Middleware for Pervasive Spaces: Balancing Privacy and Utility D. Massaguer , B. Hore, M. H. Diallo, S. Mehrotra, and N. Venkatasubramanian Presenter: Daniel Massaguer PhD candidate dani.massaguer@gmail.com Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al Massaguer et al 1 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces Balancing Privacy and Utility for Pervasive Spaces
Cyber-Physical Spaces Control Loop[ Reason Observe Adapt Responsphere: Pervasive Sensing, Computing, Storing, and Communications Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 2 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Cyber-Physical Spaces Control Loop[ Alice, Susan, John Alice Mar John coding Mary lunch Susan y Staff, Staff payroll Daniel, Sharad, Staff, meeting meeting Staff break Staff student, MD-FF1 toClass Reason Professor, reading Office Collaboration Emergency Response Reason Observe Adapt Responsphere: Pervasive Sensing, Computing, Storing, and Communications Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 3 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Sentient Spaces Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 4 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Challenges Programming complexity Due to heterogeneity of Sensors, computers, networks, complex event detection algorithms. Shared cyber-physical infrastructure Used by several applications Shared by people and their activities Real-world changes non-functional requirements of observations Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 5 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
This talk Mechanisms to be able to release observations while protecting privacy of the people in the space [Middleware09] Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 6 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Distributed and Stream Architecture Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 7 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
ODB.Base A Semanic View of the Space for Applications ODB. B.Ba Base Object Id At t Name At t Value Time Alice Location Kitchen1 10:12:50 03/04/09 Alice HeartRate 60 10:12:54 03/04/09 J ohn Location ConfRoom 1 10:12:40 03/04/09 FireTeam Location Kitchen1 10:12:50 03/04/09 FireTeam Location Kitchen2 10:12:51 03/04/09 A virtual table that would contain the latest values observed SELECT * FROM ODB.Base WHERE ObjectId = 'Peter' AND AttName = 'Location'; Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 8 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Privacy John, ? Alice,? Mary, coding Susan, lunch Privacy challenges: Marc, email 1.- Inference Daniel, Sharad, meeting Jason,? ? Jim,? Public knowledge: Hojjat “Alice and Paul always have lunch together.” break Nalini, → Alice is having lunch reading + → Paul is at Alice's office Unknown Location Timy, email Paul, lunch 2.- What is privacy and how do users express it? Office monitor Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 9 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Our Setting observer Application Disclosure {<id,att,value,t>} control Pervasive Infrastructure targets Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 10 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Our Approach Utility-based framework Model privacy as negative utility of query targets Model information requirements as positive utility of observers Utility dynamically specified with policies and utility-elicitation mechanisms Compute Inferable Data Total Privacy is Impossible Closed-world approach Represent background knowledge with pDatalog KB Generalize Data to reduce risk of disclosure Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 11 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Privacy as Negative Utility Intuition: 1.- “some information is more private than other e.g., my location if I am closer to a deadline” 2.- privateness of information depends on consequences of misusage e.g., being interrupted Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 12 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Privacy as Negative Utility EU T (y) = Pr(y | Y rel ∧ BK) * P(y) * neg_utility(y) * ω( y.t) How (un)happy if info is Pr info being (mis)used (mis)used (e.g., being interrupted) EU O (y) = Pr(y | Y rel ∧ GH) * P(y) * pos_utility(y) Now Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 13 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Privacy as Negative Utility EU T (y) = Pr(y | Y rel ∧ BK) * P(y) * neg_utility (y) * ω( y.t) How (un)happy if info is Pr info being (mis)used (mis)used (e.g., being interrupted) EU O (y) = Pr(y | Y rel ∧ GH) * P(y) * pos_utility(y) Y rel : information released Now Y req : information before disclosure control Y i rel : independent partition in Y rel Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 14 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Privacy as Negative Utility EU T (y) = Pr(y | Y rel ∧ BK) * P(y) * neg_utility (y) * ω( y.t) How (un)happy if info is Pr info being (mis)used (mis)used (e.g., being interrupted) EU O (y) = Pr(y | Y rel ∧ GH) * P(y) * pos_utility(y) Y rel : information released Now Y req : information before disclosure control Y i rel : independent partition in Y rel s.t. Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 15 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Background Knowledge pDatalog Knowledge Base (association rules): Tuple(Alice, Location, l, t) : p ∗ 0.8 ← Tuple(Mary, Location, l, t) : p Feasible approach Populated by admins (intended space usage) + learned by system (calibration + rule mining) Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 16 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Background Knowledge pDatalog Knowledge Base (association rules): Tuple(Alice, Location, l, t) : p ∗ 0.8 ← Tuple(Mary, Location, l, t) : p Feasible approach Populated by admins (intended space usage) + learned by system (calibration + rule mining) Identical facts combined with MAX (i.e., worst-case inference) Bo Xing: A Middleware Approach to Information Bo Xing: A Middleware Approach to Information Massaguer et al 17 Dissemination in Heterogeneous Wireless Networks Dissemination in Heterogeneous Wireless Networks Balancing Privacy and Utility for Pervasive Spaces
Recommend
More recommend
