idnits 2.17.1 draft-mm-wg-effect-encrypt-12.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 a both a reference to RFC 2119 and the recommended RFC 2119 boilerplate, even if it appears to use RFC 2119 keywords. RFC 2119 keyword, line 770: '... SHOULD be turned off [RFC4787]....' Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (June 30, 2017) is 2485 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- == Missing Reference: 'RFC 1984' is mentioned on line 142, but not defined == Missing Reference: 'RFC 2804' is mentioned on line 142, but not defined == Missing Reference: 'RFC 2775' is mentioned on line 142, but not defined == Missing Reference: 'RFC 3724' is mentioned on line 142, but not defined == Missing Reference: 'RFC 7754' is mentioned on line 142, but not defined == Unused Reference: 'EFF' is defined on line 1823, but no explicit reference was found in the text == Unused Reference: 'RFC2775' is defined on line 1886, but no explicit reference was found in the text == Unused Reference: 'RFC2804' is defined on line 1890, but no explicit reference was found in the text == Outdated reference: A later version (-11) exists of draft-ietf-mile-iodef-guidance-10 -- Obsolete informational reference (is this intentional?): RFC 3315 (Obsoleted by RFC 8415) -- Obsolete informational reference (is this intentional?): RFC 7230 (Obsoleted by RFC 9110, RFC 9112) -- Obsolete informational reference (is this intentional?): RFC 7234 (Obsoleted by RFC 9111) -- Obsolete informational reference (is this intentional?): RFC 7525 (Obsoleted by RFC 9325) -- Obsolete informational reference (is this intentional?): RFC 7540 (Obsoleted by RFC 9113) Summary: 1 error (**), 0 flaws (~~), 10 warnings (==), 6 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group K. Moriarty 3 Internet-Draft Dell EMC 4 Intended status: Informational A. Morton 5 Expires: January 1, 2018 AT&T Labs 6 June 30, 2017 8 Effect of Pervasive Encryption on Operators 9 draft-mm-wg-effect-encrypt-12 11 Abstract 13 Pervasive Monitoring (PM) attacks on the privacy of Internet users is 14 of serious concern to both the user and the operator communities. 15 RFC7258 discussed the critical need to protect users' privacy when 16 developing IETF specifications and also recognized making networks 17 unmanageable to mitigate PM is not an acceptable outcome, an 18 appropriate balance is needed. This document discusses current 19 security and network management practices that may be impacted by the 20 shift to increased use of encryption to help guide protocol 21 development in support of manageable, secure networks. 23 Status of This Memo 25 This Internet-Draft is submitted in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF). Note that other groups may also distribute 30 working documents as Internet-Drafts. The list of current Internet- 31 Drafts is at http://datatracker.ietf.org/drafts/current/. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 This Internet-Draft will expire on January 1, 2018. 40 Copyright Notice 42 Copyright (c) 2017 IETF Trust and the persons identified as the 43 document authors. All rights reserved. 45 This document is subject to BCP 78 and the IETF Trust's Legal 46 Provisions Relating to IETF Documents 47 (http://trustee.ietf.org/license-info) in effect on the date of 48 publication of this document. Please review these documents 49 carefully, as they describe your rights and restrictions with respect 50 to this document. Code Components extracted from this document must 51 include Simplified BSD License text as described in Section 4.e of 52 the Trust Legal Provisions and are provided without warranty as 53 described in the Simplified BSD License. 55 Table of Contents 57 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 58 1.1. Additional Background on Encryption Changes . . . . . . . 4 59 2. Network Service Provider Monitoring . . . . . . . . . . . . . 6 60 2.1. Passive Monitoring . . . . . . . . . . . . . . . . . . . 7 61 2.1.1. Traffic Surveys . . . . . . . . . . . . . . . . . . . 7 62 2.1.2. Troubleshooting . . . . . . . . . . . . . . . . . . . 7 63 2.1.3. Traffic Analysis Fingerprinting . . . . . . . . . . . 9 64 2.2. Traffic Optimization and Management . . . . . . . . . . . 10 65 2.2.1. Load Balancers . . . . . . . . . . . . . . . . . . . 10 66 2.2.2. Differential Treatment based on Deep Packet 67 Inspection (DPI) . . . . . . . . . . . . . . . . . . 12 68 2.2.3. Network Congestion Management . . . . . . . . . . . . 13 69 2.2.4. Performance-enhancing Proxies . . . . . . . . . . . . 13 70 2.2.5. Caching and Content Replication Near the Network Edge 13 71 2.2.6. Content Compression . . . . . . . . . . . . . . . . . 14 72 2.3. Network Access and Accounting . . . . . . . . . . . . . . 14 73 2.3.1. Content Filtering . . . . . . . . . . . . . . . . . . 15 74 2.3.2. Network Access and Data Usage . . . . . . . . . . . . 15 75 2.3.3. Application Layer Gateways . . . . . . . . . . . . . 16 76 2.3.4. HTTP Header Insertion . . . . . . . . . . . . . . . . 17 77 3. Encryption in Hosting SP Environments . . . . . . . . . . . . 17 78 3.1. Management Access Security . . . . . . . . . . . . . . . 18 79 3.1.1. Customer Access Monitoring . . . . . . . . . . . . . 19 80 3.1.2. SP Content Monitoring of Applications . . . . . . . . 20 81 3.2. Hosted Applications . . . . . . . . . . . . . . . . . . . 21 82 3.2.1. Monitoring Managed Applications . . . . . . . . . . . 21 83 3.2.2. Mail Service Providers . . . . . . . . . . . . . . . 22 84 3.3. Data Storage . . . . . . . . . . . . . . . . . . . . . . 22 85 3.3.1. Host-level Encryption . . . . . . . . . . . . . . . . 23 86 3.3.2. Disk Encryption, Data at Rest . . . . . . . . . . . . 23 87 3.3.3. Cross Data Center Replication Services . . . . . . . 24 88 4. Encryption for Enterprises . . . . . . . . . . . . . . . . . 24 89 4.1. Monitoring Practices of the Enterprise . . . . . . . . . 25 90 4.1.1. Security Monitoring in the Enterprise . . . . . . . . 25 91 4.1.2. Application Performance Monitoring in the Enterprise 26 92 4.1.3. Enterprise Network Diagnostics and Troubleshooting . 27 93 4.2. Techniques for Monitoring Internet Session Traffic . . . 28 94 5. Security Monitoring for Specific Attack Types . . . . . . . . 30 95 5.1. Mail Abuse and SPAM . . . . . . . . . . . . . . . . . . . 30 96 5.2. Denial of Service . . . . . . . . . . . . . . . . . . . . 31 97 5.3. Phishing . . . . . . . . . . . . . . . . . . . . . . . . 31 98 5.4. Botnets . . . . . . . . . . . . . . . . . . . . . . . . . 32 99 5.5. Malware . . . . . . . . . . . . . . . . . . . . . . . . . 32 100 5.6. Spoofed Source IP Address Protection . . . . . . . . . . 32 101 5.7. Further work . . . . . . . . . . . . . . . . . . . . . . 33 102 6. Application-based Flow Information Visible to a Network . . . 33 103 6.1. TLS Server Name Indication . . . . . . . . . . . . . . . 33 104 6.2. Application Layer Protocol Negotiation (ALPN) . . . . . . 34 105 6.3. Content Length, BitRate and Pacing . . . . . . . . . . . 34 106 7. Impact on Mobility Network Optimizations and New Services . . 34 107 7.1. Effect of Encypted ACKs . . . . . . . . . . . . . . . . . 34 108 7.2. Effect of Encrypted Transport Headers . . . . . . . . . . 35 109 7.3. Effect of Encryption on New or Emerging Services . . . . 36 110 7.4. Effect of Encryption on Mobile Network Evolution . . . . 36 111 8. Response to Increased Encryption and Looking Forward . . . . 37 112 9. Security Considerations . . . . . . . . . . . . . . . . . . . 38 113 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 38 114 11. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 38 115 12. Informative References . . . . . . . . . . . . . . . . . . . 38 116 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 44 118 1. Introduction 120 In response to pervasive monitoring revelations and the IETF 121 consensus that Pervasive Monitoring is an Attack [RFC7258], efforts 122 are underway to increase encryption of Internet traffic. Pervasive 123 Monitoring (PM) is of serious concern to users, operators, and 124 application providers. RFC7258 discussed the critical need to 125 protect users' privacy when developing IETF specifications and also 126 recognized that making networks unmanageable to mitigate PM is not an 127 acceptable outcome, but rather that an appropriate balance would 128 emerge over time. 130 This document discusses practices currently used by network operators 131 to manage, operate, and secure their networks and how those practices 132 may be impacted by a shift to increased use of encryption. It 133 provides network operators' perspectives about the motivations and 134 objectives of those practices as well as effects anticipated by 135 operators as use of encryption increases. It is a summary of 136 concerns of the operational community as they transition to managing 137 networks with less visibility. The document does not endorse the use 138 of the practices described herein. Nor does it aim to provide a 139 comprehensive treatment of the effects of current practices, some of 140 which have been considered controversial from a technical or business 141 perspective or contradictory to previous IETF statements (e.g., [RFC 142 1958], [RFC 1984], [RFC 2804], [RFC 2775], [RFC 3724], [RFC 7754]). 144 This document aims to help IETF participants understand network 145 operators' perspectives about the impact of pervasive encryption, 146 both opportunistic and strong end-to-end encryption, on operational 147 practices. The goal is to help inform future protocol development to 148 ensure that operational impact is part of the conversation. Ideally, 149 new methods could be developed to accomplish the goals of current 150 practices despite changes in the extent to which cleartext will be 151 available to network operators. Discussion of current practices and 152 the potential future changes is provided as a prerequisite to 153 potential future cross-industry and cross-layer work to support the 154 ongoing evolution towards a functional Internet with pervasive 155 encryption. 157 Traditional network management, planning, security operations, and 158 performance optimization have been developed in an Internet where a 159 large majority of data traffic flows without encryption. While 160 unencrypted traffic has made information that aids operations and 161 troubleshooting at all layers accessible, it has also made pervasive 162 monitoring by unseen parties possible. With broad support and 163 increased awareness of the need to consider privacy in all aspects 164 across the Internet, it is important to catalog existing management, 165 operational, and security practices that have depended upon the 166 availability of cleartext to function. 168 This document includes a sampling of current practices and does not 169 attempt to describe every nuance. Some sections cover technologies 170 used over a broad spectrum of devices and use cases. 172 1.1. Additional Background on Encryption Changes 174 Session encryption helps to prevent both passive and active attacks 175 on transport protocols; more on pervasive monitoring can be found in 176 Confidentiality in the Face of Pervasive Surveillance: A Threat Model 177 and Problem Statement [RFC7624]. The Internet Architecture Board 178 (IAB) released a statement advocating for increased use of encryption 179 in November 2014. Perspectives on encryption paradigms have also 180 shifted. One such shift is documented in "Opportunistic Security" 181 (OS) [RFC7435], which suggests that when use of authenticated 182 encryption is not possible, cleartext sessions should be upgraded to 183 unauthenticated session encryption, rather than no encryption. OS 184 encourages upgrading from cleartext, but cannot require or guarantee 185 such upgrades. Once OS is used, it allows for an evolution to 186 authenticated encryption. These efforts are necessary to improve end 187 user's expectation of privacy, making pervasive monitoring cost 188 prohibitive. With OS in use, active attacks are still possible on 189 unauthenticated sessions. OS has been implemented as NULL 190 Authentication with IPsec [RFC7619] and there are a number of 191 infrastructure use cases such as server to server encryption, where 192 this mode is deployed. IPsec with authentication has many useful 193 applications and usage has increased for infrastructure applications 194 such as for virtual private networks between data centers. 196 Although there is a push for OS, there is also work being done to 197 improve the implementation, development and configuration of TLS and 198 DTLS sessions to prevent active attacks used to monitor or intercept 199 session data. The 201 (UTA) working group has been publishing documentation to improve the 202 security of TLS and DTLS sessions. They have documented the known 203 attack vectors in [RFC7457] and have documented Best Practices for 204 TLS and DTLS in [RFC7525] and have other documents in the queue. 206 In addition to encrypted web site access (HTTP over TLS), there are 207 other well-deployed application level transport encryption efforts 208 such as mail transfer agent (MTA)-to-MTA session encryption transport 209 for email (SMTP over TLS) and gateway-to-gateway for instant 210 messaging (Extensible Messaging and Presence Protocol (XMPP) over 211 TLS). Although this does provide protection from transport layer 212 attacks, the servers could be a point of vulnerability if user-to- 213 user encryption is not provided for these messaging protocols. User- 214 to-user content encryption schemes, such as S/MIME and PGP for email 215 and encryption (e.g. Off-the-Record (OTR)) for XMPP are used by 216 those interested to protect their data as it crosses intermediary 217 servers, preventing the vulnerability described by providing an end- 218 to-end solution. User-to-user schemes are under review and 219 additional options will emerge to ease the configuration 220 requirements, making this type of option more accessible to non- 221 technical users interested in protecting their privacy. 