . . . . . . . . . . . . . . . . . 21 106 8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 21 107 8.1. Normative References . . . . . . . . . . . . . . . . . . . 21 108 8.2. Informative References . . . . . . . . . . . . . . . . . . 22 109 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 24 111 1. Introduction 113 The Real-Time Communications on the Web (RTCWEB) working group is 114 tasked with standardizing protocols for real-time communications 115 between Web browsers, generally called "WebRTC" 116 [I-D.ietf-rtcweb-overview]. The major use cases for WebTC technology 117 are real-time audio and/or video calls, Web conferencing, and direct 118 data transfer. Unlike most conventional real-time systems, (e.g., 119 SIP-based[RFC3261] soft phones) WebRTC communications are directly 120 controlled by some Web server. A simple case is shown below. 122 +----------------+ 123 | | 124 | Web Server | 125 | | 126 +----------------+ 127 ^ ^ 128 / \ 129 HTTP / \ HTTP 130 or / \ or 131 WebSockets / \ WebSockets 132 v v 133 JS API JS API 134 +-----------+ +-----------+ 135 | | Media | | 136 | Browser |<---------->| Browser | 137 | | | | 138 +-----------+ +-----------+ 140 Figure 1: A simple WebRTC system 142 In the system shown in Figure 1, Alice and Bob both have WebRTC 143 enabled browsers and they visit some Web server which operates a 144 calling service. Each of their browsers exposes standardized 145 JavaScript calling APIs (implementated as browser built-ins) which 146 are used by the Web server to set up a call between Alice and Bob. 147 The Web server also serves as the signaling channel to transport 148 control messages between the browsers. While this system is 149 topologically similar to a conventional SIP-based system (with the 150 Web server acting as the signaling service and browsers acting as 151 softphones), control has moved to the central Web server; the browser 152 simply provides API points that are used by the calling service. As 153 with any Web application, the Web server can move logic between the 154 server and JavaScript in the browser, but regardless of where the 155 code is executing, it is ultimately under control of the server. 157 It should be immediately apparent that this type of system poses new 158 security challenges beyond those of a conventional VoIP system. In 159 particular, it needs to contend with malicious calling services. For 160 example, if the calling service can cause the browser to make a call 161 at any time to any callee of its choice, then this facility can be 162 used to bug a user's computer without their knowledge, simply by 163 placing a call to some recording service. More subtly, if the 164 exposed APIs allow the server to instruct the browser to send 165 arbitrary content, then they can be used to bypass firewalls or mount 166 denial of service attacks. Any successful system will need to be 167 resistant to this and other attacks. 169 A companion document [I-D.ietf-rtcweb-security-arch] describes a 170 security architecture intended to address the issues raised in this 171 document. 173 2. Terminology 175 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 176 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 177 document are to be interpreted as described in RFC 2119 [RFC2119]. 179 3. The Browser Threat Model 181 The security requirements for WebRTC follow directly from the 182 requirement that the browser's job is to protect the user. Huang et 183 al. [huang-w2sp] summarize the core browser security guarantee as: 185 Users can safely visit arbitrary web sites and execute scripts 186 provided by those sites. 188 It is important to realize that this includes sites hosting arbitrary 189 malicious scripts. The motivation for this requirement is simple: 190 it is trivial for attackers to divert users to sites of their choice. 191 For instance, an attacker can purchase display advertisements which 192 direct the user (either automatically or via user clicking) to their 193 site, at which point the browser will execute the attacker's scripts. 194 Thus, it is important that it be safe to view arbitrarily malicious 195 pages. Of course, browsers inevitably have bugs which cause them to 196 fall short of this goal, but any new WebRTC functionality must be 197 designed with the intent to meet this standard. The remainder of 198 this section provides more background on the existing Web security 199 model. 201 In this model, then, the browser acts as a TRUSTED COMPUTING BASE 202 (TCB) both from the user's perspective and to some extent from the 203 server's. While HTML and JavaScript (JS) provided by the server can 204 cause the browser to execute a variety of actions, those scripts 205 operate in a sandbox that isolates them both from the user's computer 206 and from each other, as detailed below. 208 Conventionally, we refer to either WEB ATTACKERS, who are able to 209 induce you to visit their sites but do not control the network, and 210 NETWORK ATTACKERS, who are able to control your network. Network 211 attackers correspond to the [RFC3552] "Internet Threat Model". Note 212 that for HTTP traffic, a network attacker is also a Web attacker, 213 since it can inject traffic as if it were any non-HTTPS Web site. 214 Thus, when analyzing HTTP connections, we must assume that traffic is 215 going to the attacker. 217 3.1. Access to Local Resources 219 While the browser has access to local resources such as keying 220 material, files, the camera and the microphone, it strictly limits or 221 forbids web servers from accessing those same resources. For 222 instance, while it is possible to produce an HTML form which will 223 allow file upload, a script cannot do so without user consent and in 224 fact cannot even suggest a specific file (e.g., /etc/passwd); the 225 user must explicitly select the file and consent to its upload. 226 [Note: in many cases browsers are explicitly designed to avoid 227 dialogs with the semantics of "click here to screw yourself", as 228 extensive research shows that users are prone to consent under such 229 circumstances.] 231 Similarly, while Flash programs (SWFs) [SWF] can access the camera 232 and microphone, they explicitly require that the user consent to that 233 access. In addition, some resources simply cannot be accessed from 234 the browser at all. For instance, there is no real way to run 235 specific executables directly from a script (though the user can of 236 course be induced to download executable files and run them). 