| Internet-Draft | DULT Threat Model | September 2024 |
| Delano & Lowell | Expires 22 March 2025 | [Page] |
Lightweight location tracking tags are in wide use to allow users to locate items. These tags function as a component of a crowdsourced tracking network in which devices belonging to other network users (e.g., phones) report which tags they see and their location, thus allowing the owner of the tag to determine where their tag was most recently seen. While there are many legitimate uses of these tags, they are also susceptible to misuse for the purpose of stalking and abuse. A protocol that allows others to detect unwanted location trackers must incorporate an understanding of the unwanted tracking landscape today. This document provides a threat analysis for this purpose, will define what is in and out of scope for the unwanted location tracking protocols, and will provide some design considerations for implementation of protocols to detect unwanted location tracking.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://ietf-wg-dult.github.io/threat-model/draft-ietf-dult-threat-model.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-ietf-dult-threat-model/.¶
Discussion of this document takes place on the Detecting Unwanted Location Trackers Working Group mailing list (mailto:[email protected]), which is archived at https://mailarchive.ietf.org/arch/browse/unwanted-trackers/. Subscribe at https://www.ietf.org/mailman/listinfo/unwanted-trackers/.¶
Source for this draft and an issue tracker can be found at https://github.com/ietf-wg-dult/draft-ietf-dult-threat-model.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 22 March 2025.¶
Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Location tracking tags are widely-used devices that allow users to locate items. These tags function as a component of a crowdsourced tracking network in which devices belonging to other network users (e.g., phones) report on the location of tags they have seen. At a high level, this works as follows:¶
Tags ("accessories") broadcast an advertisement payload containing accessory-specific information. The payload also indicates whether the accessory is separated from its owner and thus potentially lost.¶
Devices belonging to other users ("non-owner devices") observe those payloads and if the payload is in a separated mode, reports its location to some central service.¶
The owner queries the central service for the location of their accessory.¶
A naive implementation of this design exposes both a tag’s user and anyone who might be targeted for location tracking by a tag’s user, to considerable privacy risk. In particular:¶
If accessories simply have a fixed identifier that is reported back to the tracking network, then the central server is able to track any accessory without the user's assistance, which is clearly undesirable.¶
Any attacker who can guess a tag ID can query the central server for its location.¶
An attacker can surreptitiously plant an accessory on a target and thus track them by tracking their "own" accessory.¶
In order to minimize these privacy risks, it is necessary to analyze and be able to model different privacy threats. This document uses a flexible framework to provide analysis and modeling of different threat actors, as well as models of potential victims based on their threat context. It defines how these attacker and victim persona models can be combined into threat models. It is intended to work in concert with the requirements defined in [I-D.detecting-unwanted-location-trackers], which facilitate detection of unwanted tracking tags.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
Incorporation of this threat analysis into the DULT protocol does not introduce any security risks not already inherent in the underlying Bluetooth tracking tag protocols. Existing attempts to prevent unwanted tracking by the owner of a tag have been criticized as potentially making it easier to engage in unwanted tracking of the owner of a tag. However, Beck et al. have demonstrated a technological solution that employs secret sharing and error correction coding.¶
To create a taxonomy of threat actors, we can borrow from Dev et al.’s Models of Applied Privacy (MAP) framework. This framework is intended for organizations and includes organizational threats and taxonomies of potential privacy harms. Therefore, it cannot be applied wholesale. However, its flexibility, general approach to personas, and other elements, are applicable or can be modified to fit the DULT context.¶
The characteristics of threat actors may be described as follows. This is not intended to be a full and definitive taxonomy, but an example of how existing persona modeling concepts can be applied and modified.¶
Expertise level¶
Proximity to victim¶
High: Lives with victim or has easy physical access to victim and/or victim’s possessions.¶
Medium: Has some physical access to the person and possessions of someone who lives with victim, such as when the attacker and victim are co-parenting a child.¶
Low: Does not live with or have physical access to victim and/or victim’s possessions.¶
Access to resources¶
High: The attacker has access to resources that may amplify the impact of other characteristics. These could include, but are not limited to, funds (or control over “shared” funds), persons assisting them in stalking behavior, or employment that provides privileged access to technology or individuals’ personal information.¶
Low: The attacker has access to few or no such resources.¶