223 Increased use of encryption, either opportunistic or authenticated, 224 at the transport, network or application layer, impacts how networks 225 are operated, managed, and secured. In some cases, new methods to 226 operate, manage, and secure networks will evolve in response. In 227 other cases, currently available capabilities for monitoring or 228 troubleshooting networks could become unavailable. This document 229 lists a collection of functions currently employed by network 230 operators that may be impacted by the shift to increased use of 231 encryption. This draft does not attempt to specify responses or 232 solutions to these impacts, but rather documents the current state. 234 This document refers to several different forms of service providers, 235 distinguished with adjectives. For example, network service 236 providers (or network operators) provide IP-packet transport 237 primarily, though they may bundle other services with packet 238 transport. Alternatively, application service providers primarily 239 offer systems that participate as an end-point in communications with 240 the application user, and hosting service providers lease computing, 241 storage, and communications systems in datacenters. In practice, 242 many companies perform two or more service provider roles, but may be 243 historically associated with one. 245 2. Network Service Provider Monitoring 247 Network Service Providers (SP) for this definition include the 248 backbone Internet Service providers as well as those providing 249 infrastructure at scale for core Internet use (hosted infrastructure 250 and services such as email). 252 Following the Snowden revelations, application service providers 253 responded by encrypting traffic between their data centers (IPsec) to 254 prevent passive monitoring from taking place unbeknownst to them 255 (Yahoo, Google, etc.). Large mail service providers also began to 256 encrypt session transport (TLS) to hosted mail services. This and 257 other increases in the use of encryption had the immediate effect of 258 helping protect the privacy of users' data, but created a problem for 259 some network management functions. They could no longer gain access 260 to some session streams resulting in actions by several to regain 261 their operational practices that previously depended on cleartext 262 data sessions. 264 The EFF reported [EFF2014] several network service providers taking 265 steps to prevent the use of SMTP over TLS by breaking STARTTLS 266 (section 3.2 of [RFC7525]), essentially preventing the negotiation 267 process resulting in fallback to the use of clear text. In other 268 cases, some service providers have relied on middle boxes having 269 access to clear text for the purposes of load balancing, monitoring 270 for attack traffic, meeting regulatory requirements, or for other 271 purposes. These middle box implementations, whether performing 272 functions considered legitimate by the IETF or not, have been 273 impacted by increases in encrypted traffic. Only methods keeping 274 with the goal of balancing network management and PM mitigation in 275 [RFC7258] should be considered in solution work resulting from this 276 document. 278 Network service providers use various techniques to operate, manage, 279 and secure their networks. The following subsections detail the 280 purpose of each technique and which protocol fields are used to 281 accomplish each task. In response to increased encryption of these 282 fields, some network service providers may be tempted to undertake 283 undesirable security practices in order to gain access to the fields 284 in unencrypted data flows. To avoid this situation, ideally new 285 methods could be developed to accomplish the same goals without 286 service providers having the ability to see session data. 288 2.1. Passive Monitoring 290 2.1.1. Traffic Surveys 292 Internet traffic surveys are useful in many pursuits, such as input 293 for CAIDA studies [CAIDA], network planning and optimization. 294 Tracking the trends in Internet traffic growth, from earlier peer-to- 295 peer communication to the extensive adoption of unicast video 296 streaming applications, has relied on a view of traffic composition 297 with a particular level of assumed accuracy, based on access to 298 cleartext by those conducting the surveys. 300 Passive monitoring makes inferences about observed traffic using the 301 maximal information available, and is subject to inaccuracies 302 stemming from incomplete sampling (of packets in a stream) or loss 303 due to monitoring system overload. When encryption conceals more 304 layers in each packet, reliance on pattern inferences and other 305 heuristics grows, and accuracy suffers. For example, the traffic 306 patterns between server and browser are dependent on browser supplier 307 and version, even when the sessions use the same server application 308 (e.g., web e-mail access). It remains to be seen whether more 309 complex inferences can be mastered to produce the same monitoring 310 accuracy. 312 2.1.2. Troubleshooting 314 Network operators use packet captures and protocol-dissecting 315 analyzers when responding to customer problems, to identify the 316 presence of attack traffic, and to identify root causes of the 317 problem such as misconfiguration. The protocol dissection is 318 generally limited to supporting protocols (e.g., DNS, DHCP), network 319 and transport (e.g., IP, TCP), and some higher layer protocols (e.g., 320 RTP, RTCP). 322 Network operators are often the first ones called upon to investigate 323 application problems (e.g., "my HD video is choppy"). When 324 diagnosing a customer problem, the starting point may be a particular 325 application that isn't working. The ability to identify the problem 326 application's traffic is important and deep packet inspection (DPI) 327 is often used for this purpose; IP address filtering is not useful 328 for applications using CDNs or cloud providers. After identifying 329 the traffic, an operator may analyze the traffic characteristics and 330 routing of the traffic. 332 For example, by investigating packet loss (from TCP sequence and 333 acknowledgement numbers), round-trip-time (from TCP timestamp options 334 or application-layer transactions, e.g., DNS or HTTP response time), 335 TCP receive-window size, packet corruption (from checksum 336 verification), inefficient fragmentation, or application-layer 337 problems, the operator can narrow the problem to a portion of the 338 network, server overload, client or server misconfiguration, etc. 339 Network operators may also be able to identify the presence of attack 340 traffic as not conforming to the application the user claims to be 341 using. 343 One way of quickly excluding the network as the bottleneck during 344 troubleshooting is to check whether the speed is limited by the 345 endpoints. For example, the connection speed might instead be 346 limited by suboptimal TCP options, the sender's congestion window, 347 the sender temporarily running out of data to send, the sender 348 waiting for the receiver to send another request, or the receiver 349 closing the receive window. All this information can be derived from 350 the cleartext TCP header. 352 Packet captures and protocol-dissecting analyzers have been important 353 tools. Automated monitoring has also been used to proactively 354 identify poor network conditions, leading to maintenance and network 355 upgrades before user experience declines. For example, findings of 356 loss and jitter in VoIP traffic can be a predictor of future customer 357 dissatisfaction (supported by metadata from RTP/RTCP protocol 358 )[RFC3550], or increases in DNS response time can generally make 359 interactive web browsing appear sluggish. But to detect such 360 problems, the application or service stream must first be 361 distinguished from others. 363 When using increased encryption, operators lose a source of data that 364 may be used to debug user issues. Because of this, application 365 server operators using increased encryption should expect to be 366 called upon more frequently to assist with debugging and 367 troubleshooting, and thus may want to consider what tools can be put 368 in the hands of their clients or network operators. 370 Further, the performance of some services can be more efficiently 371 managed and repaired when information on user transactions is 372 available to the service provider. It may be possible to continue 373 such monitoring activities without clear text access to the 374 application layers of interest, but inaccuracy will increase and 375 efficiency of repair activities will decrease. For example, an 376 application protocol error or failure would be opaque to network 377 troubleshooters when transport encryption is applied, making root 378 cause location more difficult and therefore increasing the time-to- 379 repair. Repair time directly reduces the availability of the 380 service, and most network operators have made availability a key 381 metric in their Service Level Agreements and/or subscription rebates. 382 Also, there may be more cases of user communication failures when the 383 additional encryption processes are introduced (e.g., key management 384 at large scale), leading to more customer service contacts and (at 385 the same time) less information available to network operations 386 repair teams. 388 In mobile networks, knowledge about TCP's stream transfer progress 389 (by observing ACKs, retransmissions, packet drops, and the Sector 390 Utilization Level etc.) is further used to measure the performance of 391 Network Segments (Sector, eNodeB (eNB) etc.). This information is 392 used as Key Performance Indicators (KPIs) and for the estimation of 393 User/Service Key Quality Indicators at network edges for circuit 394 emulation (CEM) as well as input for mitigation methods. If the 395 make-up of active services per user and per sector are not visible to 396 a server that provides Internet Access Point Names (APN), it cannot 397 perform mitigation functions based on network segment view. 399 It is important to note that the push for encryption by application 400 providers has been motivated by the application of the described 401 techniques. Some application providers have noted degraded 402 performance and/or user experience when network-based optimization or 403 enhancement of their traffic has occurred, and such cases may result 404 in additional operator troubleshooting, as well. 406 2.1.3. Traffic Analysis Fingerprinting 408 Fingerprinting is used in traffic analysis and monitoring to identify 409 traffic streams that match certain patterns. This technique is 410 sometimes used with clear text or encrypted sessions. Some 411 Distributed Denial of Service (DDoS) prevention techniques at the 412 network provider level rely on the ability to fingerprint traffic in 413 order to mitigate the effect of this type of attack. Thus, 414 fingerprinting may be an aspect of an attack or part of attack 415 countermeasures. 417 A common, early trigger for DDoS mitigation includes observing 418 uncharacteristic traffic volumes or sources; congestion; or 419 degradation of a given network or service. One approach to mitigate 420 such an attack involves distinguishing attacker traffic from 421 legitimate user traffic. The ability to examine layers and payloads 422 above transport provides a new range of filtering opportunities at 423 each layer in the clear. If fewer layers are in the clear, this 424 means that there are reduced filtering opportunities available to 425 mitigate attacks. However, fingerprinting is still possible. 427 Passive monitoring of network traffic can lead to invasion of privacy 428 by external actors at the endpoints of the monitored traffic. 429 Encryption of traffic end-to-end is one method to obfuscate some of 430 the potentially identifying information. Many DoS mitigation systems 431 perform this manner of passive monitoring. 433 For example, browser fingerprints are comprised of many 434 characteristics, including User Agent, HTTP Accept headers, browser 435 plug-in details, screen size and color details, system fonts and time 436 zone. A monitoring system could easily identify a specific browser, 437 and by correlating other information, identify a specific user. 439 2.2. Traffic Optimization and Management 441 2.2.1. Load Balancers 443 A standalone load balancer is a function one can take off the shelf, 444 place in front of a pool of servers, configure appropriately, and it 445 will balance the traffic load among servers in the pool. This is a 446 typical setup for load balancers. Standalone load balancers rely on 447 the plainly observable information in the packets they are forwarding 448 and rely on industry-accepted standards in interpreting the plainly 449 observable information. Typically, this is a 5-tuple of the 450 connection. This configuration terminates TLS sessions at the load 451 balancer, making it the end point instead of the server. Standalone 452 load balancers are considered middleboxes, but are an integral part 453 of server infrastructure that scales. 455 In contrast, an integrated load balancer is developed to be an 456 integral part of the service provided by the server pool behind that 457 load balancer. These load balancers can communicate state with their 458 pool of servers to better route flows to the appropriate servers. 459 They rely on non-standard system-specific information and operational 460 knowledge shared between the load balancer and its servers. 