Séminaire de Cryptographie

Accueil     Présentation     Archives

Marc Manulis

Cryptographic Approach to Private Social Clouds

Users increasingly rely on the ``social cloud'' for storing and sharing personal information, for establishing new contacts, and for interacting with their friends and colleagues. Even though social media platforms may differ in the target audience, in the nature of collected and disseminated information, and in services offered to the users, there are several building blocks that enable user's social interactions and are deployed by the majority of these platforms. Designing these building blocks in a secure and privacy-preserving way is of utmost importance, should potential damage from the misuse of user-provided content in the existing ``social cloud'' be averted.

Due to the absence of appropriate models and security/privacy definitions it is, so far, impossible to judge, whether a proposed protection mechanism is suitable or not, whether it can be applied in general, or is tailored to a specific social media platform. Instead of using ``ad-hoc'' and ``best practice'' solutions with questionable guarantees, it is advisable to strive for a more formal treatment of the underlying building blocks by providing sophisticated security/privacy definitions and designing solutions amenable to formal reasoning and security proofs.

In the first part of my talk I will focus on the management of personal user profiles. User profiles serve as a main building block for most social media platforms. I will introduce a formal cryptographic model for \emph{private} user profiles, supporting generation of the digital content owned by the user and its controlled disclosure to other users of the social community. I will present formal definitions of security and privacy that reflect the natural expectations on private profiles. I will further describe two general solutions based on different encryption techniques and highlight their trade-offs from the security/privacy and complexity point of view. These solutions do not require any specific network infrastructure or any trusted third parties. The provided analysis takes into account statistics of several real-world social media platforms. The second part of my talk will address the problem of discovery of common social contacts, for which I will introduce the first privacy-friendly practical solution that lets two users, on input their respective contact lists, learn their common contacts (if any), and nothing else. The proposed protocol prevents arbitrary list manipulation by means of contact certification, and guarantees user authentication and revocability. I will explain why current approaches such as private set intersection or anonymous credentials, although being related, do not provide an appropriate solution to this problem. I will discuss the modeling of contact-hiding security that private contact discovery should provide. I will also highlight efficiency considerations for the proposed protocol that does not require involvement of any trusted third parties and can be deployed in resource-constraint environments.

The content of this talk is based on the recent results from FC 2011/RLCSP, ASIACCS 2011, and ACNS 2011.