260 likes | 268 Views
This research paper discusses the limitations of current privacy preferences on social media platforms and proposes a unified model for privacy preferences that can be exported and reused across platforms. The implementation of this model using the OpenSocial framework is also explored.
E N D
Guarding a Walled Garden - Semantic Privacy Preferences for the Social WebPhilipp Kärger and Wolf SiberskiL3S Research Center, Hannover, Germany Philipp Kärger - L3S Research Center
Outline • Observations and Problem Statement • A Unified Privacy Model • Implementation using OpenSocial • Future Work and Conclusions Philipp Kärger - L3S Research Center
Observations and Problem Statement Philipp Kärger - L3S Research Center
Privacy Decisions: who and what • Who is requesting? • Do I know the guy? • Did I ever meet him? • Did I ever talk to him? • Is he a colleague? • Didn’t I meet him at ESWC? • What is requested? • Is the requested thing/action private? • Do I want to share the requested thing? • Isn't it the picture I took once at this nice ESWC talk? Philipp Kärger - L3S Research Center
The Who and the What reflected on the Social Web: Philipp Kärger - L3S Research Center
But what about … • You are allowed if you … • posted in my forum. • are my friend in either Facebook or Skype. • are in one of my Orkut groups. • are in any of my Flickr pictures. • attended ESWC. • like the same music. • are interested in Semantic Web. • are listed in my FOAF profile. … the social context? Philipp Kärger - L3S Research Center
Privacy Preferences are Trapped • Platforms share similar concepts like • friends • groups • messages and chats • profile information • blocked users • shared resources • But privacy preferences cannot be exported and reused. Philipp Kärger - L3S Research Center
Summary: MotivationPrivacy Preferences on the Social Web • are restricted to predefined categories (fixed Whos and Whats) • cannot refer to social context - although it is available (Walled Garden No. 1) • cannot be exported and reused crossing platforms(Walled Garden No. 2) Philipp Kärger - L3S Research Center
A Unified Model for Privacy Preferences Philipp Kärger - L3S Research Center
“Somebody is allowed to do/see something.” • A Privacy Preference is a mapping from • What to Who • Set of actions/objects to Set of persons • Object category to Subject category Philipp Kärger - L3S Research Center
Object and Subject Categories • users shall be allowed to define • new subject categories • e.g., “A family member is every person who is in my FOAF profile or in my Facebook group ‘family’.” • family_member(X) :- foafFriend(X) ; inFacebookGroup(X,’family’). • new object categories • e.g., “A party picture is everything which is a picture and tagged with ‘party’.” • party_picture(X) :- picture(X) , taggedWith(X,’party’). Philipp Kärger - L3S Research Center
And-Or-Tree visualization of categories • Object Categories – the “What” Subject Categories – the “Who” Philipp Kärger - L3S Research Center
User-defined Mappings on Categories enforcement happens acc. to the Mappings common vocabulary unified format Philipp Kärger - L3S Research Center
Summary • new categories can be defined by rules party_picture(X) :- picture(X) , taggedWith(X,’party’). • mappings between these categories define privacy preferences Every party_picture can be accessed by every friend. • platform independent when exploiting a suitable vocabulary Philipp Kärger - L3S Research Center
Implementation Philipp Kärger - L3S Research Center
Goals • a uniform representation of Privacy Preferences RDF • extend standard Social Web platform to enforce such preferences the OpenSocial Implementation Apache Shindig • support the integration of arbitrary social contexts for the preferences • Policy Engine “Protune” plus Social Web wrapper Philipp Kärger - L3S Research Center
Apache Shindig – an Open Social Implementation • Open Social – an API for Social Web sites defining methods for • retrieving personal information (profile data) • getting activity notification • storing/retrieving application data • sending/receiving messages • Apache Shindig • container for hosting Social Web applications • implements Open Social API • basis to easily develop Social Web platforms Philipp Kärger - L3S Research Center
Shindig Extension • filters requests (Open Social API calls) for information about persons • interprets our Privacy Preference Model • Protune is used as policy engine for privacy enforcement • privacy enforcement calls external sources used to check social context family_member(X) :- foafFriend(X) ; inFacebookGroup(X,’family’). Philipp Kärger - L3S Research Center
Social Context • arbitrary RDF sources, e.g., FOAF, SIOC, DOAP • relationships on Twitter • friendships on Flickr • any OpenSocial information (e.g., age, current location, address) • co-authorship on DBLP (via the DBLP SPARQL endpoint) Privacy Preferences can relate to Philipp Kärger - L3S Research Center
Implementation Summary • each Shindig-based platform can adopt it • reuse and share RDF format of Privacy Preferences • integrate of social context gathered from the Web Philipp Kärger - L3S Research Center
Future Work & Conclusions Philipp Kärger - L3S Research Center
Future Work • Authentication and identification • FOAF+SSL • OpenID • certificates, credentials, Trust Negotiation • User Interface Philipp Kärger - L3S Research Center
Conclusions • Social Web Privacy Preferences • share a simple scheme • are limited to concepts of one platform only • are not portable to other platforms • We introduced a new Privacy Preference Model • easily extendable with complex categories • crosses the Walled Garden • Our Implementation • exploits OpenSocial and Protune • uses RDF representation of Privacy Preferences Philipp Kärger - L3S Research Center
Thanks for your attention. • Philipp Kärger • L3S Research Center • kaerger@L3S.de • www.L3S.de/~kaerger Philipp Kärger - L3S Research Center