462 Both standalone and integrated load balancers can be deployed in 463 pools for redundancy and load sharing. For high availability, it is 464 important that when packets belonging to a flow start to arrive at a 465 different load balancer in the load balancer pool, the packets 466 continue to be forwarded to the original server in the server pool. 467 The importance of this requirement increases as the chances of such 468 load balancer change event increases. 470 Mobile operators deploy integrated load balancers to assist with 471 maintaining connection state as devices migrate. With the 472 proliferation of mobile connected devices, there is an acute need for 473 connection-oriented protocols that maintain connections after a 474 network migration by an endpoint. This connection persistence 475 provides an additional challenge for multi-homed anycast-based 476 services typically employed by large content owners and Content 477 Distribution Networks (CDNs). The challenge is that a migration to a 478 different network in the middle of the connection greatly increases 479 the chances of the packets routed to a different anycast point-of- 480 presence (POP) due to the new network's different connectivity and 481 Internet peering arrangements. The load balancer in the new POP, 482 potentially thousands of miles away, will not have information about 483 the new flow and would not be able to route it back to the original 484 POP. 486 To help with the endpoint network migration challenges, anycast 487 service operations are likely to employ integrated load balancers 488 that, in cooperation with their pool servers, are able to ensure that 489 client-to-server packets contain some additional identification in 490 plainly-observable parts of the packets (in addition to the 5-tuple). 491 As noted in Section 2 of [RFC7258], careful consideration in protocol 492 design to mitigate PM is important, while ensuring manageability of 493 the network. 495 Some integrated load balancers have the ability to use additional 496 plainly observable information even for today's protocols that are 497 not network migration tolerant. This additional information allows 498 for improved availability and scaleability of the load balancing 499 operation. For example, BGP reconvergence can cause a flow to switch 500 anycast POPs even without a network change by any endpoint. 501 Additionally, a system that is able to encode the identity of the 502 pool server in plain text information available in each incoming 503 packet is able to provide stateless load balancing. This ability 504 confers great reliability and scaleability advantages even if the 505 flow remains in a single POP, because the load balancing system is 506 not required to keep state of each flow. Even more importantly, 507 there's no requirement to continuously synchronize such state among 508 the pool of load balancers. 510 Current protocols, such as TCP, allow the development of stateless 511 integrated load balancers by availing such load balancers of 512 additional plain text information in client-to-server packets. In 513 case of TCP, such information can be encoded by having server- 514 generated sequence numbers (that are ACK'd by the client), segment 515 values, lengths of the packet sent, etc. 517 Mobile operators apply Self Organizing Networks (3GPP SON) for 518 intelligent workflows such as content-aware MLB (Mobility Load 519 Balancing). Where network load balancers have been configured to 520 route according to application-layer semantics, an encrypted payload 521 is effectively invisible. This has resulted in practices of 522 intercepting TLS in front of load balancers to regain that 523 visibility, but at a cost to security and privacy. 525 In future Network Function Virtualization (NFV) architectures, load 526 balancing functions are likely to be more prevalent (deployed at 527 locations throughout operators' networks), so they would be handling 528 traffic using encrypted tunnels whenever it is present. 530 2.2.2. Differential Treatment based on Deep Packet Inspection (DPI) 532 Data transfer capacity resources in cellular radio networks tend to 533 be more constrained than in fixed networks. This is a result of 534 variance in radio signal strength as a user moves around a cell, the 535 rapid ingress and egress of connections as users hand off between 536 adjacent cells, and temporary congestion at a cell. Mobile networks 537 alleviate this by queuing traffic according to its required bandwidth 538 and acceptable latency: for example, a user is unlikely to notice a 539 20ms delay when receiving a simple Web page or email, or an instant 540 message response, but will very likely notice a re-buffering pause in 541 a video playback or a VoIP call de-jitter buffer. Ideally, the 542 scheduler manages the queue so that each user has an acceptable 543 experience as conditions vary, but inferences of the traffic type 544 have been used to make bearer assignments and set scheduler priority. 546 Deep Packet Inspection (DPI) allows identification of applications 547 based on payload signatures, in contrast to trusting well-known port 548 numbers. Application and transport layer encryption make the traffic 549 type estimation more complex and less accurate, and therefore it may 550 not be effectual to use this information as input for queue 551 management. With the use of WebSockets [RFC6455], for example, many 552 forms of communications (from isochronous/real-time to bulk/elastic 553 file transfer) will take place over HTTP port 80 or port 443, so only 554 the messages and higher-layer data will make application 555 differentiation possible. If the monitoring system sees only "HTTP 556 port 443", it cannot distinguish application streams that would 557 benefit from priority queueing from others that would not. 559 Mobile networks especially rely on content/application based 560 prioritization of Over-the-Top (OTT) services - each application-type 561 or service has different delay/loss/throughput expectations, and each 562 type of stream will be unknown to an edge device if encrypted; this 563 impedes dynamic-QoS adaptation. An alternate way to achieve 564 encrypted application separation is possible when the User Equipment 565 (UE) requests a dedicated bearer for the specific application stream 566 (known by the UE), using a mechanism such as the one described in 567 Section 6.5 of 3GPP TS 24.301 [TS3GPP]. The UE's request includes 568 the Quality Class Indicator (QCI) appropriate for each application, 569 based on their different delay/loss/throughput expectations. 570 However, UE requests for dedicated bearers and QCI may not be 571 supported at the subscriber's service level, or in all mobile 572 networks. 574 These effects and potential alternative solutions have been discussed 575 at the accord BoF [ACCORD] at IETF95. 577 2.2.3. Network Congestion Management 579 For User Plane Congestion Management (3GPP UPCON) - ability to 580 understand content and manage network during congestion. Mitigating 581 techniques such as deferred download, off-peak acceleration, and 582 outbound roamers. 584 2.2.4. Performance-enhancing Proxies 586 Due to the characteristics of the mobile link, performance-enhancing 587 TCP proxies may perform local retransmission at the mobile edge. In 588 TCP, duplicated ACKs are detected and potentially concealed when the 589 proxy retransmits a segment that was lost on the mobile link without 590 involvement of the far end (see section 2.1.1 of [RFC3135] and 591 section 3.5 of [I-D.dolson-plus-middlebox-benefits]). 593 This optimization at network edges measurably improves real-time 594 transmission over long delay Internet paths or networks with large 595 capacity-variation (such as mobile/cellular networks). 597 In general, performance-enhancing proxies have a lower Round-Trip 598 Time (RTT) to the client and therefore determine the responsiveness 599 of flow control. A lower RTT makes the flow control loop more 600 responsive to changing in the mobile network conditions and enables 601 faster adaptation in a delay and capacity varying network due to user 602 mobility. 604 Further, service-provider-operated proxies are used to reduce the 605 control delay between the sender and a receiver on a mobile network 606 where resources are limited. The RTT determines how quickly an 607 user's attempt to cancel a video is recognized and therefore how 608 quickly the traffic is stopped, thus keeping un-wanted video packets 609 from entering the radio scheduler queue. 611 An application-type-aware network edge (middlebox) can further 612 control pacing, limit simultaneous HD videos, or prioritize active 613 videos against new videos, etc. 615 2.2.5. Caching and Content Replication Near the Network Edge 617 The features and efficiency of some Internet services can be 618 augmented through analysis of user flows and the applications they 619 provide. For example, network caching of popular content at a 620 location close to the requesting user can improve delivery efficiency 621 (both in terms of lower request response times and reduced use of 622 International Internet links when content is remotely located), and 623 authorized parties acting on their behalf use DPI in combination with 624 content distribution networks to determine if they can intervene 625 effectively. Web proxies are widely used [WebCache], and caching is 626 supported by the recent update of "Hypertext Transfer Protocol 627 (HTTP/1.1): Caching" in [RFC7234]. Encryption of packet contents at 628 a given protocol layer usually makes DPI processing of that layer and 629 higher layers impossible. That being said, it should be noted that 630 some content providers prevent caching to control content delivery 631 through the use of encrypted end-to-end sessions. CDNs vary in their 632 deployment options of end-to-end encryption. The business risk is a 633 motivation outside of privacy and pervasive monitoring that are 634 driving end-to-end encryption for these content providers. 636 Content replication in caches (for example live video, DRM protected 637 content) is used to most efficiently utilize the available limited 638 bandwidth and thereby maximize the user's Quality of Experience 639 (QoE). Especially in mobile networks, duplicating every stream 640 through the transit network increases backhaul cost for live TV. The 641 Enhanced Multimedia Broadcast/Multicast Services (3GPP eMBMS) - 642 trusted edge proxies facilitate delivering same stream to different 643 users, using either unicast or multicast depending on channel 644 conditions to the user. 646 Alternate approaches such as blind caches [I-D.thomson-http-bc] are 647 being explored to allow caching of encrypted content; however, they 648 still need to intercept the end-to-end transport connection. 650 2.2.6. Content Compression 652 In addition to caching, various applications exist to provide data 653 compression in order to conserve the life of the user's mobile data 654 plan and optimize delivery over the mobile link. The compression 655 proxy access can be built into a specific user level application, 656 such as a browser, or it can be available to all applications using a 657 system level application. The primary method is for the mobile 658 application to connect to a centralized server as a proxy, with the 659 data channel between the client application and the server using 660 compression to minimize bandwidth utilization. The effectiveness of 661 such systems depends on the server having access to unencrypted data 662 flows. 664 2.3. Network Access and Accounting 666 Mobile Networks and many ISPs operate under the regulations of their 667 licensing government authority. These regulations include Lawful 668 Intercept, adherence to Codes of Practice on content filtering, and 669 application of court order filters. Such regulations assume network 670 access to provide content filtering and accounting, as discussed 671 below. 673 2.3.1. Content Filtering 675 There are numerous reasons why service providers might block content: 676 to comply with requests from law enforcement or regulatory 677 authorities, to effectuate parental controls, to enforce content- 678 based billing, or for other reasons, possibly considered 679 inappropriate by some. See RFC7754 [RFC7754] for a survey of 680 Internet filtering techniques and motivations. This section is 681 intended to document a selection of current content blocking 682 practices by operators and the effects of encryption on those 683 practices. Content blocking may also happen at endpoints or at the 684 edge of enterprise networks, but those are not addressed in this 685 section. 687 In a mobile network content filtering usually occurs in the core 688 network. A proxy is installed which analyses the transport metadata 689 of the content users are viewing and either filters content based on 690 a blacklist of sites or based on the user's pre-defined profile (e.g. 691 for age sensitive content). Although filtering can be done by many 692 methods, one commonly used method involves a trigger based on the 693 proxy identifying a DNS lookup of a host name in a URL which appears 694 on a blacklist being used by the operator. The subsequent requests 695 to that domain will be re-routed to a proxy which checks whether the 696 full URL matches a blocked URL on the list, and will return a 404 if 697 a match is found. All other requests should complete. This 698 technique does not work in situations where DNS traffic is encrypted 699 (e.g., by employing [RFC7858] ). 