Understand the privacy expectations of mobile users: a recommendation-based crowdsourcing approach

Should learn the need, recently read a paper, close-up summary, if there are flaws, also hope to point out.

Directory

Introduction

Referral mechanism

1.1 Why to understand the privacy expectations of mobile users

1. There are some potential privacy threats and information leaks in the widespread use of mobile devices. 2, the system supplier has proposed the corresponding measures to this question, for example: Apple's iOS system lets the user control whether the application can access the specific sensitive data source. The Android platform also has a similar fine-grained permission control mechanism. However, there are drawbacks: not all users have the knowledge background to properly configure their privacy. It is also a tedious and challenging job.　　　　　　　　　　　　　　　　　　　　　　　　　The user experience is not high. 3. There is no simple rule that satisfies all privacy requirements, some people may be willing to provide some information for better service and experience, while others may not want to share sensitive data because of privacy issues. In order to achieve the best balance for each user, it is critical to understand their expectations of privacy and to help them set privacy permissions accordingly. That's why we don't use the security experts ' recommendations that follow.

1.2 What is the crowdsourcing method and the Priwe system

This approach is based on some key insights about how users decide to grant permissions to an application: 1, which depends on the specific privacy preferences or concerns of the user (as opposed to the "user-based" reference later); 2, the decision　　　　　　　　　　　　　　　　　　　　 The expectations of some applications (as opposed to "project-based" mentioned later). The article proposes a method that first compares the user's privacy preferences and privacy expectations, and then recommends appropriate permissions settings to the user based on these similarities. The rationale behind this approach is that users who have similar preferences in certain private data and/or privacy expectations are more likely to make similar decisions in relevant privacy projects. 　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　Pursue a balance between usability and privacy. Priwe Architecture: First, Priwe can help users make better decisions about privacy settings on their smartphones. Second, the process of analyzing crowdsourcing data and generating recommendations should be done on the server due to the limited functionality of the smartphone. As shown in the following:

The server side has two key components: 1, the data of the handset to preprocess, such as authentication and classification; 2. For different users Recommendations for various mobile applications.

2.1 Basic Ideas

The traditional recommendation system is designed to recommend attractive and emotive products to customers in some e-commerce markets.　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　　Therefore, each user corresponds to a customer, and each privacy setting is mapped to a commodity. Collaborative filtering algorithm two main categories: Memory-based and Model-based methods. "User-based" and "project-based" are two key algorithms for the memory-based approach. The memory-based algorithm has the following advantages: 1, non-parametric method, less dependent on the hypothetical model, 2, it is easy to be extended to more High dimensions, easy to calculate and understand, 3, the number of users and projects are more robust, 4, requires a small number of parameters, and the calculation speed is fast.

2.2 Item-and user-based Collaborative filtering

We assume that there are k users, each with a M application. Each application has n data access rights.define RI,A,G as user I sets the settings for application A of data permissions G. binary {0,1},whereri,a,g=0 (indicates that the user does not like to share data with anyone);whereri,a,g=1 (indicates that participants allow public information) Ex Ample 1: Two users, I and J, have installed two applications, a, and each app has two permissions g,h. Both user I and J allow app a to get Data permissions by setting ri,a,g=1& ri,a,h=1 and rj,a,g=1&rj,a,h=1. In this case, we think they may have similar privacy preferences. If user I sets ri,a,g=0 to organize access for application B G, user J may have the same option on this setting. 　　　　　　　　　　　　　　　　　　　　　　

Understanding the privacy expectations of mobile users: a recommended crowdsourcing method