238 3.2. Same Origin Policy 240 Many other resources are accessible but isolated. For instance, 241 while scripts are allowed to make HTTP requests via the 242 XMLHttpRequest() API those requests are not allowed to be made to any 243 server, but rather solely to the same ORIGIN from whence the script 244 came xref target="RFC6454"/> (although CORS [CORS] and WebSockets 245 [RFC6455] provide a escape hatch from this restriction, as described 246 below.) This SAME ORIGIN POLICY (SOP) prevents server A from 247 mounting attacks on server B via the user's browser, which protects 248 both the user (e.g., from misuse of his credentials) and the server B 249 (e.g., from DoS attack). 251 More generally, SOP forces scripts from each site to run in their 252 own, isolated, sandboxes. While there are techniques to allow them 253 to interact, those interactions generally must be mutually consensual 254 (by each site) and are limited to certain channels. For instance, 255 multiple pages/browser panes from the same origin can read each 256 other's JS variables, but pages from the different origins--or even 257 iframes from different origins on the same page--cannot. 259 3.3. Bypassing SOP: CORS, WebSockets, and consent to communicate 261 While SOP serves an important security function, it also makes it 262 inconvenient to write certain classes of applications. In 263 particular, mash-ups, in which a script from origin A uses resources 264 from origin B, can only be achieved via a certain amount of hackery. 265 The W3C Cross-Origin Resource Sharing (CORS) spec [CORS] is a 266 response to this demand. In CORS, when a script from origin A 267 executes what would otherwise be a forbidden cross-origin request, 268 the browser instead contacts the target server to determine whether 269 it is willing to allow cross-origin requests from A. If it is so 270 willing, the browser then allows the request. This consent 271 verification process is designed to safely allow cross-origin 272 requests. 274 While CORS is designed to allow cross-origin HTTP requests, 275 WebSockets [RFC6455] allows cross-origin establishment of transparent 276 channels. Once a WebSockets connection has been established from a 277 script to a site, the script can exchange any traffic it likes 278 without being required to frame it as a series of HTTP request/ 279 response transactions. As with CORS, a WebSockets transaction starts 280 with a consent verification stage to avoid allowing scripts to simply 281 send arbitrary data to another origin. 283 While consent verification is conceptually simple--just do a 284 handshake before you start exchanging the real data--experience has 285 shown that designing a correct consent verification system is 286 difficult. In particular, Huang et al. [huang-w2sp] have shown 287 vulnerabilities in the existing Java and Flash consent verification 288 techniques and in a simplified version of the WebSockets handshake. 289 In particular, it is important to be wary of CROSS-PROTOCOL attacks 290 in which the attacking script generates traffic which is acceptable 291 to some non-Web protocol state machine. In order to resist this form 292 of attack, WebSockets incorporates a masking technique intended to 293 randomize the bits on the wire, thus making it more difficult to 294 generate traffic which resembles a given protocol. 296 4. Security for WebRTC Applications 297 4.1. Access to Local Devices 299 As discussed in Section 1, allowing arbitrary sites to initiate calls 300 violates the core Web security guarantee; without some access 301 restrictions on local devices, any malicious site could simply bug a 302 user. At minimum, then, it MUST NOT be possible for arbitrary sites 303 to initiate calls to arbitrary locations without user consent. This 304 immediately raises the question, however, of what should be the scope 305 of user consent. 307 In order for the user to make an intelligent decision about whether 308 to allow a call (and hence his camera and microphone input to be 309 routed somewhere), he must understand either who is requesting 310 access, where the media is going, or both. As detailed below, there 311 are two basic conceptual models: 313 You are sending your media to entity A because you want to talk to 314 Entity A (e.g., your mother). 315 Entity A (e.g., a calling service) asks to access the user's 316 devices with the assurance that it will transfer the media to 317 entity B (e.g., your mother) 319 In either case, identity is at the heart of any consent decision. 320 Moreover, identity is all that the browser can meaningfully enforce; 321 if you are calling A, A can simply forward the media to C. Similarly, 322 if you authorize A to place a call to B, A can call C instead. In 323 either case, all the browser is able to do is verify and check 324 authorization for whoever is controlling where the media goes. The 325 target of the media can of course advertise a security/privacy 326 policy, but this is not something that the browser can enforce. Even 327 so, there are a variety of different consent scenarios that motivate 328 different technical consent mechanisms. We discuss these mechanisms 329 in the sections below. 331 It's important to understand that consent to access local devices is 332 largely orthogonal to consent to transmit various kinds of data over 333 the network (see Section 4.2. Consent for device access is largely a 334 matter of protecting the user's privacy from malicious sites. By 335 contrast, consent to send network traffic is about preventing the 336 user's browser from being used to attack its local network. Thus, we 337 need to ensure communications consent even if the site is not able to 338 access the camera and microphone at all (hence WebSockets's consent 339 mechanism) and similarly we need to be concerned with the site 340 accessing the user's camera and microphone even if the data is to be 341 sent back to the site via conventional HTTP-based network mechanisms 342 such as HTTP POST. 344 4.1.1. Threats from Screen Sharing 346 In addition to camera and microphone access, there has been demand 347 for screen and/or application sharing functionality. Unfortunately, 348 the security implications of this functionality are much harder for 349 users to intuitively analyze than for camera and microphone access. 