701 Another form of content filtering is called parental control, where 702 some users are deliberately denied access to age-sensitive content as 703 a feature to the service subscriber. Some sites involve a mixture of 704 universal and age-sensitive content and filtering software. In these 705 cases, more granular (application layer) metadata may be used to 706 analyze and block traffic. Methods that accessed cleartext 707 application-layer metadata no longer work when sessions are 708 encrypted. This type of granular filtering could occur at the 709 endpoint. However, the lack of ability to efficiently manage 710 endpoints as a service reduces providers' ability to offer parental 711 control. 713 2.3.2. Network Access and Data Usage 715 Approved access to a network is a prerequisite to requests for 716 Internet traffic. 718 However, there are cases (beyond parental control) when a network 719 service provider currently redirects customer requests for content 720 (affecting content accessibility): 722 1. The network service provider is performing the accounting and 723 billing for the content provider, and the customer has not (yet) 724 purchased the requested content. 726 2. Further content may not be allowed as the customer has reached 727 their usage limit and needs to purchase additional data service, 728 which is the usual billing approach in mobile networks. 730 Currently, some mobile network service providers redirect the 731 customer using HTTP redirect to a page that explains to those 732 customers the reason for the blockage and the steps to proceed. When 733 the HTTP headers and content are encrypted, this prevents mobile 734 carriers from intercepting the traffic and performing an HTTP 735 redirect. As a result, some mobile carriers block customer's 736 encrypted requests, which is a far less optimal customer experience 737 because the blocking reason must be conveyed by some other means. 738 The customer may need to call customer care to find out the reason, 739 both an inconvenience to the customer and additional overhead to the 740 mobile network service provider. 742 Further, when the requested service is about to consume the remainder 743 of the user's plan limits, the transmission could be terminated and 744 advance notifications may be sent to the user by their service 745 provider to warn the user ahead of the exhausted plan. If web 746 content is encrypted, the network provider cannot know the data 747 transfer size at request time. Lacking this visibility of the 748 application type and content size, the network would continue the 749 transmission and stop the transfer when the limit was reached. A 750 partial transfer may not be usable by the client wasting both network 751 and user resources, possibly leading to customer complaints. The 752 content provider does not know user's service plans or current usage, 753 and cannot warn the user of plan exhaustion. 755 In addition, mobile network operator often sell tariffs that allow 756 free-data access to certain sites, known as 'zero rating'. A session 757 to visit such a site incurs no additional cost or data usage to the 758 user. This feature is impacted if encryption hides the details of 759 the content domain from the network. 761 2.3.3. Application Layer Gateways 763 Application Layer Gateways (ALG) assist applications to set 764 connectivity across Network Address Translators (NAT), Firewalls, 765 and/or Load Balancers for specific applications running across mobile 766 networks. Section 2.9 of [RFC2663] describes the role of ALGs and 767 their interaction with NAT and/or application payloads. ALG are 768 deployed with an aim to improve connectivity. However, it is an IETF 769 Best Common Practice recommendation that ALGs for UDP-based protocols 770 SHOULD be turned off [RFC4787]. 772 One example of an ALG in current use is aimed at video applications 773 that use the Real Time Session Protocol (RTSP) [RFC7826] primary 774 stream as a means to identify related Real Time Protocol/Real Time 775 Control Protocol (RTP/RTCP) [RFC3550] flows at set-up. The ALG in 776 this case relies on the 5-tuple flow information derived from RTSP to 777 provision NAT or other middle boxes and provide connectivity. 778 Implementations vary, and two examples follow: 780 1. Parse the content of the RTSP stream and identify the 5-tuple of 781 the supporting streams as they are being negotiated. 783 2. Intercept and modify the 5-tuple information of the supporting 784 media streams as they are being negotiated on the RTSP stream, 785 which is more intrusive to the media streams. 787 When RTSP stream content is encrypted, the 5-tuple information within 788 the payload is not visible to these ALG implementations, and 789 therefore they cannot provision their associated middelboxes with 790 that information. 792 2.3.4. HTTP Header Insertion 794 Some mobile carriers use HTTP header insertion (see section 3.2.1 of 795 [RFC7230]) to provide information about their customers to third 796 parties or to their own internal systems [Enrich]. Third parties use 797 the inserted information for analytics, customization, advertising, 798 to bill the customer, or to selectively allow or block content. HTTP 799 header insertion is also used to pass information internally between 800 a mobile service provider's sub-systems, thus keeping the internal 801 systems loosely coupled. When HTTP connections are encrypted, mobile 802 network service providers cannot insert headers to accomplish the 803 functions above. 805 3. Encryption in Hosting SP Environments 807 Hosted environments have had varied requirements in the past for 808 encryption, with many businesses choosing to use these services 809 primarily for data and applications that are not business or privacy 810 sensitive. A shift prior to the revelations on surveillance/passive 811 monitoring began where businesses were asking for hosted environments 812 to provide higher levels of security so that additional applications 813 and service could be hosted externally. Businesses understanding the 814 threats of monitoring in hosted environments only increased that 815 pressure to provide more secure access and session encryption to 816 protect the management of hosted environments as well as for the data 817 and applications. 819 3.1. Management Access Security 821 Hosted environments may have multiple levels of management access, 822 where some may be strictly for the Hosting SP (infrastructure that 823 may be shared among customers) and some may be accessed by a specific 824 customer for application management. In some cases, there are 825 multiple levels of hosting service providers, further complicating 826 the security of management infrastructure and the associated 827 requirements. 829 Hosting service provider management access is typically segregated 830 from other traffic with a control channel and may or may not be 831 encrypted depending upon the isolation characteristics of the 832 management session. Customer access may be through a dedicated 833 connection, but discussion for that connection method is out-of-scope 834 for this document. 836 Application Service Providers may offer content-level monitoring 837 options to detect intellectual property leakage, or other attacks. 838 In service provider environments where Data Loss Prevention (DLP) has 839 been implemented on the basis of the service provider having 840 cleartext access to session streams, the use of encrypted streams 841 prevents these implementations from conducting content searches for 842 the keywords or phrases configured in the DLP system. DLP is often 843 used to prevent the leakage of Personally Identifiable Information 844 (PII) as well as financial account information, Personal Health 845 Information (PHI), and Payment Card Information (PCI). If session 846 encryption is terminated at a gateway prior to accessing these 847 services, DLP on session data can still be performed. The decision 848 of where to terminate encryption to hosted environments will be a 849 risk decision made between the application service provider and 850 customer organization according to their priorities. DLP can be 851 performed at the server for the hosted application and on an end 852 user's system in an organization as alternate or additional 853 monitoring points of content; however, this is not frequently done in 854 a service provider environment. 856 Application service providers, by their very nature, control the 857 application endpoint. As such, much of the information gleaned from 858 sessions are still available on that endpoint. However, when a gap 859 exists in the application's logging and debugging capabilities, this 860 has led the application service provider to access data-in-transport 861 for monitoring and debugging. 863 Overlay networks (e.g. VXLAN, Geneve, etc.) may be used to indicate 864 desired isolation, but this is not sufficient to prevent deliberate 865 attacks that are aware of the use of the overlay network. It is 866 possible to use an overlay header in combination with IPsec, but this 867 adds the requirement for authentication infrastructure and may reduce 868 packet transfer performance. Additional extension mechanisms to 869 provide integrity and/or privacy protections are being investigated 870 for overlay encapsulations. Section 7 of [RFC7348] describes some of 871 the security issues possible when deploying VXLAN on Layer 2 872 networks. Rogue endpoints can join the multicast groups that carry 873 broadcast traffic, for example. 875 3.1.1. Customer Access Monitoring 877 Hosted applications that allow some level of customer management 878 access may also require monitoring by the hosting service provider. 879 Monitoring could include access control restrictions such as 880 authentication, authorization, and accounting for filtering and 881 firewall rules to ensure they are continuously met. Customer access 882 may occur on multiple levels, including user-level and administrative 883 access. The hosting service provider may need to monitor access 884 either through session monitoring or log evaluation to ensure 885 security service level agreements (SLA) for access management are 886 met. The use of session encryption to access hosted environments 887 limits access restrictions to the metadata described below. 888 Monitoring and filtering may occur at an: 890 2-tuple IP-level with source and destination IP addresses alone, or 892 5-tuple IP and protocol-level with source IP address, destination IP 893 address, protocol number, source port number, and destination port 894 number. 896 Session encryption at the application level, TLS for example, 897 currently allows access to the 5-tuple. IP-level encryption, such as 898 IPsec in tunnel mode prevents access to the original 5-tuple and may 899 limit the ability to restrict traffic via filtering techniques. This 900 shift may not impact all hosting service provider solutions as 901 alternate controls may be used to authenticate sessions or access may 902 require that clients access such services by first connecting to the 903 organization before accessing the hosted application. Shifts in 904 access may be required to maintain equivalent access control 905 management. Logs may also be used for monitoring that access control 906 restrictions are met, but would be limited to the data that could be 907 observed due to encryption at the point of log generation. Log 908 analysis is out of scope for this document. 910 3.1.2. SP Content Monitoring of Applications 912 The following observations apply to any IT organization that is 913 responsible for delivering services, whether to third-parties, for 914 example as a web based service, or to internal customers in an 915 enterprise, e.g. a data processing system that forms a part of the 916 enterprise's business. 918 Organizations responsible for the operation of a data center have 919 many processes which access the contents of IP packets (passive 920 methods of measurement, as defined in [RFC7799]). These processes 921 are typically for service assurance or security purposes as part of 922 their data center operations. 924 Examples include: 926 - Network Performance Monitoring/Application Performance 927 Monitoring 929 - Intrusion defense/prevention systems 931 - Malware detection 933 - Fraud Monitoring 935 - Application DDOS protection 937 - Cyber-attack investigation 939 - Proof of regulatory compliance 941 Many application service providers simply terminate sessions to/from 942 the Internet at the edge of the data center in the form of SSL/TLS 943 offload in the load balancer. Not only does this reduce the load on 944 application servers, it simplifies the processes to enable monitoring 945 of the session content. 947 However, in some situations, encryption deeper in the data center may 948 be necessary to protect personal information or in order to meet 949 industry regulations, e.g. those set out by the Payment Card Industry 950 (PCI). In such situations, various methods have been used to allow 951 service assurance and security processes to access unencrypted data. 952 These include SSL/TLS decryption in dedicated units, which then 953 forward packets to SP-controlled tools, or by real-time or post- 954 capture decryption in the tools themselves. The use of tools that 955 perform SSL/TLS decryption are impacted by the increased use of 956 encryption that prevents interception. 958 Data center operators may also maintain packet recordings in order to 959 be able to investigate attacks, breach of internal processes, etc. 960 In some industries, organizations may be legally required to maintain 961 such information for compliance purposes. Investigations of this 962 nature have used access to the unencrypted contents of the packet. 963 Alternate methods to investigate attacks or breach of process will 964 rely on endpoint information, such as logs. As previously noted, 965 logs often lack complete information, and this is seen as a concern 966 resulting in some relying on session access for additional 967 information. 