In addition, the victim also has characteristics which influence the threat analysis. As with attacker characteristics, these are not intended as a definitive taxonomy.¶
Expertise level¶
Access to resources¶
High: The victim is generally able to safely access practical and relevant resources. These might include funds to pay a car mechanic or private investigator, law enforcement or legal assistance, or other resources.¶
Low: The victim is generally unable to safely access practical and relevant resources. These might include money to pay a car mechanic or private investigator, law enforcement or legal assistance, or other resources.¶
Access to technological safeguards¶
High: The victim is able to safely use, and has access to, technological safeguards such as active scanning apps.¶
Limited: The victim is able to safely use, and has access to, technological safeguards such as active scanning apps, but is unable to use their full capacity.¶
Low: The victim is not able to use technological safeguards such as active scanning apps, due to reasons of safety or access.¶
It is also appropriate to define who is using the tracking tags and incorporate this into a model. This is because if protocols overly deprioritize the privacy of tracking tags’ users, an attacker could use a victim’s own tag to track them. Beck et al. describe a possible technological solution to the problem of user privacy vs privacy of other potential victims.¶
Tracking tag usage¶
The following scenarios are composite cases based upon reports from the field. They are intended to illustrate different angles of the problem. They are not only technological, but meant to provide realistic insights into the constraints of people being targeted through these tags. There is no identifying information for any real person contained within them. In accordance with research on how designers understand personas, the characters are given non-human names without attributes such as gender or race. The analysis of each scenario provides an example usage of the modeling framework described above. It includes a tracking tag usage element for illustrative purposes. However, as discussed previously, this element becomes more or less relevant depending on protocol evolution. Note that once a given attacker persona has been modeled, it could be recombined with a different victim persona, or vice versa, to model a different scenario. For example, a non-expert victim persona could be combined with both non-expert and expert attacker personas.¶
Mango and Avocado have two young children. Mango, Avocado, and the children all use smartphones, but have no specialized technical knowledge. Mango left because Avocado was abusive. They were homeless for a month, and the children have been living with Avocado. They now have an apartment two towns away. They do not want Avocado to know where it is, but they do want to see the children. They and Avocado meet at a public playground. They get there early so that Avocado will not see which bus route they arrived on and keep playing with the children on the playground until after Avocado leaves, so that Avocado will not see which bus route they get on. Two days later, Avocado shows up at Mango’s door, pounding on the door and shouting.¶
In this case, the attacker has planted a tag on a child. Co-parenting after separation is common in cases of intimate partner violence where the former partners have a child together. Child visits can be an opportunity to introduce technology for purposes of stalking the victim.¶
| Attacker Profile | Avocado |
|---|---|
| Expertise Level | Non-Expert |
| Proximity to Victim | Medium |
| Access to Resources | Unknown, but can be presumed higher than Mango’s due to Mango’s recent homelessness |
| Victim Profile | Mango |
|---|---|
| Expertise Level | Non-Expert |
| Access to Resources | Low |
| Access to Technological Safeguards | Normal |
| Other Characteristics | Avocado and Mango |
|---|---|
| Accessory Usage | Attacker Only |
Strawberry and Elderberry live together. Neither has any specialized technological knowledge. Strawberry has noticed that Elderberry has become excessively jealous – every time they go to visit a friend by themselves, Elderberry accuses them of infidelity. To their alarm, over the last week, on multiple occasions, Elderberry has somehow known which friend they visited at any given time and has started to harass the friends. Strawberry eventually gets a notification that a tracker is traveling with them, and thinks it may be in their car, but they cannot find it. They live in a car-dependent area and cannot visit friends without the car, and Elderberry controls all of the “family” money, so their cannot take the car to the mechanic without Elderberry knowing.¶
Here, the attacker and the victim are still cohabiting, and the attacker is monitoring the victim’s independent activities. This would allow the attacker to know if, for instance, the victim went to a police station or a domestic violence agency. The victim has reason to think that they are being tracked, but they cannot find the device. This can happen if the sound emitted by the device is insufficiently loud, and is particularly a risk in a car, where seat cushions or other typical features of a car may provide sound insulation for a hidden tag. The victim could benefit from having a mechanism to increase the volume of the sound emitted by the tag. Another notable feature of this scenario is that because of the cohabitation, the tag will spend most of the time in “near-owner state” as defined by the proposed industry consortium specification [I-D.detecting-unwanted-location-trackers]. In near-owner state it would not provide alerts under that specification.¶