350 (See 351 http://lists.w3.org/Archives/Public/public-webrtc/2013Mar/0024.html 352 for a full analysis.) 354 The most obvious threats are simply those of "oversharing". I.e., 355 the user may believe they are sharing a window when in fact they are 356 sharing an application, or may forget they are sharing their whole 357 screen, icons, notifications, and all. This is already an issue with 358 existing screen sharing technologies and is made somewhat worse if a 359 partially trusted site is responsible for asking for the resource to 360 be shared rather than having the user propose it. 362 A less obvious threat involves the impact of screen sharing on the 363 Web security model. A key part of the Same Origin Policy is that 364 HTML or JS from site A can reference content from site B and cause 365 the browser to load it, but (unless explicitly permitted) cannot see 366 the result. However, if a web application from a site is screen 367 sharing the browser, then this violates that invariant, with serious 368 security consequences. For example, an attacker site might request 369 screen sharing and then briefly open up a new Window to the user's 370 bank or webmail account, using screen sharing to read the resulting 371 displayed content. A more sophisticated attack would be open up a 372 source view window to a site and use the screen sharing result to 373 view anti cross-site request forgery tokens. 375 These threats suggest that screen/application sharing might need a 376 higher level of user consent than access to the camera or microphone. 378 4.1.2. Calling Scenarios and User Expectations 380 While a large number of possible calling scenarios are possible, the 381 scenarios discussed in this section illustrate many of the 382 difficulties of identifying the relevant scope of consent. 384 4.1.2.1. Dedicated Calling Services 386 The first scenario we consider is a dedicated calling service. In 387 this case, the user has a relationship with a calling site and 388 repeatedly makes calls on it. It is likely that rather than having 389 to give permission for each call that the user will want to give the 390 calling service long-term access to the camera and microphone. This 391 is a natural fit for a long-term consent mechanism (e.g., installing 392 an app store "application" to indicate permission for the calling 393 service.) A variant of the dedicated calling service is a gaming 394 site (e.g., a poker site) which hosts a dedicated calling service to 395 allow players to call each other. 397 With any kind of service where the user may use the same service to 398 talk to many different people, there is a question about whether the 399 user can know who they are talking to. If I grant permission to 400 calling service A to make calls on my behalf, then I am implicitly 401 granting it permission to bug my computer whenever it wants. This 402 suggests another consent model in which a site is authorized to make 403 calls but only to certain target entities (identified via media-plane 404 cryptographic mechanisms as described in Section 4.3.2 and especially 405 Section 4.3.2.3.) Note that the question of consent here is related 406 to but distinct from the question of peer identity: I might be 407 willing to allow a calling site to in general initiate calls on my 408 behalf but still have some calls via that site where I can be sure 409 that the site is not listening in. 411 4.1.2.2. Calling the Site You're On 413 Another simple scenario is calling the site you're actually visiting. 414 The paradigmatic case here is the "click here to talk to a 415 representative" windows that appear on many shopping sites. In this 416 case, the user's expectation is that they are calling the site 417 they're actually visiting. However, it is unlikely that they want to 418 provide a general consent to such a site; just because I want some 419 information on a car doesn't mean that I want the car manufacturer to 420 be able to activate my microphone whenever they please. Thus, this 421 suggests the need for a second consent mechanism where I only grant 422 consent for the duration of a given call. As described in 423 Section 3.1, great care must be taken in the design of this interface 424 to avoid the users just clicking through. Note also that the user 425 interface chrome must clearly display elements showing that the call 426 is continuing in order to avoid attacks where the calling site just 427 leaves it up indefinitely but shows a Web UI that implies otherwise. 429 4.1.3. Origin-Based Security 431 Now that we have seen another use case, we can start to reason about 432 the security requirements. 434 As discussed in Section 3.2, the basic unit of Web sandboxing is the 435 origin, and so it is natural to scope consent to origin. 436 Specifically, a script from origin A MUST only be allowed to initiate 437 communications (and hence to access camera and microphone) if the 438 user has specifically authorized access for that origin. It is of 439 course technically possible to have coarser-scoped permissions, but 440 because the Web model is scoped to origin, this creates a difficult 441 mismatch. 443 Arguably, origin is not fine-grained enough. Consider the situation 444 where Alice visits a site and authorizes it to make a single call. 445 If consent is expressed solely in terms of origin, then at any future 446 visit to that site (including one induced via mash-up or ad network), 447 the site can bug Alice's computer, use the computer to place bogus 448 calls, etc. While in principle Alice could grant and then revoke the 449 privilege, in practice privileges accumulate; if we are concerned 450 about this attack, something else is needed. There are a number of 451 potential countermeasures to this sort of issue. 453 Individual Consent 454 Ask the user for permission for each call. 456 Callee-oriented Consent 457 Only allow calls to a given user. 459 Cryptographic Consent 460 Only allow calls to a given set of peer keying material or to a 461 cryptographically established identity. 