969 3.2. Hosted Applications 971 Organizations are increasingly using hosted applications rather than 972 in-house solutions that require maintenance of equipment and 973 software. Examples include Enterprise Resource Planning (ERP) 974 solutions, payroll service, time and attendance, travel and expense 975 reporting among others. Organizations may require some level of 976 management access to these hosted applications and will typically 977 require session encryption or a dedicated channel for this activity. 979 In other cases, hosted applications may be fully managed by a hosting 980 service provider with service level agreement expectations for 981 availability and performance as well as for security functions 982 including malware detection. Due to the sensitive nature of these 983 hosted environments, the use of encryption is already prevalent. Any 984 impact may be similar to an enterprise with tools being used inside 985 of the hosted environment to monitor traffic. Additional concerns 986 were not reported in the call for contributions. 988 3.2.1. Monitoring Managed Applications 990 Performance, availability, and other aspects of a SLA are often 991 collected through passive monitoring. For example: 993 o Availability: ability to establish connections with hosts to 994 access applications, and discern the difference between network or 995 host-related causes of unavailability. 997 o Performance: ability to complete transactions within a target 998 response time, and discern the difference between network or host- 999 related causes of excess response time. 1001 Here, as with all passive monitoring, the accuracy of inferences are 1002 dependent on the cleartext information available, and encryption 1003 would tend to reduce the information and therefore, the accuracy of 1004 each inference. Passive measurement of some metrics will be 1005 impossible with encryption that prevents inferring packet 1006 correspondence across multiple observation points, such as for packet 1007 loss metrics. 1009 Until application logging is sufficient, the ability to make accurate 1010 inferences in an environment with increased encryption will remain a 1011 gap for passive performance monitoring. 1013 3.2.2. Mail Service Providers 1015 Mail (application) service providers vary in what services they 1016 offer. Options may include a fully hosted solution where mail is 1017 stored external to an organization's environment on mail service 1018 provider equipment or the service offering may be limited to monitor 1019 incoming mail to remove SPAM [Section 5.1], malware [Section 5.6], 1020 and phishing attacks [Section 5.3] before mail is directed to the 1021 organization's equipment. In both of these cases, content of the 1022 messages and headers is monitored to detect SPAM, malware, phishing, 1023 and other messages that may be considered an attack. 1025 STARTTLS ought have zero effect on anti-SPAM efforts for SMTP 1026 traffic. Anti-SPAM services could easily be performed on an SMTP 1027 gateway, eliminating the need for TLS decryption services. The 1028 impact to Anti-SPAM service providers should be limited to a change 1029 in tools, where middle boxes were deployed to perform these 1030 functions. 1032 Many efforts are emerging to improve user-to-user encryption, 1033 including promotion of PGP and newer efforts such as Dark Mail 1034 [DarkMail]. Of course, SPAM filtering will not be possible on 1035 encrypted content. 1037 3.3. Data Storage 1039 Numerous service offerings exist that provide hosted storage 1040 solutions. This section describes the various offerings and details 1041 the monitoring for each type of service and how encryption may impact 1042 the operational and security monitoring performed. 1044 Trends in data storage encryption for hosted environments include a 1045 range of options. The following list is intentionally high-level to 1046 describe the types of encryption used in coordination with data 1047 storage that may be hosted remotely, meaning the storage is 1048 physically located in an external data center requiring transport 1049 over the Internet. Options for monitoring will vary with each 1050 encryption approach described below. 1052 3.3.1. Host-level Encryption 1054 For higher security and/or privacy of data and applications, options 1055 that provide end-to-end encryption of the data from the user's 1056 desktop or server to the storage platform may be preferred. With 1057 this description, host level encryption includes any solution that 1058 encrypts data at the object level, not transport level. Encryption 1059 of data may be performed with libraries on the system or at the 1060 application level, which includes file encryption services via a file 1061 manager. Host-level encryption is useful when data storage is 1062 hosted, or scenarios when storage location is determined based on 1063 capacity or based on a set of parameters to automate decisions. This 1064 could mean that large data sets accessed infrequently could be sent 1065 to an off-site storage platform at an external hosting service, data 1066 accessed frequently may be stored locally, or the decision could be 1067 based on the transaction type. Host-level encryption is grouped 1068 separately for the purpose of this document since data may be stored 1069 in multiple locations including off-site remote storage platforms. 1070 If session encryption is used, the protocol is likely to be TLS. 1072 3.3.1.1. Monitoring for Hosted Storage 1074 Monitoring of hosted storage solutions that use host-level (object) 1075 encryption is described in this subsection. Host-level encryption 1076 can be employed for backup services, and occasionally for external 1077 storage services (operated by a third party) when internal storage 1078 limits are exceeded. 1080 Monitoring of data flows to hosted storage solutions is performed for 1081 security and operational purposes. The security monitoring may be to 1082 detect anomalies in the data flows that could include changes to 1083 destination, the amount of data transferred, or alterations in the 1084 size and frequency of flows. Operational considerations include 1085 capacity and availability monitoring. 1087 3.3.2. Disk Encryption, Data at Rest 1089 There are multiple ways to achieve full disk encryption for stored 1090 data. Encryption may be performed on data to be stored while in 1091 transit close to the storage media with solutions like Controller 1092 Based Encryption (CBE) or in the drive system with Self-Encrypting 1093 Drives (SED). Session encryption is typically coupled with 1094 encryption of these data at rest (DAR) solutions to also protect data 1095 in transit. Transport encryption is likely via TLS. 1097 3.3.2.1. Monitoring Session Flows for DAR Solutions 1099 Monitoring for transport of data to storage platforms, where object 1100 level encryption is performed close to or on the storage platform are 1101 similar to those described in the section on Monitoring for Hosted 1102 Storage. The primary difference for these solutions is the possible 1103 exposure of sensitive information, which could include privacy 1104 related data, financial information, or intellectual property if 1105 session encryption via TLS is not deployed. Session encryption is 1106 typically used with these solutions, but that decision would be based 1107 on a risk assessment. There are use cases where DAR or disk-level 1108 encryption is required. Examples include preventing exposure of data 1109 if physical disks are stolen or lost. In the case where TLS is in 1110 use, monitoring and the exposure of data is limited to a 5-tuple. 1112 3.3.3. Cross Data Center Replication Services 1114 Storage services also include data replication which may occur 1115 between data centers and may leverage Internet connections to tunnel 1116 traffic. The traffic may use iSCSI [RFC7143] or FC/IP [RFC7146] 1117 encapsulated in IPsec. Either transport or tunnel mode may be used 1118 for IPsec depending upon the termination points of the IPsec session, 1119 if it is from the storage platform itself or from a gateway device at 1120 the edge of the data center respectively. 1122 3.3.3.1. Monitoring Of IPSec for Data Replication Services 1124 Monitoring for data replication services are described in this 1125 subsection. 1127 Monitoring of data flows between data centers may be performed for 1128 security and operational purposes and would typically concentrate 1129 more on operational aspects since these flows are essentially virtual 1130 private networks (VPN) between data centers. Operational 1131 considerations include capacity and availability monitoring. The 1132 security monitoring may be to detect anomalies in the data flows, 1133 similar to what was described in the "Monitoring for Hosted Storage 1134 Section". If IPsec tunnel mode is in use, monitoring is limited to a 1135 2-tuple, or with transport mode, a 5-tuple. 1137 4. Encryption for Enterprises 1139 Encryption of network traffic within the private enterprise is a 1140 growing trend, particularly in industries with audit and regulatory 1141 requirements. Some enterprise internal networks are almost 1142 completely TLS and/or IPsec encrypted. 1144 For each type of monitoring, different techniques and access to parts 1145 of the data stream are part of current practice. As we transition to 1146 an increased use of encryption, alternate methods of monitoring for 1147 operational purposes may be necessary to reduce the practice of 1148 breaking encryption (other policies may apply in some enterprise 1149 settings). 1151 4.1. Monitoring Practices of the Enterprise 1153 Large corporate enterprises are the owners of the platforms, data, 1154 and network infrastructure that provide critical business services to 1155 their user communities. As such, these enterprises are responsible 1156 for all aspects of the performance, availability, security, and 1157 quality of experience for all user sessions. These responsibilities 1158 break down into three basic areas: 1160 1. Security Monitoring and Control 1162 2. Application Performance Monitoring and Reporting 1164 3. Network Diagnostics and Troubleshooting 1166 In each of the above areas, technical support teams utilize 1167 collection, monitoring, and diagnostic systems. Some organizations 1168 currently use attack methods such as replicated TLS server RSA 1169 private keys to decrypt passively monitored copies of encrypted TLS 1170 packet streams. 1172 For an enterprise to avoid costly application down time and deliver 1173 expected levels of performance, protection, and availability, some 1174 forms of traffic analysis, sometimes including examination of packet 1175 payloads, are currently used. 1177 4.1.1. Security Monitoring in the Enterprise 1179 Enterprise users are subject to the policies of their organization 1180 and the jurisdictions in which the enterprise operates. As such, 1181 proxies may be in use to: 1183 1. intercept outbound session traffic to monitor for intellectual 1184 property leakage (by users or, more likely these days, through 1185 malware and trojans), 1187 2. detect viruses/malware entering the network via email or web 1188 traffic, 1190 3. detect malware/Trojans in action, possibly connecting to remote 1191 hosts, 1193 4. detect attacks (Cross site scripting and other common web related 1194 attacks), 1196 5. track misuse and abuse by employees, 1198 6. restrict the types of protocols permitted to/from the entire 1199 corporate environment, 1201 7. detect and defend against Internet DDoS attacks, including both 1202 volumetric and layer 7 attacks. 1204 A significant portion of malware hides its activity within TLS or 1205 other encrypted protocols. This includes lateral movement, Command 1206 and Control, and Data Exfiltration. Detecting these functions are 1207 important to effective monitoring and mitigation of malicious 1208 traffic, not limited to malware. 1210 Security monitoring in the enterprise may also be performed at the 1211 endpoint with numerous current solutions that mitigate the same 1212 problems as some of the above mentioned solutions. Since the 1213 software agents operate on the device, they are able to monitor 1214 traffic before it is encrypted, monitor for behavior changes, and 1215 lock down devices to use only the expected set of applications. 1216 Session encryption does not affect these solutions. Some might argue 1217 that scaling is an issue in the enterprise, but some large 1218 enterprises have used these tools effectively. 1220 4.1.2. Application Performance Monitoring in the Enterprise 1222 There are two main goals of monitoring: 1224 1. Assess traffic volume on a per-application basis, for billing, 1225 capacity planning, optimization of geographical location for 1226 servers or proxies, and other goals. 1228 2. Assess performance in terms of application response time and user 1229 perceived response time. 1231 Network-based Application Performance Monitoring tracks application 1232 response time by user and by URL, which is the information that the 1233 application owners and the lines of business request. Content 1234 Delivery Networks (CDNs) add complexity in determining the ultimate 1235 endpoint destination. By their very nature, such information is 1236 obscured by CDNs and encrypted protocols -- adding a new challenge 1237 for troubleshooting network and application problems. URL 1238 identification allows the application support team to do granular, 1239 code level troubleshooting at multiple tiers of an application. 