idnits 2.17.1 draft-johansson-linkability-bad-01.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 Security Considerations section. ** The document seems to lack an IANA Considerations section. (See Section 2.2 of https://www.ietf.org/id-info/checklist for how to handle the case when there are no actions for IANA.) Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (January 21, 2014) is 3720 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- No issues found here. Summary: 2 errors (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 General L. Johansson 3 Internet-Draft SUNET 4 Intended status: Informational L. Nordberg 5 Expires: July 25, 2014 NORDUnet 6 January 21, 2014 8 Linkability Considered Harmful 9 draft-johansson-linkability-bad-01 11 Abstract 13 Current debate on pervasive monitoring often focus on passive attacks 14 on the protocol and transport layers but even if these issues were 15 eliminated through the judicious use of encryption, roughly the same 16 information would still be available to an attacker who is able to 17 (legally or otherwise) obtain access to linked data sets which are 18 being maintained by large content and service providers. 20 Status of This Memo 22 This Internet-Draft is submitted in full conformance with the 23 provisions of BCP 78 and BCP 79. 25 Internet-Drafts are working documents of the Internet Engineering 26 Task Force (IETF). Note that other groups may also distribute 27 working documents as Internet-Drafts. The list of current Internet- 28 Drafts is at http://datatracker.ietf.org/drafts/current/. 30 Internet-Drafts are draft documents valid for a maximum of six months 31 and may be updated, replaced, or obsoleted by other documents at any 32 time. It is inappropriate to use Internet-Drafts as reference 33 material or to cite them other than as "work in progress." 35 This Internet-Draft will expire on July 25, 2014. 37 Copyright Notice 39 Copyright (c) 2014 IETF Trust and the persons identified as the 40 document authors. All rights reserved. 42 This document is subject to BCP 78 and the IETF Trust's Legal 43 Provisions Relating to IETF Documents 44 (http://trustee.ietf.org/license-info) in effect on the date of 45 publication of this document. Please review these documents 46 carefully, as they describe your rights and restrictions with respect 47 to this document. Code Components extracted from this document must 48 include Simplified BSD License text as described in Section 4.e of 49 the Trust Legal Provisions and are provided without warranty as 50 described in the Simplified BSD License. 52 1. Introduction 54 This I-D is submitted as a position paper for the joint IAB/W3C 55 STRINT workshop 2014. The authors wishes to call attention to the 56 fact that linked data sets are a source of information, sometimes 57 every bit as useful as anything that can be gleaned from passive 58 monitoring of Internet traffic. Such data sets are routinely 59 generated and maintained by service and content providers and are 60 often a source of secondary (or even primary) income for those that 61 own and generate them. 63 In the current discussion on pervasive monitoring we often overlook 64 the fact that even as more encryption is used, making passive attacks 65 harder, focus may simply shift to attacks on owners of linked data 66 sets. We should strike at the root of this problem by making it less 67 appealing to maintain these data sets and by offering users a measure 68 of control over how their information is used and shared. 70 Linkability is by no means a new concept and the authors do not 71 propose to (re)define the concept in this draft. Instead our intent 72 is to show, using some simple examples, how linkability occurs in 73 practice and what effect linkability has on privacy on the Internet. 75 2. A Simple Example 77 Service providers (we use this term in a general sense, and not with 78 a view to any particular protocols etc) typically manage users and 79 billing records. This leads to a data set being created for every 80 user of that service. Most services employ a simple pattern for user 81 enrollment which relies on an email address as a means of 82 (supposedly) uniquely identifying the user. The email address has 83 become the defacto user identifier on the Internet. 85 When a user pays for the service a pair of linked data sets is 86 created: the user data at the service provider is associated (via the 87 credit card information) to the user data held by the credit card 88 company. The value of the linked data, as well as the risk to the 89 user, is higher than the value/risk involved in the two data sets 90 taken as separate entities. For instance the linked data says 91 something about the buying habits of the user (based on the use of 92 the particular credit card) which in itself is valuable information. 94 Linking increases the risk to the user as well. With every service 95 that stores the users credit card the risk of exposure to active 96 attacks increase as events in recent years have made it painfully 97 clear. 99 If this example seems overly simplified or even naive to bring up, 100 consider the simple observation that when we visit a store in the 101 physical world we have the ability to "browse", i.e to view and 102 select among the offered goods without having to identify ourselves 103 or prove our ability to pay for any of the goods in the store. This 104 aspect of the real world has not been translated into the online 105 world where prospective customers are routinely fingerprinted and our 106 behaviour tracked even when we have shown no intention of engaging in 107 a business transaction with the store owner. 