463 Unfortunately, none of these approaches is satisfactory for all 464 cases. As discussed above, individual consent puts the user's 465 approval in the UI flow for every call. Not only does this quickly 466 become annoying but it can train the user to simply click "OK", at 467 which point the consent becomes useless. Thus, while it may be 468 necessary to have individual consent in some case, this is not a 469 suitable solution for (for instance) the calling service case. Where 470 necessary, in-flow user interfaces must be carefully designed to 471 avoid the risk of the user blindly clicking through. 473 The other two options are designed to restrict calls to a given 474 target. Callee-oriented consent provided by the calling site not 475 work well because a malicious site can claim that the user is calling 476 any user of his choice. One fix for this is to tie calls to a 477 cryptographically established identity. While not suitable for all 478 cases, this approach may be useful for some. If we consider the case 479 of advertising, it's not particularly convenient to require the 480 advertiser to instantiate an iframe on the hosting site just to get 481 permission; a more convenient approach is to cryptographically tie 482 the advertiser's certificate to the communication directly. We're 483 still tying permissions to origin here, but to the media origin 484 (and-or destination) rather than to the Web origin. 485 [I-D.ietf-rtcweb-security-arch] describes mechanisms which facilitate 486 this sort of consent. 488 Another case where media-level cryptographic identity makes sense is 489 when a user really does not trust the calling site. For instance, I 490 might be worried that the calling service will attempt to bug my 491 computer, but I also want to be able to conveniently call my friends. 492 If consent is tied to particular communications endpoints, then my 493 risk is limited. Naturally, it is somewhat challenging to design UI 494 primitives which express this sort of policy. The problem becomes 495 even more challenging in multi-user calling cases. 497 4.1.4. Security Properties of the Calling Page 499 Origin-based security is intended to secure against web attackers. 500 However, we must also consider the case of network attackers. 501 Consider the case where I have granted permission to a calling 502 service by an origin that has the HTTP scheme, e.g., 503 http://calling-service.example.com. If I ever use my computer on an 504 unsecured network (e.g., a hotspot or if my own home wireless network 505 is insecure), and browse any HTTP site, then an attacker can bug my 506 computer. The attack proceeds like this: 508 1. I connect to http://anything.example.org/. Note that this site 509 is unaffiliated with the calling service. 510 2. The attacker modifies my HTTP connection to inject an IFRAME (or 511 a redirect) to http://calling-service.example.com 512 3. The attacker forges the response apparently 513 http://calling-service.example.com/ to inject JS to initiate a 514 call to himself. 516 Note that this attack does not depend on the media being insecure. 517 Because the call is to the attacker, it is also encrypted to him. 518 Moreover, it need not be executed immediately; the attacker can 519 "infect" the origin semi-permanently (e.g., with a web worker or a 520 popped-up window that is hidden under the main window.) and thus be 521 able to bug me long after I have left the infected network. This 522 risk is created by allowing calls at all from a page fetched over 523 HTTP. 525 Even if calls are only possible from HTTPS sites, if the site embeds 526 active content (e.g., JavaScript) that is fetched over HTTP or from 527 an untrusted site, because that JavaScript is executed in the 528 security context of the page [finer-grained]. Thus, it is also 529 dangerous to allow WebRTC functionality from HTTPS origins that embed 530 mixed content. Note: this issue is not restricted to PAGES which 531 contain mixed content. If a page from a given origin ever loads 532 mixed content then it is possible for a network attacker to infect 533 the browser's notion of that origin semi-permanently. 535 4.2. Communications Consent Verification 537 As discussed in Section 3.3, allowing web applications unrestricted 538 network access via the browser introduces the risk of using the 539 browser as an attack platform against machines which would not 540 otherwise be accessible to the malicious site, for instance because 541 they are topologically restricted (e.g., behind a firewall or NAT). 542 In order to prevent this form of attack as well as cross-protocol 543 attacks it is important to require that the target of traffic 544 explicitly consent to receiving the traffic in question. Until that 545 consent has been verified for a given endpoint, traffic other than 546 the consent handshake MUST NOT be sent to that endpoint. 548 4.2.1. ICE 550 Verifying receiver consent requires some sort of explicit handshake, 551 but conveniently we already need one in order to do NAT hole- 552 punching. ICE [RFC5245] includes a handshake designed to verify that 553 the receiving element wishes to receive traffic from the sender. It 554 is important to remember here that the site initiating ICE is 555 presumed malicious; in order for the handshake to be secure the 556 receiving element MUST demonstrate receipt/knowledge of some value 557 not available to the site (thus preventing the site from forging 558 responses). In order to achieve this objective with ICE, the STUN 559 transaction IDs must be generated by the browser and MUST NOT be made 560 available to the initiating script, even via a diagnostic interface. 561 Verifying receiver consent also requires verifying the receiver wants 562 to receive traffic from a particular sender, and at this time; for 563 example a malicious site may simply attempt ICE to known servers that 564 are using ICE for other sessions. ICE provides this verification as 565 well, by using the STUN credentials as a form of per-session shared 566 secret. Those credentials are known to the Web application, but 567 would need to also be known and used by the STUN-receiving element to 568 be useful. 