1241 New methodologies to monitor user perceived response time and to 1242 separate network from server time are evolving. For example, the 1243 IPv6 Destination Option Header (DOH) implementation of Performance 1244 and Diagnostic Metrics (PDM) will provide this 1245 [I-D.ietf-ippm-6man-pdm-option]. Using PDM with IPSec Encapsulating 1246 Security Payload (ESP) Transport Mode requires placement of the PDM 1247 DOH within the ESP encrypted payload to avoid leaking timing and 1248 sequence number information that could be useful to an attacker. Use 1249 of PDM DOH also may introduce some security weaknesses, including a 1250 timing attack, as described in Section 7 of 1251 [I-D.ietf-ippm-6man-pdm-option]. For these and other reasons, 1252 [I-D.ietf-ippm-6man-pdm-option] requires that the PDM DOH option be 1253 explicitly turned on by administrative action in each host where this 1254 measurement feature will be used. 1256 4.1.3. Enterprise Network Diagnostics and Troubleshooting 1258 One primary key to network troubleshooting is the ability to follow a 1259 transaction through the various tiers of an application in order to 1260 isolate the fault domain. A variety of factors relating to the 1261 structure of the modern data center and multi-tiered application have 1262 made it difficult to follow a transaction in network traces without 1263 the ability to examine some of the packet payload. Alternate 1264 methods, such as log analysis need improvement to fill this gap. 1266 4.1.3.1. Address Sharing (NAT) 1268 Content Delivery Networks (CDNs) and NATs and Network Address and 1269 Port Translators (NAPT) obscure the ultimate endpoint designation 1270 (See [RFC6269] for types of address sharing and a list of issues). 1271 Troubleshooting a problem for a specific end user requires finding 1272 information such as the IP address and other identifying information 1273 so that their problem can be resolved in a timely manner. 1275 NAT is also frequently used by lower layers of the data center 1276 infrastructure. Firewalls, Load Balancers, Web Servers, App Servers, 1277 and Middleware servers all regularly NAT the source IP of packets. 1278 Combine this with the fact that users are often allocated randomly by 1279 load balancers to all these devices, the network troubleshooter is 1280 often left with very few options in today's environment due to poor 1281 logging implementations in applications. As such, network 1282 troubleshooting is used to trace packets at a particular layer, 1283 decrypt them, and look at the payload to find a user session. 1285 This kind of bulk packet capture and bulk decryption is frequently 1286 used when troubleshooting a large and complex application. Endpoints 1287 typically don't have the capacity to handle this level of network 1288 packet capture, so out-of-band networks of robust packet brokers and 1289 network sniffers that use techniques such as copies of TLS RSA 1290 private keys accomplish this task today. 1292 4.1.3.2. TCP Pipelining/Session Multiplexing 1294 TCP Pipelining/Session Multiplexing used mainly by middle boxes today 1295 allow for multiple end user sessions to share the same TCP 1296 connection. Today's network troubleshooter often relies upon session 1297 decryption to tell which packet belongs to which end user, since the 1298 logs are currently inadequate for the analysis performed. 1300 Increased use of HTTP/2 will likely further increase the prevalence 1301 of session multiplexing, both on the Internet and in the private data 1302 center. 1304 4.1.3.3. HTTP Service Calls 1306 When an application server makes an HTTP service call to back end 1307 services on behalf of a user session, it uses a completely different 1308 URL and a completely different TCP connection. Troubleshooting via 1309 network trace involves matching up the user request with the HTTP 1310 service call. Some organizations do this today by decrypting the TLS 1311 packet and inspecting the payload. Logging has not been adequate for 1312 their purposes. 1314 4.1.3.4. Application Layer Data 1316 Many applications use text formats such as XML to transport data or 1317 application level information. When transaction failures occur and 1318 the logs are inadequate to determine the cause, network and 1319 application teams work together, each having a different view of the 1320 transaction failure. Using this troubleshooting method, the network 1321 packet is correlated with the actual problem experienced by an 1322 application to find a root cause. The inability to access the 1323 payload prevents this method of troubleshooting. 1325 4.2. Techniques for Monitoring Internet Session Traffic 1327 Corporate networks commonly monitor outbound session traffic to 1328 detect or prevent attacks as well as to guarantee service level 1329 expectations. In some cases, alternate options are available when 1330 encryption is in use, but techniques like that of data leakage 1331 prevention tools at a proxy would not be possible if encrypted 1332 traffic cannot be intercepted, encouraging alternate options such as 1333 performing these functions at the edge. 1335 Some DLP tools intercept traffic at the Internet gateway or proxy 1336 services with the ability to man-in-the-middle (MiTM) encrypted 1337 session traffic (HTTP/TLS). These tools may use key words important 1338 to the enterprise including business sensitive information such as 1339 trade secrets, financial data, personally identifiable information 1340 (PII), or personal health information (PHI). Various techniques are 1341 used to intercept HTTP/TLS sessions for DLP and other purposes, and 1342 are described in "Summarizing Known Attacks on TLS and DTLS" 1343 [RFC7457]. Note: many corporate policies allow access to personal 1344 financial and other sites for users without interception. Another 1345 option is to terminate a TLS session prior to the point where 1346 monitoring is performed. 1348 Monitoring traffic patterns for anomalous behavior such as increased 1349 flows of traffic that could be bursty at odd times or flows to 1350 unusual destinations (small or large amounts of traffic) is common. 1351 This traffic may or may not be encrypted and various methods of 1352 encryption or just obfuscation may be used. 1354 Restrictions on traffic to approved sites: Web proxies are sometimes 1355 used to filter traffic, allowing only access to well-known sites 1356 found to be legitimate and free of malware on last check by a proxy 1357 service company. This type of restriction is usually not noticeable 1358 in a corporate setting as the typical corporate user does not access 1359 sites that are not well-known to these tools, but may be noticeable 1360 to those in research who are unable to access colleague's individual 1361 sites or new web sites that have not yet been screened. In 1362 situations where new sites are required for access, they can 1363 typically be added after notification by the user or proxy log alerts 1364 and review. Home mail account access may be blocked in corporate 1365 settings to prevent another vector for malware to enter as well as 1366 for intellectual property to leak out of the network. This method 1367 remains functional with increased use of encryption and may be more 1368 effective at preventing malware from entering the network. Web proxy 1369 solutions monitor and potentially restrict access based on the 1370 destination URL or the DNS name. A complete URL may be used in cases 1371 where access restrictions vary for content on a particular site or 1372 for the sites hosted on a particular server. 1374 Desktop DLP tools are used in some corporate environments as well. 1375 Since these tools reside on the desktop, they can intercept traffic 1376 before it is encrypted and may provide a continued method of 1377 monitoring intellectual property leakage from the desktop to the 1378 Internet or attached devices. 1380 DLP tools can also be deployed by Network Service providers, as they 1381 have the vantage point of monitoring all traffic paired with 1382 destinations off the enterprise network. This makes an effective 1383 solution for enterprises that allow "bring-your-own" devices when the 1384 traffic is not encrypted, and for devices outside the desktop 1385 category (such as mobile phones) that are used on corporate networks 1386 nonetheless. 1388 Enterprises may wish to reduce the traffic on their Internet access 1389 facilities by monitoring requests for within-policy content and 1390 caching it. In this case, repeated requests for Internet content 1391 spawned by URLs in e-mail trade newsletters or other sources can be 1392 served within the enterprise network. Gradual deployment of end to 1393 end encryption would tend to reduce the cacheable content over time, 1394 owing to concealment of critical headers and payloads. Many forms of 1395 enterprise performance management may be similarly affected. 1397 5. Security Monitoring for Specific Attack Types 1399 Effective incident response today requires collaboration at Internet 1400 scale. This section will only focus on efforts of collaboration at 1401 Internet scale that are dedicated to specific attack types. They may 1402 require new monitoring and detection techniques in an increasingly 1403 encrypted Internet. As mentioned previously, some service providers 1404 have been interfering with STARTTLS to prevent session encryption to 1405 be able to perform functions they are used to (injecting ads, 1406 monitoring, etc.). By detailing the current monitoring methods used 1407 for attack detection and response, this information can be used to 1408 devise new monitoring methods that will be effective in the changed 1409 Internet via collaboration and innovation. 1411 5.1. Mail Abuse and SPAM 1413 The largest operational effort to prevent mail abuse is through the 1414 Messaging, Malware, Mobile Anti-Abuse Working Group (M3AAWG)[M3AAWG]. 1415 Mail abuse is combatted directly with mail administrators who can 1416 shut down or stop continued mail abuse originating from large scale 1417 providers that participate in using the Abuse Reporting Format (ARF) 1418 agents standardized in the IETF [RFC5965], [RFC6430], [RFC6590], 1419 [RFC6591], [RFC6650], [RFC6651], and [RFC6652]. The ARF agent 1420 directly reports abuse messages to the appropriate service provider 1421 who can take action to stop or mitigate the abuse. Since this 1422 technique uses the actual message, the use of SMTP over TLS between 1423 mail gateways will not affect its usefulness. As mentioned 1424 previously, SMTP over TLS only protects data while in transit and the 1425 messages may be exposed on mail servers or mail gateways if a user- 1426 to-user encryption method is not used. Current user-to-user message 1427 encryption methods on email (S/MIME and PGP) do not encrypt the email 1428 header information used by ARF and the service provider operators in 1429 their abuse mitigation efforts. 1431 5.2. Denial of Service 1433 Response to Denial of Service (DoS) attacks are typically coordinated 1434 by the SP community with a few key vendors who have tools to assist 1435 in the mitigation efforts. Traffic patterns are determined from each 1436 DoS attack to stop or rate limit the traffic flows with patterns 1437 unique to that DoS attack. 1439 Data types used in monitoring traffic for DDoS are described in the 1440 DDoS Open Threat Signaling (DOTS) [DOTS] working group documents in 1441 development. 1443 Data types used in DDoS attacks have been detailed in the IODEF 1444 Guidance draft [I-D.ietf-mile-iodef-guidance], Appendix A.2, with the 1445 help of several members of the service provider community. The 1446 examples provided are intended to help identify the useful data in 1447 detecting and mitigating these attacks independent of the transport 1448 and protocol descriptions in the drafts. 1450 5.3. Phishing 1452 Investigations and response to phishing attacks follow well-known 1453 patterns, requiring access to specific fields in email headers as 1454 well as content from the body of the message. When reporting 1455 phishing attacks, the recipient has access to each field as well as 1456 the body to make content reporting possible, even when end-to-end 1457 encryption is used. The email header information is useful to 1458 identify the mail servers and accounts used to generate or relay the 1459 attack messages in order to take the appropriate actions. The 1460 content of the message often contains an embedded attack that may be 1461 in an infected file or may be a link that results in the download of 1462 malware to the user's system. 1464 Administrators often find it helpful to use header information to 1465 track down similar message in their mail queue or users inboxes to 1466 prevent further infection. Combinations of To:, From:, Subject:, 1467 Received: from header information might be used for this purpose. 1468 Administrators may also search for document attachments of the same 1469 name, size, or containing a file with a matching hash to a known 1470 phishing attack. Administrators might also add URLs contained in 1471 messages to block lists locally or this may also be done by browser 1472 vendors through larger scales efforts like that of the Anti-Phishing 1473 Working Group (APWG). See the Coordinating Attack Response at 1474 Internet Scale (CARIS) workshop Report [RFC8073] for additional 1475 information and pointers to the APWG's efforts on anti- phishing. 1477 A full list of the fields used in phishing attack incident response 1478 can be found in RFC5901. Future plans to increase privacy 1479 protections may limit some of these capabilities if some email header 1480 fields are encrypted, such as To:, From:, and Subject: header fields. 