| Attacker Profile | Elderberry |
|---|---|
| Expertise Level | Non-Expert |
| Proximity to Victim | High |
| Access to Resources | High |
| Victim Profile | Strawberry |
|---|---|
| Expertise Level | Non-Expert |
| Access to Resources | Low |
| Access to Technological Safeguards | Impaired (cannot hear alert sound) |
| Other Characteristics | Elderberry and Strawberry |
|---|---|
| Accessory Usage | Attacker Only |
Lime and Lemon have been dating for two years. Lemon works for a tech company and often emphasizes how much more they know about technology than Lime, who works at a restaurant. Lemon insists on having access to Lime’s computer and Android phone so that they can “make sure they are working well and that there are no dangerous apps.” Lemon hits Lime when angry and has threatened to out Lime as gay to their conservative parents and report them to Immigration & Customs Enforcement if Lime “talks back.” Lime met with an advocate at a local domestic violence program to talk about going to their shelter once a bed was available. The advocate did some safety planning with Lime, and mentioned that there is an app for Android that can scan for location trackers, but Lime did not feel safe installing this app because Lemon would see it. The next time Lime went to see the advocate, they chose a time when they knew Lemon had to be at work until late to make sure that Lemon did not follow them, but when Lemon got home from work they knew where Lime had been.¶
This is a case involving a high-skill attacker, with a large skill difference between attacker and victim. This situation often arises in regions with a high concentration of technology industry workers. It also may be more common in ethnic-cultural communities with high representation in the technology industry. In this case the victim is also subject to a very high level of control from the attacker due to their imbalances in technological skills and societal status, and is heavily constrained in their options as a result. It is unsafe for the victim to engage in active scanning, or to receive alerts on their phone. The victim might benefit from being able to log into an account on another phone or a computer and view logs of any recent alerts collected through passive scanning.¶
| Attacker Profile | Lemon |
|---|---|
| Expertise Level | Expert |
| Proximity to Victim | High |
| Access to Resources | High |
| Victim Profile | Lime |
|---|---|
| Expertise Level | Non-Expert |
| Access to Resources | Low |
| Access to Technological Safeguards | Low |
| Other Characteristics | Lemon and Lime |
|---|---|
| Accessory Usage | Attacker Only |
The above taxonomy and threat analysis focus on location tracking tags. They are protocol-independent; if a tag were designed using a technology other than Bluetooth, they would still apply. The key attributes are the functionalities and physical properties of the accessory from the user’s perspective. The accessory must be small enough to be easily concealed, and able to broadcast its location to other consumer devices.¶
The scope of this threat analysis includes any easily-concealable accessory that is able to broadcast its location to other consumer devices.¶
An attacker who attempts to track a victim using a tracking tag and applications readily available for end-users (e.g. native tracking application) is in scope. Additonally, an attacker who physically modifies a tracking tag (e.g. to disable a speaker) is in scope. An atacker who makes non-nation-state level alterations to the firmware of an existing tracking tag or creates a custom device that leverages the crowdsourced tracking network is in scope.¶
All victims profiles are in scope regardless of their expertise, access to resources, or access to technological safeguards. For example, protocols should account for a victim's lack of access to a smartphone, and scenarios in which victims cannot install separate software.¶
There are many types of technology that can be used for location tracking. In many cases, the threat analysis would be similar, as the contexts in which potential attackers and victims exist and use the technology are similar. However, it would be infeasible to attempt to describe a threat analysis for each possible technology in this document. We have therefore limited its scope to location-tracking accessories that are small enough to be easily concealed, and able to broadcast their locations to other devices. The following are out of scope for this document:¶
Attackers with nation-state level expertise and resources who deploy custom or altered tracking tags to bypass protocol safeguards or jailbreak a victim end-device (e.g. smartphone) are considered out of scope.¶
As discussed in Section 3, unwanted location tracking can involve a variety of attacker, victim, and tracking tag profiles. A successful implementation to preventing unwanted location tracking would:¶
Include a variety of approaches to address different scenarios, including active and passive scanning and notifications or sounds¶
Account for scenarios in which the attacker has high expertise, proximity, and/or access to resources within the scope defined in Section 3.2 and Section 3.3¶
Account for scenarios in which the victim has low expertise, access to resources, and/or access to technological safeguards within the scope defined in Section 3.2 and Section 3.3¶
Avoid privacy compromises for the tag owner when protecting against unwanted location tracking using tracking tags¶
This document has no IANA actions.¶
TODO acknowledge.¶