570 There also needs to be some mechanism for the browser to verify that 571 the target of the traffic continues to wish to receive it. Because 572 ICE keepalives are indications, they will not work here, so some 573 other mechanism is needed as described in 574 [I-D.muthu-behave-consent-freshness]. 576 4.2.2. Masking 578 Once consent is verified, there still is some concern about 579 misinterpretation attacks as described by Huang et al.[huang-w2sp]. 580 Once consent is verified, there still is some concern about 581 misinterpretation attacks as described by Huang et al.[huang-w2sp]. 582 Where TCP is used the risk is substantial due to the potential 583 presence of transparent proxies and therefore if TCP is to be used, 584 then WebSockets style masking MUST be employed. 586 Since DTLS (with the anti-chosen plaintext mechanisms required by TLS 587 1.1) does not allow the attacker to generate predictable ciphertext, 588 there is no need for masking of protocols running over DTLS (e.g. 589 SCTP over DTLS, UDP over DTLS, etc.). 591 4.2.3. Backward Compatibility 593 A requirement to use ICE limits compatibility with legacy non-ICE 594 clients. It seems unsafe to completely remove the requirement for 595 some check. All proposed checks have the common feature that the 596 browser sends some message to the candidate traffic recipient and 597 refuses to send other traffic until that message has been replied to. 598 The message/reply pair must be generated in such a way that an 599 attacker who controls the Web application cannot forge them, 600 generally by having the message contain some secret value that must 601 be incorporated (e.g., echoed, hashed into, etc.). Non-ICE 602 candidates for this role (in cases where the legacy endpoint has a 603 public address) include: 605 o STUN checks without using ICE (i.e., the non-RTC-web endpoint sets 606 up a STUN responder.) 607 o Use or RTCP as an implicit reachability check. 609 In the RTCP approach, the WebRTC endpoint is allowed to send a 610 limited number of RTP packets prior to receiving consent. This 611 allows a short window of attack. In addition, some legacy endpoints 612 do not support RTCP, so this is a much more expensive solution for 613 such endpoints, for which it would likely be easier to implement ICE. 614 For these two reasons, an RTCP-based approach does not seem to 615 address the security issue satisfactorily. 617 In the STUN approach, the WebRTC endpoint is able to verify that the 618 recipient is running some kind of STUN endpoint but unless the STUN 619 responder is integrated with the ICE username/password establishment 620 system, the WebRTC endpoint cannot verify that the recipient consents 621 to this particular call. This may be an issue if existing STUN 622 servers are operated at addresses that are not able to handle 623 bandwidth-based attacks. Thus, this approach does not seem 624 satisfactory either. 626 If the systems are tightly integrated (i.e., the STUN endpoint 627 responds with responses authenticated with ICE credentials) then this 628 issue does not exist. However, such a design is very close to an 629 ICE-Lite implementation (indeed, arguably is one). An intermediate 630 approach would be to have a STUN extension that indicated that one 631 was responding to WebRTC checks but not computing integrity checks 632 based on the ICE credentials. This would allow the use of standalone 633 STUN servers without the risk of confusing them with legacy STUN 634 servers. If a non-ICE legacy solution is needed, then this is 635 probably the best choice. 637 Once initial consent is verified, we also need to verify continuing 638 consent, in order to avoid attacks where two people briefly share an 639 IP (e.g., behind a NAT in an Internet cafe) and the attacker arranges 640 for a large, unstoppable, traffic flow to the network and then 641 leaves. The appropriate technologies here are fairly similar to 642 those for initial consent, though are perhaps weaker since the 643 threats is less severe. 645 4.2.4. IP Location Privacy 647 Note that as soon as the callee sends their ICE candidates, the 648 caller learns the callee's IP addresses. The callee's server 649 reflexive address reveals a lot of information about the callee's 650 location. In order to avoid tracking, implementations may wish to 651 suppress the start of ICE negotiation until the callee has answered. 652 In addition, either side may wish to hide their location entirely by 653 forcing all traffic through a TURN server. 655 In ordinary operation, the site learns the browser's IP address, 656 though it may be hidden via mechanisms like Tor 657 [http://www.torproject.org] or a VPN. However, because sites can 658 cause the browser to provide IP addresses, this provides a mechanism 659 for sites to learn about the user's network environment even if the 660 user is behind a VPN that masks their IP address. Implementations 661 wish to provide settings which suppress all non-VPN candidates if the 662 user is on certain kinds of VPN, especially privacy-oriented systems 663 such as Tor. 665 4.3. Communications Security 667 Finally, we consider a problem familiar from the SIP world: 668 communications security. For obvious reasons, it MUST be possible 669 for the communicating parties to establish a channel which is secure 670 against both message recovery and message modification. (See 671 [RFC5479] for more details.) This service must be provided for both 672 data and voice/video. Ideally the same security mechanisms would be 673 used for both types of content. Technology for providing this 674 service (for instance, SRTP [RFC3711], DTLS [RFC4347] and DTLS-SRTP 675 [RFC5763]) is well understood. However, we must examine this 676 technology to the WebRTC context, where the threat model is somewhat 677 different. 679 In general, it is important to understand that unlike a conventional 680 SIP proxy, the calling service (i.e., the Web server) controls not 681 only the channel between the communicating endpoints but also the 682 application running on the user's browser. While in principle it is 683 possible for the browser to cut the calling service out of the loop 684 and directly present trusted information (and perhaps get consent), 685 practice in modern browsers is to avoid this whenever possible. "In- 686 flow" modal dialogs which require the user to consent to specific 687 actions are particularly disfavored as human factors research 688 indicates that unless they are made extremely invasive, users simply 689 agree to them without actually consciously giving consent. 