1481 This does not mean that those fields should not be encrypted, only 1482 that we should be aware of how they are currently used. 1484 Some products protect users from phishing by maintaining lists of 1485 known phishing domains (such as misspelled bank names) and blocking 1486 access. This can be done by observing DNS, clear-text HTTP, or SNI 1487 in TLS, in addition to analyzing email. Alternate options to detect 1488 and prevent phishing attacks may be needed. More recent examples of 1489 data exchanged in spear phishing attacks has been detailed in the 1490 IODEF Guidance draft [I-D.ietf-mile-iodef-guidance], Appendix A.3. 1492 5.4. Botnets 1494 Botnet detection and mitigation is complex and may involve hundreds 1495 or thousands of hosts with numerous Command and Control (C&C) 1496 servers. The techniques and data used to monitor and detect each may 1497 vary. Connections to C&C servers are typically encrypted, therefore 1498 a move to an increasingly encrypted Internet may not affect the 1499 detection and sharing methods used. 1501 5.5. Malware 1503 Malware monitoring and detection techniques vary. As mentioned in 1504 the enterprise section, malware monitoring may occur at gateways to 1505 the organization analyzing email and web traffic. These services can 1506 also be provided by service providers, changing the scale and 1507 location of this type of monitoring. Additionally, incident 1508 responders may identify attributes unique to types of malware to help 1509 track down instances by their communication patterns on the Internet 1510 or by alterations to hosts and servers. 1512 Data types used in malware investigations have been summarized in an 1513 example of the IODEF Guidance draft [I-D.ietf-mile-iodef-guidance], 1514 Appendix A.1. 1516 5.6. Spoofed Source IP Address Protection 1518 The IETF has reacted to spoofed source IP address-based attacks, 1519 recommending the use of network ingress filtering [RFC2827] and the 1520 unicast Reverse Path Forwarding (uRPF) mechanism [RFC2504]. But uRPF 1521 suffers from limitations regarding its granularity: a malicious node 1522 can still use a spoofed IP address included inside the prefix 1523 assigned to its link. The Source Address Validation Improvements 1524 (SAVI) mechanisms try to solve this issue. Basically, a SAVI 1525 mechanism is based on the monitoring of a specific address 1526 assignment/management protocol (e.g., SLAAC [RFC4862], SEND 1528 [RFC3971], DHCPv4/v6 [RFC2131][RFC3315]) and, according to this 1529 monitoring, set-up a filtering policy allowing only the IP flows with 1530 a correct source IP address (i.e., any packet with a source IP 1531 address, from a node not owning it, is dropped). The encryption of 1532 parts of the address assignment/management protocols, critical for 1533 SAVI mechanisms, can result in a dysfunction of the SAVI mechanisms. 1535 5.7. Further work 1537 Although incident response work will continue, new methods to prevent 1538 system compromise through security automation and continuous 1539 monitoring [SACM] may provide alternate approaches where system 1540 security is maintained as a preventative measure. 1542 6. Application-based Flow Information Visible to a Network 1544 This section describes specific techniques used in monitoring 1545 applications that may apply to various network types. 1547 6.1. TLS Server Name Indication 1549 When initiating the TLS handshake, the Client may provide an 1550 extension field (server_name) which indicates the server to which it 1551 is attempting a secure connection. TLS SNI was standardized in 2003 1552 to enable servers to present the "correct TLS certificate" to clients 1553 in a deployment of multiple virtual servers hosted by the same server 1554 infrastructure and IP-address. Although this is an optional 1555 extension, it is today supported by all modern browsers, web servers 1556 and developer libraries. Akamai [Nygren] reports that many of their 1557 customer see client TLS SNI usage over 99%. It should be noted that 1558 HTTP/2 introduces the Alt-SVC method for upgrading the connection 1559 from HTTP/1 to either unencrypted or encrypted HTTP/2. If the 1560 initial HTTP/1 request is unencrypted, the destination alternate 1561 service name can be identified before the communication is 1562 potentially upgraded to encrypted HTTP/2 transport. HTTP/2 requires 1563 the TLS implementation to support the Server Name Indication (SNI) 1564 extension (see section 9.2 of [RFC7540]). 1566 This information is only visible if the client is populating the 1567 Server Name Indication extension. This need not be done, but may be 1568 done as per TLS standard and as stated above this has been 1569 implemented by all major browsers. Therefore, even if existing 1570 network filters look out for seeing a Server Name Indication 1571 extension, they may not find one. It is also possible that SNI may 1572 be encrypted in future versions of TLS, but there is no way to do 1573 that currently. The per-domain nature of SNI may not reveal the 1574 specific service or media type being accessed, especially where the 1575 domain is of a provider offering a range of email, video, Web pages 1576 etc. For example, certain blog or social network feeds may be deemed 1577 'adult content', but the Server Name Indication will only indicate 1578 the server domain rather than a URL path. 1580 6.2. Application Layer Protocol Negotiation (ALPN) 1582 ALPN is a TLS extension which may be used to indicate the application 1583 protocol within the TLS session. This is likely to be of more value 1584 to the network where it indicates a protocol dedicated to a 1585 particular traffic type (such as video streaming) rather than a 1586 multi-use protocol. ALPN is used as part of HTTP/2 'h2', but will 1587 not indicate the traffic types which may make up streams within an 1588 HTTP/2 multiplex. ALPN will be encrypted in TLS 1.3. 1590 6.3. Content Length, BitRate and Pacing 1592 The content length of encrypted traffic is effectively the same as 1593 the cleartext. Although block ciphers utilise padding, this makes a 1594 negligible difference. Bitrate and pacing are generally application 1595 specific, and do not change much when the content is encrypted. 1596 Multiplexed formats (such as HTTP/2 and QUIC) may however incorporate 1597 several application streams over one connection, which makes the 1598 bitrate/pacing no longer application-specific. 1600 7. Impact on Mobility Network Optimizations and New Services 1602 This section considers the effects of transport level encryption on 1603 existing forms of mobile network optimization techniques, as well as 1604 potential new services. The material in this section assumes 1605 familiarity with mobile network concepts, specifications, and 1606 architectures. Readers who need additional background should start 1607 with the 3GPP's web pages on various topics of interest[Web3GPP], 1608 especially the article on Long Term Evolution (LTE). 3GPP provides a 1609 mapping between their expanding technologies and the different series 1610 of technical specifications [Map3GPP]. 3GPP also has a canonical 1611 specification of their vocabulary, definitions, and acronyms [Vocab], 1612 as does the RFC Editor for abbreviations [RFCEdit]. 1614 7.1. Effect of Encypted ACKs 1616 The stream of TCP ACKs that flow from a receiver of a byte stream 1617 using TCP for reliability, flow-control, and NAT/firewall transversal 1618 is called an ACK stream. The ACKs contain segment numbers that 1619 confirm successful transmission and their RTT, or indicate packet 1620 loss (duplicate ACKs). If this view of progress of stream transfer 1621 is lost, then the mobile network has greatly reduced ability to 1622 monitor transport layer performance. When the ACK stream is 1623 encrypted, it prevents the following mobile network functions from 1624 operating: 1626 a. Measurement of Network Segment (Sector, eNodeB (eNB) etc.) 1627 characterization KPIs (Retransmissions, packet drops, Sector 1628 Utilization Level etc.), estimation of User/Service KQIs at 1629 network edges for circuit emulation (CEM), and mitigation 1630 methods. The active services per user and per sector are not 1631 visible to a server that only services Internet Access Point 1632 Names (APN), and thus could not perform mitigation functions 1633 based on network segment view. 1635 b. Ability to deploy SP-operated proxies that reduce control round- 1636 trip time (RTT) between the TCP transmitter and receiver. The 1637 RTT determines how quickly a user's attempt to cancel a video is 1638 recognized (how quickly the traffic is stopped, thus keeping un- 1639 wanted video packets from entering the radio scheduler queue). 1641 c. Performance-enhancing proxy with low RTT determines the 1642 responsiveness of TCP flow control, and enables faster adaptation 1643 in a delay & capacity varying network due to user mobility. Low 1644 RTT permits use of a smaller send window, which makes the flow 1645 control loop more responsive to changing mobile network 1646 conditions. 1648 7.2. Effect of Encrypted Transport Headers 1650 When the Transport Header is encrypted, it prevents the following 1651 mobile network features from operating: 1653 a. Application-type-aware network edge (middlebox) that could 1654 control pacing, limit simultaneous HD videos, prioritize active 1655 videos against new videos, etc. 1657 b. For Self Organizing Networks (3GPP SON) - intelligent SON 1658 workflows such as content-aware MLB (Mobility Load Balancing) 1660 c. Reduces the benefits IP/DSCP-based transit network delivery 1661 optimizations where a mobile<->transit marking agreement exists; 1662 since multiple applications are multiplexed within the same 1663 5-tuple transport connection, a reasonable assumption is that the 1664 DSCP markings would be withheld from the outer IP header to 1665 further obscure which packets belong to each application flow. 1667 d. Advance notification for dense data usages - If the application 1668 types are visible, transit network element could warn (ahead of 1669 usage) that the requested service consumes user plan limits, and 1670 transmission could be terminated. Without such visibility, the 1671 network might have to continue the operation and stop the 1672 operation at the limit. Partially loaded content wastes 1673 resources and may not be usable by the client, thus increasing 1674 customer complaints. Content publisher will not know user- 1675 service plans, and Network Edge would not know data transfer 1676 lengths before large object is requested. 1678 7.3. Effect of Encryption on New or Emerging Services 1680 This section describes some new/emerging mobile services and how they 1681 might be affected with transport encryption: 1683 1. Content/Application based Prioritization of Over-the-Top (OTT) 1684 services - each application-type or service has different 1685 delay/loss/throughput expectations, and each type of stream will 1686 be unknown to an edge device if encrypted; this impedes dynamic- 1687 QoS adaptation. 1689 2. Rich Communication Services (3GPP-RCS) using different Quality 1690 Class Indicators (QCIs in LTE) - Operators offer different QoS 1691 classes for value-added services. The QCI type is visible in RAN 1692 control plane and invisible in user plane, thus the QCI cannot be 1693 set properly when the application -type is unknown. 1695 3. 1697 7.4. Effect of Encryption on Mobile Network Evolution 1699 The transport header encryption prevents trusted transit proxies. It 1700 may be that the benefits of such proxies could be achieved by end to 1701 end client & server optimizations and distribution using CDNs, plus 1702 the ability to continue connections across different access 1703 technologies (across dynamic user IP addresses). The following 1704 aspects need to be considered in this approach: 1706 1. In a wireless mobile network, the delay and channel capacity per 1707 user and sector varies due to coverage, contention, user 1708 mobility, and scheduling balances fairness, capacity and service 1709 QoE. If most users are at the cell edge, the controller cannot 1710 use more complex QAM, thus reducing total cell capacity; 1711 similarly if a UMTS edge is serving some number of CS-Voice 1712 Calls, the remaining capacity for packet services is reduced. 1714 2. Roamers: Mobile wireless networks service in-bound roamers (Users 1715 of Operator A in a foreign operator Network B) by backhauling 1716 their traffic though Operator B's network to Operator A's Network 1717 and then serving through the P-Gateway (PGW), General GPRS 1718 Support Node (GGSN), Content Distribution Network (CDN) etc., of 1719 Operator A (User's Home Operator). Increasing window sizes to 1720 compensate for the path RTT will have the limitations outlined 1721 earlier for TCP. The outbound roamer scenario has a similar TCP 1722 performance impact. 1724 3. Issues in deploying CDNs in RAN: Decreasing Client-Server control 1725 loop requires deploying CDNs/Cloud functions that terminate 1726 encryption closer to the edge. In Cellular RAN, the user IP 1727 traffic is encapsulated into General Packet Radio Service (GPRS) 1728 Tunneling Protocol-User Plane (GTP-U in UMTS and LTE) tunnels to 1729 handle user mobility; the tunnels terminate in APN/GGSN/PGW that 1730 are in central locations. One user's traffic may flow through 1731 one or more APN's (for example Internet APN, Roaming APN for 1732 Operator X, Video-Service APN, OnDeckAPN etc.). The scope of 1733 operator private IP addresses may be limited to specific APN. 1734 Since CDNs generally operate on user IP flows, deploying them 1735 would require enhancing them with tunnel translation, etc., 1736 tunnel management functions. 1738 4. While CDNs that de-encrypt flows or split-connection proxy 1739 (similar to split-tcp) could be deployed closer to the edges to 1740 reduce control loop RTT, with transport header encryption, such 1741 CDNs perform optimization functions only for partner client 1742 flows; thus content from some Small-Medium Businesses (SMBs) 1743 would not get such CDN benefits. 