Privacy Risk Models

Jason Hong, Jennifer D. Ng, Scott Lederer and James A. Landay present their framework for modelling privacy risks in ubiquitous computing environments.

The privacy risk models framework consists of two parts: privacy risk analysis, that proposes a list of questions to help defining the context of use of the future application and the privacy risk management, which is a cost-benefit analysis that is used to prioritise the privacy risks and develop the system.

Privacy risk analysis

The privacy risk analysis starts with the formulation of the following questions grouped in the categories Social and Organisational Context and Technology:

Social and Organizational Context

  • Who are the users of the system? Who are the data sharers, the people sharing personal information? Who are the data observers, the people that see that personal information?
  • What kinds of personal information are shared? Under what circumstances?
  • What is the value proposition for sharing personal information?
  • What are the relationships between data sharers and data observers? What is the relevant level, nature, and symmetry of trust? What incentives do data observers have to protect data sharers’ personal information (or not, as the case may be)?
  • Is there the potential for malicious data observers (e.g., spammers and stalkers)? What kinds of personal information are they interested in?
  • Are there other stakeholders or third parties that might be directly
    or indirectly impacted by the system?


  • How is personal information collected? Who has control over the
  • computers and sensors used to collect information?
  • How is personal information shared? Is it opt-in or is it opt-out (or do data sharers even have a choice at all)? Do data sharers push personal information to data observers? Or do data observers pull personal information from data sharers?
  • How much information is shared? Is it discrete and one-time? Is it continuous?
  • What is the quality of the information shared? With respect to space, is the data at the room, building, street, or neighborhood level? With respect to time, is it real-time, or is it several hours or even days old? With respect to identity, is it a specific person, a pseudonym, or anonymous?
  • How long is personal data retained? Where is it stored? Who has access to it?

Privacy Risk Management

This part consists on the prioritisation of privacy risks applying the inequality known as the Hand’s rule.

C < L×D


  • L: The likelihood that an unwanted disclosure of personal information occurs
  • D: The damage that will happen on such a disclosure
  • C: The cost of protecting this privacy in an adequate manner


J. I. Hong, J. D. Ng, S. Lederer, and J. A. Landay, “Privacy risk models for designing privacy-sensitive ubiquitous computing systems,” in Proceedings of the 5th conference on Designing interactive systems: processes, practices, methods, and techniques, 2004, pp. 91–100.

Three-Layer Privacy Responsibility Framework

Sarah Spiekermann and Lorrie Faith Cranor in their work “Engineering Privacy” state that software engineers have a major responsibility when it comes to developing privacy-friendly systems “because they are the ones devising the technical architecture and creating the code”. They present the three-layer model of user privacy concerns and responsibility framework. Based on this model they elaborate a set of guidelines, categorising them in “privacy-by-policy” and “privacy-by-architecture”

Three-Layer Privacy Responsibility Framework

The authors distinguish from three spheres of privacy: User Sphere (constrained to the user environment, i.e. laptop, mobile phone, integrated systems etc), Recipient Sphere (company centric sphere involving their back-ends infrastructure) and Joint Sphere (related to companies that host users information, like email or facebook). For each of the privacy layers, the following table describes where is the data stored, what is the responsibility of the engineer and what are the issues that they need to face.


Privacy Spheres Where Data is Stored Engineer’s Responsibility Engineering issues

User Sphere

Users’ desktop personal computers, laptops, mobile phones, RFID chips

  • Give users control over access to themselves (in terms of access to data and attention)
  • What data is transferred from the client to a data recipient?
  • Is the user explicitly involved in the transfer?
  • Is the user aware of remote and/or local application storing data on his system?
  • Is data storage transient or persistent?

Joint Sphere

Web service provider’s servers and databases

  • Give users some control over access to themselves (in terms of access to data and attention)
  • Minimize users’ future privacy risks
  • Is the user fully aware of how his data is used and can he control this?

Recipient Sphere

Any data recipients: servers and databases of network providers, service providers or other parties with whom data recipient shares data

  • Minimize users’ future privacy risks
  • What data is being shared by the data recipient with other parties?
  • Can the user expect or anticipate a transfer of his data by the recipient?
  • Is personal Data adequately secured?
  • Is data storage transient or persistent?
  • Can the processing of personal data be foreseen by the user?
  • Are there secondary uses of data that may not be foreseen by the user?
  • Is there a way to minimize processing? (e.g. by delegating some pre-processing to User Sphere)

Framework for Privacy-Friendly System Design

Spiekermann and Cranor propose a framework to develop privacy friendly systems. There is a rank of privacy levels lowest to highest that corresponds to the degree of identifiability (identified, pseudonymous, anonymous) of a user. In the cases where the user is totally identified, privacy needs to be provided by policy, while, in those cases where users are anonymous or pseudonymous, privacy can also be provided by architecture. The following table matches this attributes with the characteristics of the corresponding systems.


Privacy stages identifiability Approach to privacy protection Linkability of data to personal identifiers System Characteristics



privacy by policy (notice and choice)


  • unique identifiers across databases
  • contact information stored with profile information



linkable with reasonable & automatable effort

  • no unique identifiers across databases
  • common attributes across databases
  • contact information stored separately from profile or transaction information


privacy by architecture

not linkable with reasonable effort

  • no unique identifiers across databases
  • no common attributes across databases
  • random identifiers
  • contact information stored separately from profile or transaction information
  • collection of long term person characteristics on a low level of granularity
  • technically enforced deletion of profile details at regular intervals




  • no collection of contact information
  • no collection of long term person characteristics
  • k-anonymity with large value of k


S. Spiekermann and L. F. Cranor, “Engineering privacy,” IEEE Transactions on software engineering, vol. 35, no. 1, pp. 67–82, 2009.