690 [abarth-rtcweb]. Thus, nearly all the UI will necessarily be 691 rendered by the browser but under control of the calling service. 692 This likely includes the peer's identity information, which, after 693 all, is only meaningful in the context of some calling service. 695 This limitation does not mean that preventing attack by the calling 696 service is completely hopeless. However, we need to distinguish 697 between two classes of attack: 699 Retrospective compromise of calling service. 700 The calling service is is non-malicious during a call but 701 subsequently is compromised and wishes to attack an older call 702 (often called a "passive attack") 704 During-call attack by calling service. 705 The calling service is compromised during the call it wishes to 706 attack (often called an "active attack"). 708 Providing security against the former type of attack is practical 709 using the techniques discussed in Section 4.3.1. However, it is 710 extremely difficult to prevent a trusted but malicious calling 711 service from actively attacking a user's calls, either by mounting a 712 MITM attack or by diverting them entirely. (Note that this attack 713 applies equally to a network attacker if communications to the 714 calling service are not secured.) We discuss some potential 715 approaches and why they are likely to be impractical in 716 Section 4.3.2. 718 4.3.1. Protecting Against Retrospective Compromise 720 In a retrospective attack, the calling service was uncompromised 721 during the call, but that an attacker subsequently wants to recover 722 the content of the call. We assume that the attacker has access to 723 the protected media stream as well as having full control of the 724 calling service. 726 If the calling service has access to the traffic keying material (as 727 in SDES [RFC4568]), then retrospective attack is trivial. This form 728 of attack is particularly serious in the Web context because it is 729 standard practice in Web services to run extensive logging and 730 monitoring. Thus, it is highly likely that if the traffic key is 731 part of any HTTP request it will be logged somewhere and thus subject 732 to subsequent compromise. It is this consideration that makes an 733 automatic, public key-based key exchange mechanism imperative for 734 WebRTC (this is a good idea for any communications security system) 735 and this mechanism SHOULD provide perfect forward secrecy (PFS). The 736 signaling channel/calling service can be used to authenticate this 737 mechanism. 739 In addition, if end-to-end keying is in used, the system MUST NOT 740 provide any APIs to extract either long-term keying material or to 741 directly access any stored traffic keys. Otherwise, an attacker who 742 subsequently compromised the calling service might be able to use 743 those APIs to recover the traffic keys and thus compromise the 744 traffic. 746 4.3.2. Protecting Against During-Call Attack 748 Protecting against attacks during a call is a more difficult 749 proposition. Even if the calling service cannot directly access 750 keying material (as recommended in the previous section), it can 751 simply mount a man-in-the-middle attack on the connection, telling 752 Alice that she is calling Bob and Bob that he is calling Alice, while 753 in fact the calling service is acting as a calling bridge and 754 capturing all the traffic. Protecting against this form of attack 755 requires positive authentication of the remote endpoint such as 756 explicit out-of-band key verification (e.g., by a fingerprint) or a 757 third-party identity service as described in 758 [I-D.ietf-rtcweb-security-arch]. 760 4.3.2.1. Key Continuity 762 One natural approach is to use "key continuity". While a malicious 763 calling service can present any identity it chooses to the user, it 764 cannot produce a private key that maps to a given public key. Thus, 765 it is possible for the browser to note a given user's public key and 766 generate an alarm whenever that user's key changes. SSH [RFC4251] 767 uses a similar technique. (Note that the need to avoid explicit user 768 consent on every call precludes the browser requiring an immediate 769 manual check of the peer's key). 771 Unfortunately, this sort of key continuity mechanism is far less 772 useful in the WebRTC context. First, much of the virtue of WebRTC 773 (and any Web application) is that it is not bound to particular piece 774 of client software. Thus, it will be not only possible but routine 775 for a user to use multiple browsers on different computers which will 776 of course have different keying material (SACRED [RFC3760] 777 notwithstanding.) Thus, users will frequently be alerted to key 778 mismatches which are in fact completely legitimate, with the result 779 that they are trained to simply click through them. As it is known 780 that users routinely will click through far more dire warnings 781 [cranor-wolf], it seems extremely unlikely that any key continuity 782 mechanism will be effective rather than simply annoying. 784 Moreover, it is trivial to bypass even this kind of mechanism. 785 Recall that unlike the case of SSH, the browser never directly gets 786 the peer's identity from the user. Rather, it is provided by the 787 calling service. Even enabling a mechanism of this type would 788 require an API to allow the calling service to tell the browser "this 789 is a call to user X". All the calling service needs to do to avoid 790 triggering a key continuity warning is to tell the browser that "this 791 is a call to user Y" where Y is close to X. Even if the user actually 792 checks the other side's name (which all available evidence indicates 793 is unlikely), this would require (a) the browser to trusted UI to 794 provide the name and (b) the user to not be fooled by similar 795 appearing names. 797 4.3.2.2. Short Authentication Strings 799 ZRTP [RFC6189] uses a "short authentication string" (SAS) which is 800 derived from the key agreement protocol. This SAS is designed to be 801 compared by the users (e.g., read aloud over the the voice channel or 802 transmitted via an out of band channel) and if confirmed by both 803 sides precludes MITM attack. The intention is that the SAS is used 804 once and then key continuity (though a different mechanism from that 805 discussed above) is used thereafter. 