1745 8. Response to Increased Encryption and Looking Forward 1747 In the best case scenario, engineers and other innovators would work 1748 to solve the problems at hand in new ways rather than prevent the use 1749 of encryption. As stated in [RFC7258], "an appropriate balance 1750 (between network management and PM mitigations) will emerge over time 1751 as real instances of this tension are considered." 1753 There has already been documented cases of service providers 1754 preventing STARTTLS [NoEncrypt] to prevent session encryption 1755 negotiation on some session to inject a super cookie. In order to 1756 effectively deploy encryption and prevent interception, 1757 considerations for protocol design should factor in network 1758 management functions to work toward the balance called out in 1759 RFC7258. 1761 It is well known that national surveillance programs monitor traffic 1762 [JNSLP] as Internet security practitioners monitor for criminal 1763 activities. Governments vary on their balance between monitoring 1764 versus the protection of user privacy, data, and assets. Those that 1765 favor unencrypted access to data ignore the real need to protect 1766 users' identity, financial transactions and intellectual property, 1767 which requires security and encryption to prevent crime. A clear 1768 understanding of technology, encryption, and monitoring goals will 1769 aid in the development of solutions to appropriately balance these 1770 with privacy. As this understanding increases, hopefully the 1771 discussions will improve; this draft is meant to help further the 1772 discussion. 1774 Changes to improve encryption or to deploy OS methods have little 1775 impact on the detection of malicious actors; they already have access 1776 to strong encryption. The current push to increase encryption is 1777 aimed at increasing users' privacy and providing application 1778 integrity. There is already protection in place for purchases, 1779 financial transactions, systems management infrastructure, and 1780 intellectual property although this too can be improved. The 1781 Opportunistic Security (OS) [RFC7435] efforts aim to increase the 1782 costs of monitoring through the use of encryption that can be subject 1783 to active attacks, but make passive monitoring broadly cost 1784 prohibitive. This is meant to restrict monitoring to sessions where 1785 there is reason to have suspicion. 1787 9. Security Considerations 1789 There are no additional security considerations as this is a summary 1790 and does not include a new protocol or functionality. 1792 10. IANA Considerations 1794 This memo makes no requests of IANA. 1796 11. Acknowledgements 1798 Thanks to our reviewers, Natasha Rooney, Kevin Smith, Ashutosh Dutta, 1799 Brandon Williams, Jean-Michel Combes, Nalini Elkins, Paul Barrett, 1800 Badri Subramanyan, Igor Lubashev, Suresh Krishnan, Dave Dolson, 1801 Mohamed Boucadair, Stephen Farrell, Warren Kumari, Alia Atlas, Roman 1802 Danyliw, and Mirja Kuhlewind for their editorial and content 1803 suggestions. Surya K. Kovvali provided material for section 7. 1804 Chris Morrow and Nik Teague provided reviews and updates specific to 1805 the DoS fingerprinting text. 1807 12. Informative References 1809 [ACCORD] "Acord BoF IETF95 https://www.ietf.org/proceedings/95/ 1810 accord.html". 1812 [CAIDA] "CAIDA *Anonymized Internet Traces* 1813 [http://www.caida.org/data/overview/ and 1814 http://www.caida.org/data/passive/ 1815 passive_2016_dataset.xml]". 1817 [DarkMail] 1818 "The Dark Mail Technical Aliance https://darkmail.info/". 1820 [DOTS] https://datatracker.ietf.org/wg/dots/charter/, "DDoS Open 1821 Threat Signaling IETF Working Group". 1823 [EFF] "Electronic Frontier Foundation https://www.eff.org/". 1825 [EFF2014] "EFF Report on STARTTLS Downgrade Attacks 1826 https://www.eff.org/deeplinks/2014/11/starttls-downgrade- 1827 attacks". 1829 [Enrich] Narseo Vallina-Rodriguez, et al., "Header Enrichment or 1830 ISP Enrichment? Emerging Privacy Threats in Mobile 1831 Networks, Hot Middlebox'15, August 17-21 2015, London, 1832 United Kingdom", 2015. 1834 [I-D.dolson-plus-middlebox-benefits] 1835 Dolson, D., Snellman, J., Boucadair, M., and C. Jacquenet, 1836 "Beneficial Functions of Middleboxes", draft-dolson-plus- 1837 middlebox-benefits-03 (work in progress), March 2017. 1839 [I-D.ietf-ippm-6man-pdm-option] 1840 Elkins, N., Hamilton, R., and m. mackermann@bcbsm.com, 1841 "IPv6 Performance and Diagnostic Metrics (PDM) Destination 1842 Option", draft-ietf-ippm-6man-pdm-option-13 (work in 1843 progress), June 2017. 1845 [I-D.ietf-mile-iodef-guidance] 1846 Kampanakis, P. and M. Suzuki, "IODEF Usage Guidance", 1847 draft-ietf-mile-iodef-guidance-10 (work in progress), May 1848 2017. 1850 [I-D.thomson-http-bc] 1851 Thomson, M., Eriksson, G., and C. Holmberg, "Caching 1852 Secure HTTP Content using Blind Caches", draft-thomson- 1853 http-bc-01 (work in progress), October 2016. 1855 [JNSLP] Surveillance, Vol. 8 No. 3, "10 Standards for Oversight 1856 and Transparency of National Intelligence Services 1857 http://jnslp.com/". 1859 [M3AAWG] "Messaging, Malware, Mobile Anti-Abuse Working Group 1860 (M3AAWG) https://www.maawg.org/". 1862 [Map3GPP] http://www.3gpp.org/technologies, "Mapping between 1863 technologies and specifications". 1865 [NoEncrypt] 1866 "ISPs Removing their Customers EMail Encryption 1867 https://www.eff.org/deeplinks/2014/11/starttls-downgrade- 1868 attacks/". 1870 [Nygren] https://blogs.akamai.com/2017/03/ reaching-toward- 1871 universal-tls-sni.html, "Erik Nygren, personal reference". 1873 [RFC2131] Droms, R., "Dynamic Host Configuration Protocol", 1874 RFC 2131, DOI 10.17487/RFC2131, March 1997, 1875 . 1877 [RFC2504] Guttman, E., Leong, L., and G. Malkin, "Users' Security 1878 Handbook", FYI 34, RFC 2504, DOI 10.17487/RFC2504, 1879 February 1999, . 1881 [RFC2663] Srisuresh, P. and M. Holdrege, "IP Network Address 1882 Translator (NAT) Terminology and Considerations", 1883 RFC 2663, DOI 10.17487/RFC2663, August 1999, 1884 . 1886 [RFC2775] Carpenter, B., "Internet Transparency", RFC 2775, 1887 DOI 10.17487/RFC2775, February 2000, 1888 . 1890 [RFC2804] IAB and IESG, "IETF Policy on Wiretapping", RFC 2804, 1891 DOI 10.17487/RFC2804, May 2000, 1892 . 1894 [RFC2827] Ferguson, P. and D. Senie, "Network Ingress Filtering: 1895 Defeating Denial of Service Attacks which employ IP Source 1896 Address Spoofing", BCP 38, RFC 2827, DOI 10.17487/RFC2827, 1897 May 2000, . 1899 [RFC3135] Border, J., Kojo, M., Griner, J., Montenegro, G., and Z. 1900 Shelby, "Performance Enhancing Proxies Intended to 1901 Mitigate Link-Related Degradations", RFC 3135, 1902 DOI 10.17487/RFC3135, June 2001, 1903 . 1905 [RFC3315] Droms, R., Ed., Bound, J., Volz, B., Lemon, T., Perkins, 1906 C., and M. Carney, "Dynamic Host Configuration Protocol 1907 for IPv6 (DHCPv6)", RFC 3315, DOI 10.17487/RFC3315, July 1908 2003, . 1910 [RFC3550] Schulzrinne, H., Casner, S., Frederick, R., and V. 1911 Jacobson, "RTP: A Transport Protocol for Real-Time 1912 Applications", STD 64, RFC 3550, DOI 10.17487/RFC3550, 1913 July 2003, . 1915 [RFC3971] Arkko, J., Ed., Kempf, J., Zill, B., and P. Nikander, 1916 "SEcure Neighbor Discovery (SEND)", RFC 3971, 1917 DOI 10.17487/RFC3971, March 2005, 1918 . 1920 [RFC4787] Audet, F., Ed. and C. Jennings, "Network Address 1921 Translation (NAT) Behavioral Requirements for Unicast 1922 UDP", BCP 127, RFC 4787, DOI 10.17487/RFC4787, January 1923 2007, . 1925 [RFC4862] Thomson, S., Narten, T., and T. Jinmei, "IPv6 Stateless 1926 Address Autoconfiguration", RFC 4862, 1927 DOI 10.17487/RFC4862, September 2007, 1928 . 1930 [RFC5965] Shafranovich, Y., Levine, J., and M. Kucherawy, "An 1931 Extensible Format for Email Feedback Reports", RFC 5965, 1932 DOI 10.17487/RFC5965, August 2010, 1933 . 1935 [RFC6269] Ford, M., Ed., Boucadair, M., Durand, A., Levis, P., and 1936 P. Roberts, "Issues with IP Address Sharing", RFC 6269, 1937 DOI 10.17487/RFC6269, June 2011, 1938 . 1940 [RFC6430] Li, K. and B. Leiba, "Email Feedback Report Type Value: 1941 not-spam", RFC 6430, DOI 10.17487/RFC6430, November 2011, 1942 . 1944 [RFC6455] Fette, I. and A. Melnikov, "The WebSocket Protocol", 1945 RFC 6455, DOI 10.17487/RFC6455, December 2011, 1946 . 1948 [RFC6590] Falk, J., Ed. and M. Kucherawy, Ed., "Redaction of 1949 Potentially Sensitive Data from Mail Abuse Reports", 1950 RFC 6590, DOI 10.17487/RFC6590, April 2012, 1951 . 1953 [RFC6591] Fontana, H., "Authentication Failure Reporting Using the 1954 Abuse Reporting Format", RFC 6591, DOI 10.17487/RFC6591, 1955 April 2012, . 1957 [RFC6650] Falk, J. and M. Kucherawy, Ed., "Creation and Use of Email 1958 Feedback Reports: An Applicability Statement for the Abuse 1959 Reporting Format (ARF)", RFC 6650, DOI 10.17487/RFC6650, 1960 June 2012, . 1962 [RFC6651] Kucherawy, M., "Extensions to DomainKeys Identified Mail 1963 (DKIM) for Failure Reporting", RFC 6651, 1964 DOI 10.17487/RFC6651, June 2012, 1965 . 1967 [RFC6652] Kitterman, S., "Sender Policy Framework (SPF) 1968 Authentication Failure Reporting Using the Abuse Reporting 1969 Format", RFC 6652, DOI 10.17487/RFC6652, June 2012, 1970 . 1972 [RFC7143] Chadalapaka, M., Satran, J., Meth, K., and D. Black, 1973 "Internet Small Computer System Interface (iSCSI) Protocol 1974 (Consolidated)", RFC 7143, DOI 10.17487/RFC7143, April 1975 2014, . 1977 [RFC7146] Black, D. and P. Koning, "Securing Block Storage Protocols 1978 over IP: RFC 3723 Requirements Update for IPsec v3", 1979 RFC 7146, DOI 10.17487/RFC7146, April 2014, 1980 . 1982 [RFC7230] Fielding, R., Ed. and J. Reschke, Ed., "Hypertext Transfer 1983 Protocol (HTTP/1.1): Message Syntax and Routing", 1984 RFC 7230, DOI 10.17487/RFC7230, June 2014, 1985 . 1987 [RFC7234] Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke, 1988 Ed., "Hypertext Transfer Protocol (HTTP/1.1): Caching", 1989 RFC 7234, DOI 10.17487/RFC7234, June 2014, 1990 . 1992 [RFC7258] Farrell, S. and H. Tschofenig, "Pervasive Monitoring Is an 1993 Attack", BCP 188, RFC 7258, DOI 10.17487/RFC7258, May 1994 2014, . 1996 [RFC7348] Mahalingam, M., Dutt, D., Duda, K., Agarwal, P., Kreeger, 1997 L., Sridhar, T., Bursell, M., and C. Wright, "Virtual 1998 eXtensible Local Area Network (VXLAN): A Framework for 1999 Overlaying Virtualized Layer 2 Networks over Layer 3 2000 Networks", RFC 7348, DOI 10.17487/RFC7348, August 2014, 2001 . 2003 [RFC7435] Dukhovni, V., "Opportunistic Security: Some Protection 2004 Most of the Time", RFC 7435, DOI 10.17487/RFC7435, 2005 December 2014, . 2007 [RFC7457] Sheffer, Y., Holz, R., and P. Saint-Andre, "Summarizing 2008 Known Attacks on Transport Layer Security (TLS) and 2009 Datagram TLS (DTLS)", RFC 7457, DOI 10.17487/RFC7457, 2010 February 2015, . 2012 [RFC7525] Sheffer, Y., Holz, R., and P. Saint-Andre, 2013 "Recommendations for Secure Use of Transport Layer 2014 Security (TLS) and Datagram Transport Layer Security 2015 (DTLS)", BCP 195, RFC 7525, DOI 10.17487/RFC7525, May 2016 2015, . 2018 [RFC7540] Belshe, M., Peon, R., and M. Thomson, Ed., "Hypertext 2019 Transfer Protocol Version 2 (HTTP/2)", RFC 7540, 2020 DOI 10.17487/RFC7540, May 2015, 2021 . 2023 [RFC7619] Smyslov, V. and P. Wouters, "The NULL Authentication 2024 Method in the Internet Key Exchange Protocol Version 2 2025 (IKEv2)", RFC 7619, DOI 10.17487/RFC7619, August 2015, 2026 . 2028 [RFC7624] Barnes, R., Schneier, B., Jennings, C., Hardie, T., 2029 Trammell, B., Huitema, C., and D. Borkmann, 2030 "Confidentiality in the Face of Pervasive Surveillance: A 2031 Threat Model and Problem Statement", RFC 7624, 2032 DOI 10.17487/RFC7624, August 2015, 2033 . 2035 [RFC7754] Barnes, R., Cooper, A., Kolkman, O., Thaler, D., and E. 2036 Nordmark, "Technical Considerations for Internet Service 2037 Blocking and Filtering", RFC 7754, DOI 10.17487/RFC7754, 2038 March 2016, . 2040 [RFC7799] Morton, A., "Active and Passive Metrics and Methods (with 2041 Hybrid Types In-Between)", RFC 7799, DOI 10.17487/RFC7799, 2042 May 2016, . 2044 [RFC7826] Schulzrinne, H., Rao, A., Lanphier, R., Westerlund, M., 2045 and M. Stiemerling, Ed., "Real-Time Streaming Protocol 2046 Version 2.0", RFC 7826, DOI 10.17487/RFC7826, December 2047 2016, . 2049 [RFC7858] Hu, Z., Zhu, L., Heidemann, J., Mankin, A., Wessels, D., 2050 and P. Hoffman, "Specification for DNS over Transport 2051 Layer Security (TLS)", RFC 7858, DOI 10.17487/RFC7858, May 2052 2016, . 2054 [RFC8073] Moriarty, K. and M. Ford, "Coordinating Attack Response at 2055 Internet Scale (CARIS) Workshop Report", RFC 8073, 2056 DOI 10.17487/RFC8073, March 2017, 2057 . 2059 [RFCEdit] https://www.rfc-editor.org/materials/abbrev.expansion.txt, 2060 "RFC Editor Abbreviation List". 2062 [SACM] https://datatracker.ietf.org/wg/sacm/charter/, "Security 2063 Automation and Continuous Monitoring (sacm) IETF Working 2064 Group". 2066 [TS3GPP] "3GPP TS 24.301, "Non-Access-Stratum (NAS) protocol for 2067 Evolved Packet System (EPS); Stage 3"", 2017. 2069 [Vocab] https://portal.3gpp.org/desktopmodules/Specifications/ 2070 SpecificationDetails.aspx?specificationId=558, "3GPP TR 2071 21.905 V13.1.0 (2016-06) Vocabulary for 3GPP 2072 Specifications". 2074 [Web3GPP] http://www.3gpp.org/technologies/95-keywords-acronyms, 2075 "3GPP Web pages on specific topics of interest". 2077 [WebCache] 2078 Xing Xu, et al., "Investigating Transparent Web Proxies in 2079 Cellular Networks, Passive and Active Measurement 2080 Conference (PAM)", 2015. 2082 Authors' Addresses 2084 Kathleen Moriarty 2085 Dell EMC 2086 176 South St 2087 Hopkinton, MA 2088 USA 2090 Phone: +1 2091 Email: Kathleen.Moriarty@dell.com 2092 Al Morton 2093 AT&T Labs 2094 200 Laurel Avenue South 2095 Middletown,, NJ 07748 2096 USA 2098 Phone: +1 732 420 1571 2099 Fax: +1 732 368 1192 2100 Email: acmorton@att.com