This paper studies model-inversion attacks, in which the access to a model is abused to infer information about the training data. Since its first introduction by~\cite{fredrikson2014privacy}, such attacks have raised serious concerns given that training data usually contain privacy sensitive information. Thus far, successful model-inversion attacks have only been demonstrated on simple models, such as linear regression and logistic regression. Previous attempts to invert neural networks, even the ones with simple architectures, have failed to produce convincing results. Here we present a novel attack method, termed the \emph{generative model-inversion attack}, which can invert deep neural networks with high success rates. Rather than reconstructing private training data from scratch, we leverage partial public information, which can be very generic, to learn a distributional prior via generative adversarial networks (GANs) and use it to guide the inversion process. Moreover, we theoretically prove that a model's predictive power and its vulnerability to inversion attacks are indeed two sides of the same coin---highly predictive models are able to establish a strong correlation between features and labels, which coincides exactly with what an adversary exploits to mount the attacks. Our extensive experiments demonstrate that the proposed attack improves identification accuracy over the existing work by about $75\%$ for reconstructing face images from a state-of-the-art face recognition classifier. We also show that differential privacy, in its canonical form, is of little avail to defend against our attacks.

/pdf/the-secret-revealer-generative-model-inversion-attacks-4bizqps1ct.pdf

The Secret Revealer: Generative Model-Inversion Attacks Against Deep Neural Networks

The privacy of individuals does not only depend on their own actions and data but may also be affected by the privacy decisions and by the data shared by other individuals. This interdependence is an essential aspect of privacy and ignoring it can lead to serious privacy violations. In this survey, we summarize and analyze research on interdependent privacy risks and on the associated (cooperative and non-cooperative) solutions. We also demonstrate that interdependent privacy has been studied in isolation in different research communities. By doing so, we systematize knowledge on interdependent privacy research and provide insights on how this research should be conducted and which challenges it should address.

/pdf/a-survey-on-interdependent-privacy-2wp6tdyda2.pdf

A Survey on Interdependent Privacy

Information shared through Social Media may affect more than one user’s privacy e.g., Information that depicts different users, comments that mention different users, events in which different users are invited, etc. Many types of privacy management support in present mainstream Social Media foundation makes users unable to appropriately control the sender and receiver. Computational mechanisms that are able to merge the privacy preferences of different users into a single policy for an item can help solve this problem. Merging different user’s personal preferences is difficult hence conflicts occur in privacy preferences, so methods to resolve conflicts are needed. Moreover, these techniques need to consider how users’ would actually reach an engagement about a solution to the conflict in order to propose solutions that can be acceptable by all of the users affected by the information to be shared. present approaches are either too demanding or only consider fixed ways of aggregating privacy preferences. Here, we introduce the basic computational procedure to overcome problems in Social Media that is able to adapt to different situations by modeling the concessions that users make to reach answers to the conflicts. The present results of a user study in which our introduced mechanism outperformed other present approaches in terms of how many times each approach matched users’ action. Computational mechanisms that are able to merge the privacy preferences of multiple users into a single policy for an item can help solve this problem. However, merging multiple users’ privacy preferences is not an easy task, because privacy preferences may conflict, so methods to resolve conflicts are needed. Moreover, these methods need to consider how users’ would actually reach an agreement about a solution to the conflict in order to propose solutions that can be acceptable by all of the users affected by the item to be shared. Current approaches are either too demanding or only consider fixed ways of aggregating privacy preferences. In this project, we propose the first computational mechanism to resolve conflicts for multi-party privacy management in Social Media that is able to adapt to different situations by modeling the concessions that users make to reach a solution to the conflicts. We also present results of a user study in which our proposed mechanism outperformed other existing approaches in terms of how many times each approach matched users’ behavior.

/pdf/resolving-multi-party-privacy-conflicts-in-social-media-32ca6vy5y5.pdf

Resolving Multi-party Privacy Conflicts in Social Media

Photo sharing on Online Social Networks (OSNs) has become one of the most popular social activities in our daily life. However, some associated friends or bystanders in the photos may not want to be viewed due to privacy concerns. In this paper, we propose the design, implementation and evaluation of HideMe, a framework to preserve the associated users’ privacy for online photo sharing. HideMe acts as a plugin to existing photo sharing OSNs, and it enables the following: a) extraction of factors when users upload their photos, b) associated friends in the uploaded photos are able to set their own privacy policies based on scenarios, instead of a photo-by-photo setting, c) any user in other friend’s uploaded photos could be hidden away from unwanted viewers based on one time policy generation. We also design a distance-based algorithm to identify and protect the privacy of bystanders. Moreover, HideMe not only protects users’ privacy but also reduces the system overhead by a carefully designed face matching algorithm. We have implemented a prototype of HideMe, and evaluation results have demonstrated its effectiveness and efficiency.

HideMe: Privacy-Preserving Photo Sharing on Social Networks

Privacy Computing: Concept, Computing Framework, and Future Development Trends

The QoS and privacy trade-off of adversarial deep learning: An evolutionary game approach

The massive data collection, transmission, storage and processing in IoT are challenging to the cloud computing environment. Aiming at the problem of data sharing and privacy protection of IoT, this paper designed an IoT data sharing model that is based on the edge computing service. The model establishes the virtual data management service by the data abstraction in the edge service layer to provide data service for IoT devices, and further proposed a privacy preserving scheme for data sharing based on attribute encryption. The scheme realized anonymous data sharing and access control and finally is proved to be secure and has a good performance.

An IoT data sharing privacy preserving scheme

In instant messaging systems, photos have become an expressive form of real-time content sharing among users. Unfortunately, faces of stakeholders, who are depicted in the shared photos, may be exposed to unexpected viewers. This may lead to privacy leakages. In this paper, an approach is proposed to prevent such leakages, which is based on access control and face recognition. Each time a photo is sent, all stakeholders are recognized, and their faces are hidden from viewers. At the same time, stakeholders will be notified about the photo, and can decide to reveal their own faces to some proper viewers later. On the other hand, similar photos are exploited to automate this revealing process when possible, so as to reduce the impact of access control on user experiences. Besides, a proof-of-concept system is constructed and its performance is evaluated. The results indicate that this approach can be applied in instant messaging systems without too much overhead.

Sun Zhe

Papers

HideMe: Privacy-Preserving Photo Sharing on Social Networks

The QoS and privacy trade-off of adversarial deep learning: An evolutionary game approach

An IoT data sharing privacy preserving scheme

A Privacy-Preserving Method for Photo Sharing in Instant Message Systems

SRIM Scheme: An Impression-Management Scheme for Privacy-Aware Photo-Sharing Users