807 Unfortunately, the SAS does not offer a practical solution to the 808 problem of a compromised calling service. "Voice conversion" 809 systems, which modify voice from one speaker to make it sound like 810 another, are an active area of research. These systems are already 811 good enough to fool both automatic recognition systems 812 [farus-conversion] and humans [kain-conversion] in many cases, and 813 are of course likely to improve in future, especially in an 814 environment where the user just wants to get on with the phone call. 815 Thus, even if SAS is effective today, it is likely not to be so for 816 much longer. 818 Additionally, it is unclear that users will actually use an SAS. As 819 discussed above, the browser UI constraints preclude requiring the 820 SAS exchange prior to completing the call and so it must be 821 voluntary; at most the browser will provide some UI indicator that 822 the SAS has not yet been checked. However, it it is well-known that 823 when faced with optional security mechanisms, many users simply 824 ignore them [whitten-johnny]. 826 Once uses have checked the SAS once, key continuity is required to 827 avoid them needing to check it on every call. However, this is 828 problematic for reasons indicated in Section 4.3.2.1. In principle 829 it is of course possible to render a different UI element to indicate 830 that calls are using an unauthenticated set of keying material 831 (recall that the attacker can just present a slightly different name 832 so that the attack shows the same UI as a call to a new device or to 833 someone you haven't called before) but as a practical matter, users 834 simply ignore such indicators even in the rather more dire case of 835 mixed content warnings. 837 4.3.2.3. Third Party Identity 839 The conventional approach to providing communications identity has of 840 course been to have some third party identity system (e.g., PKI) to 841 authenticate the endpoints. Such mechanisms have proven to be too 842 cumbersome for use by typical users (and nearly too cumbersome for 843 administrators). However, a new generation of Web-based identity 844 providers (BrowserID, Federated Google Login, Facebook Connect, 845 OAuth, OpenID, WebFinger), has recently been developed and use Web 846 technologies to provide lightweight (from the user's perspective) 847 third-party authenticated transactions. It is possible to use 848 systems of this type to authenticate WebRTC calls, linking them to 849 existing user notions of identity (e.g., Facebook adjacencies). 850 Specifically, the third-party identity system is used to bind the 851 user's identity to cryptographic keying material which is then used 852 to authenticate the calling endpoints. Calls which are authenticated 853 in this fashion are naturally resistant even to active MITM attack by 854 the calling site. 856 Note that there is one special case in which PKI-style certificates 857 do provide a practical solution: calls from end-users to large 858 sites. For instance, if you are making a call to Amazon.com, then 859 Amazon can easily get a certificate to authenticate their media 860 traffic, just as they get one to authenticate their Web traffic. 861 This does not provide additional security value in cases in which the 862 calling site and the media peer are one in the same, but might be 863 useful in cases in which third parties (e.g., ad networks or 864 retailers) arrange for calls but do not participate in them. 866 4.3.2.4. Page Access to Media 868 Identifying the identity of the far media endpoint is a necessary but 869 not sufficient condition for providing media security. In WebRTC, 870 media flows are rendered into HTML5 MediaStreams which can be 871 manipulated by the calling site. Obviously, if the site can modify 872 or view the media, then the user is not getting the level of 873 assurance they would expect from being able to authenticate their 874 peer. In many cases, this is acceptable because the user values 875 site-based special effects over complete security from the site. 876 However, there are also cases where users wish to know that the site 877 cannot interfere. In order to facilitate that, it will be necessary 878 to provide features whereby the site can verifiably give up access to 879 the media streams. This verification must be possible both from the 880 local side and the remote side. I.e., I must be able to verify that 881 the person I am calling has engaged a secure media mode. In order to 882 achieve this it will be necessary to cryptographically bind an 883 indication of the local media access policy into the cryptographic 884 authentication procedures detailed in the previous sections. 886 4.3.3. Malicious Peers 888 One class of attack that we do not generally try to prevent is 889 malicious peers. For instance, no matter what confidentiality 890 measures you employ the person you are talking to might record the 891 call and publish it on the Internet. Similarly, we do not attempt to 892 prevent them from using voice or video processing technology from 893 hiding or changing their appearance. While technologies (DRM, etc.) 894 do exist to attempt to address these issues, they are generally not 895 compatible with open systems and WebRTC does not address them. 897 Similarly, we make no attempt to prevent prank calling or other 898 unwanted calls. In general, this is in the scope of the calling 899 site, though because WebRTC does offer some forms of strong 900 authentication, that may be useful as part of a defense against such 901 attacks. 903 4.4. Privacy Considerations 905 4.4.1. Correlation of Anonymous Calls 907 While persistent endpoint identifiers can be a useful security 908 feature (see Section 4.3.2.1 they can also represent a privacy threat 909 in settings where the user wishes to be anonymous. WebRTC provides a 910 number of possible persistent identifiers such as DTLS certificates 911 (if they are reused between connections) and RTCP CNAMES (if 912 generated according to [RFC6222] rather than the privacy preserving 913 mode of [I-D.ietf-avtcore-6222bis]). In order to prevent this type 914 of correlation, browsers need to provide mechanisms to reset these 915 identifiers (e.g., with the same lifetime as cookies). Moreover, the 916 API should provide mechanisms to allow sites intended for anonymous 917 calling to force the minting of fresh identifiers. 