109 Naturally there must be ways to "conduct business on the Internet", 110 but there are ways to enable business without the need for linkable 111 attributes. In fact there are ways to enable business using non- 112 linkable attributes in such a way that the risk to business owners 113 are reduced. 115 3. Avoiding Linkable Attributes 117 The way to avoid linking is simple (and yet so hard in practice): 118 avoid the use of linkable attributes. In our e-commerce example 119 above, the credit card number is a linkable attribute. However in 120 this case the credit card is strictly speaking not needed at the 121 service provider. When the user provides her credit card information 122 to the service provider she is actually providing an authorization to 123 the service giving the service provider the right to obtain payment 124 from the credit card company. 126 Instead of using the credit card number as an implicit grant (of a 127 right to obtain payment), a token that isn't linkable across 128 identifier domains could be used to represent an explicit grant 129 issued on behalf of the user by the credit card company to the 130 service provider. This is a simple example of a general pattern: 131 instead of using a linkable user identifier, provide access to an 132 attribute representing some property of the user that used to grant 133 specific access. 135 Some credit card companies have actually taken first steps towards 136 this by involving the user directly in a second factor authentication 137 (typically to reduce the risk of fraud). This practice follows a 138 model for 3:rd party authentication services (aka identity providers) 139 commonly used in the enterprise and R&E community. Experience from 140 the R&E identity federation community shows that access control using 141 identity providers and non-linkable pseudonymous identifiers is by no 142 means problem free, but can be made to work in many situations. 144 4. Linkability and Probability 146 Links between data sets do not necessarily have to have be uniquely 147 linkable in order to blur the line between private and public actions 148 on the Internet. Linkability is really about probability and often 149 absolute certainty (demonstrated in the example above) is not needed 150 for an adversary to conclude that two actions were likely performed 151 by the same client. 153 Web browsers are typically fingerprintable even when the user tries 154 to avoid sticking out when compared to other users of a given web 155 site. Regardless of encryption, the operator of a web site can more 156 often than not tell one user from another by looking at information 157 sent by the browser without the users knowledge. 159 Some examples of sources for fingerprintability in web browsers are 160 information about browser window and desktop resolution, browser 161 toolbar presence, title bar font size and window manager settings 162 [TORBUTTON-DESIGN]. This has been confirmed by empirical studies 163 like the [PANOPTICLICK] study done by the EFF. 165 5. Incentives for Collecting Data 167 There are strong incentives for service providers to enrich the value 168 of their data set using attribute linking. The value of the 169 attribute naturally increase with the inverse of the size of the set 170 of users who share that attribute: the more specific the attribute 171 the more valuable it is, because it can be used to profile a user 172 with a higher degree of certainty. 174 Unfortunately there seem to be few costs associated with keeping 175 large linked data sets around - stolen user credentials in the 10s of 176 thousands rarely result in more than a brief notice in the news 177 anymore. To date the IETF community have focused on how to avoid the 178 use of long-term credentials (passwords) to reduce the effects of 179 such attacks. We need to broaden our scope to find ways to 180 disincentivize the (over)use of linkable attributes. 182 6. The Least common Denominator of Privacy 184 Out of all the transactions that a given user performs over the time 185 her traffic is being observed by an adversary, the most linkable one 186 will define her level of privacy towards that adversary. In other 187 words, linkability is the least common denominator of privacy and 188 must be treated as the privacy killer that it is. When we allow for 189 too much linkability in protocols we must acknowledge the fact that 190 we're building something that can't provide privacy. 192 Part of the Internet economy seems to be based on linked data sets 193 and linkable attributes. Changing this will require creating 194 negative incentives for service providers, making it less attractive 195 to keep data around as well as establishing technical mechanisms that 196 allow service providers access to the attributes they do need in 197 order to conduct their business without having to rely on linkable 198 attributes. Success will depend on carefully engineering the 199 negative incentives to match the technical mechanisms in order to 200 promote good behaviour. 202 The authors believe that the IETF community should attempt to design 203 technical controls into existing and future protocols that make it 204 possible for users of Internet technology to have a choice when to 205 provide linkable data to services and eavesdroppers and when not to. 207 7. Acknowledgements 209 Many thanks to Nick Mathewson for important contributions on the 210 topic of linkability. Many thanks also to Lucy Lynch who is the 211 source of much wisdom, the "I'm just browsing" response to 212 identification on the web in particular. 214 8. Informative References 216 [TORBUTTON-DESIGN] 217 Perry, M., "Torbutton Design Documentation", n.d., . 220 [PANOPTICLICK] 221 Eckersley, P., "How Unique Is Your Web Browser?", n.d., 222 . 224 Authors' Addresses 226 Leif Johansson 227 SUNET 229 Email: leifj@sunet.se 230 Linus Nordberg 231 NORDUnet 233 Email: linus@nordu.net