919 4.4.2. Browser Fingerprinting 921 Any new set of API features adds a risk of browser fingerprinting, 922 and WebRTC is no exception. Specifically, sites can use the presence 923 or absence of specific devices as a browser fingerprint. In general, 924 the API needs to be balanced between functionality and the 925 incremental fingerprint risk. 927 5. Security Considerations 929 This entire document is about security. 931 6. Acknowledgements 933 Bernard Aboba, Harald Alvestrand, Dan Druta, Cullen Jennings, Alan 934 Johnston, Hadriel Kaplan (S 4.2.1), Matthew Kaufman, Martin Thomson, 935 Magnus Westerland. 937 7. Changes Since -04 939 o Replaced RTCWEB and RTC-Web with WebRTC, except when referring to 940 the IETF WG 941 o Removed discussion of the IFRAMEd advertisement case, since we 942 decided not to treat it specially. 943 o Added a privacy section considerations section. 944 o Significant edits to the SAS section to reflect Alan Johnston's 945 comments. 946 o Added some discussion if IP location privacy and Tor. 947 o Updated the "communications consent" section to reflrect draft- 948 muthu. 949 o Added a section about "malicious peers". 950 o Added a section describing screen sharing threats. 951 o Assorted editorial changes. 953 8. References 955 8.1. Normative References 957 [I-D.ietf-rtcweb-overview] 958 Alvestrand, H., "Overview: Real Time Protocols for Brower- 959 based Applications", draft-ietf-rtcweb-overview-08 (work 960 in progress), September 2013. 962 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 963 Requirement Levels", BCP 14, RFC 2119, March 1997. 965 8.2. Informative References 967 [CORS] van Kesteren, A., "Cross-Origin Resource Sharing". 969 [I-D.ietf-avtcore-6222bis] 970 Begen, A., Perkins, C., Wing, D., and E. Rescorla, 971 "Guidelines for Choosing RTP Control Protocol (RTCP) 972 Canonical Names (CNAMEs)", draft-ietf-avtcore-6222bis-06 973 (work in progress), July 2013. 975 [I-D.ietf-rtcweb-security-arch] 976 Rescorla, E., "WebRTC Security Architecture", 977 draft-ietf-rtcweb-security-arch-07 (work in progress), 978 July 2013. 980 [I-D.kaufman-rtcweb-security-ui] 981 Kaufman, M., "Client Security User Interface Requirements 982 for RTCWEB", draft-kaufman-rtcweb-security-ui-00 (work in 983 progress), June 2011. 985 [I-D.muthu-behave-consent-freshness] 986 Perumal, M., Wing, D., R, R., and T. Reddy, "STUN Usage 987 for Consent Freshness", 988 draft-muthu-behave-consent-freshness-04 (work in 989 progress), July 2013. 991 [RFC2818] Rescorla, E., "HTTP Over TLS", RFC 2818, May 2000. 993 [RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, 994 A., Peterson, J., Sparks, R., Handley, M., and E. 995 Schooler, "SIP: Session Initiation Protocol", RFC 3261, 996 June 2002. 998 [RFC3552] Rescorla, E. and B. Korver, "Guidelines for Writing RFC 999 Text on Security Considerations", BCP 72, RFC 3552, 1000 July 2003. 1002 [RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K. 1003 Norrman, "The Secure Real-time Transport Protocol (SRTP)", 1004 RFC 3711, March 2004. 1006 [RFC3760] Gustafson, D., Just, M., and M. Nystrom, "Securely 1007 Available Credentials (SACRED) - Credential Server 1008 Framework", RFC 3760, April 2004. 1010 [RFC4251] Ylonen, T. and C. Lonvick, "The Secure Shell (SSH) 1011 Protocol Architecture", RFC 4251, January 2006. 1013 [RFC4347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer 1014 Security", RFC 4347, April 2006. 1016 [RFC4568] Andreasen, F., Baugher, M., and D. Wing, "Session 1017 Description Protocol (SDP) Security Descriptions for Media 1018 Streams", RFC 4568, July 2006. 1020 [RFC5245] Rosenberg, J., "Interactive Connectivity Establishment 1021 (ICE): A Protocol for Network Address Translator (NAT) 1022 Traversal for Offer/Answer Protocols", RFC 5245, 1023 April 2010. 1025 [RFC5479] Wing, D., Fries, S., Tschofenig, H., and F. Audet, 1026 "Requirements and Analysis of Media Security Management 1027 Protocols", RFC 5479, April 2009. 1029 [RFC5763] Fischl, J., Tschofenig, H., and E. Rescorla, "Framework 1030 for Establishing a Secure Real-time Transport Protocol 1031 (SRTP) Security Context Using Datagram Transport Layer 1032 Security (DTLS)", RFC 5763, May 2010. 1034 [RFC6189] Zimmermann, P., Johnston, A., and J. Callas, "ZRTP: Media 1035 Path Key Agreement for Unicast Secure RTP", RFC 6189, 1036 April 2011. 1038 [RFC6222] Begen, A., Perkins, C., and D. Wing, "Guidelines for 1039 Choosing RTP Control Protocol (RTCP) Canonical Names 1040 (CNAMEs)", RFC 6222, April 2011. 1042 [RFC6454] Barth, A., "The Web Origin Concept", RFC 6454, 1043 December 2011. 1045 [RFC6455] Fette, I. and A. Melnikov, "The WebSocket Protocol", 1046 RFC 6455, December 2011. 1048 [SWF] Adobe, "SWF File Format Specification Version 19". 1050 [abarth-rtcweb] 1051 Barth, A., "Prompting the user is security failure", RTC- 1052 Web Workshop. 1054 [cranor-wolf] 1055 Sunshine, J., Egelman, S., Almuhimedi, H., Atri, N., and 1056 L. cranor, "Crying Wolf: An Empirical Study of SSL Warning 1057 Effectiveness", Proceedings of the 18th USENIX Security 1058 Symposium, 2009. 1060 [farus-conversion] 1061 Farrus, M., Erro, D., and J. Hernando, "Speaker 1062 Recognition Robustness to Voice Conversion". 1064 [finer-grained] 1065 Barth, A. and C. Jackson, "Beware of Finer-Grained 1066 Origins", W2SP, 2008. 1068 [huang-w2sp] 1069 Huang, L-S., Chen, E., Barth, A., Rescorla, E., and C. 1070 Jackson, "Talking to Yourself for Fun and Profit", W2SP, 1071 2011. 1073 [kain-conversion] 1074 Kain, A. and M. Macon, "Design and Evaluation of a Voice 1075 Conversion Algorithm based on Spectral Envelope Mapping 1076 and Residual Prediction", Proceedings of ICASSP, May 1077 2001. 1079 [whitten-johnny] 1080 Whitten, A. and J. Tygar, "Why Johnny Can't Encrypt: A 1081 Usability Evaluation of PGP 5.0", Proceedings of the 8th 1082 USENIX Security Symposium, 1999. 1084 Author's Address 1086 Eric Rescorla 1087 RTFM, Inc. 1088 2064 Edgewood Drive 1089 Palo Alto, CA 94303 1090 USA 1092 Phone: +1 650 678 2350 1093 Email: ekr@rtfm.com