Needs for Anonymized Mobile Data
120 likes | 229 Vues
Needs for Anonymized Mobile Data. Discussion Topic / Working Group Seminar 08471. What do we need to learn?. Applications Importance Societal Supportable Privacy constraints Knowledge What information must be present in the data? Structure
Needs for Anonymized Mobile Data
E N D
Presentation Transcript
Needs for Anonymized Mobile Data Discussion Topic / Working Group Seminar 08471
What do we need to learn? • Applications • Importance • Societal • Supportable • Privacy constraints • Knowledge • What information must be present in the data? • Structure • How should the data be represented to make learning easy? Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
The Killer App(s) for Anonymized Data • Context and Location Aware Services • When can we have expectation of privacy (sensors)? • Expectation “in a crowd” vs. “in the Wald” • Public Safety • Emergency response, evacuation • Public security / law enforcement • Lookup/location advertising • Business workflows – factory, logistics – real-time response • Traffic / transportation • Mixed-reality games • Enhanced tourism / Edutainment • Location Microdata • Public Safety • Planning • Investigation • Health research • Personal health-related data (e.g., exercise data, environmental sensors) • Epidemiology, pathology • Collaborative filtering / collaborative recommendation • Geomarketing • Business workflows – factory, logistics – real-time response • Urban planning Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Information Required • Frequent vs. outlier • Location vs. trajectory • Data quality • Exact? • Probabilistic? • Generalization of truth? Trajectory Patterns (Dino) example of learning that involves approximation Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Real-time traffic analysis and services(Infomobility): Information Required • Frequent vs. outlier • Outlier events • Frequent normality • Location vs. trajectory • Generally want trajectory, planned destination • Aggregate data largely sufficient • Sometimes point data sufficient (e.g., accident) • Service: Need to know current location, destination • Can this be provided anonymously? • Background information • Road network • Calendar / events • Data quality / Granularity • Granularity: road segment • Outlier events – exact • Frequency – probably want relatively close to exact, particularly when near capacity Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Research on anonymized (geo) Health Info.: Information Required • Geospatial information • Sensor-based / atmospheric conditions • Geography – relevant semantics • Telemedicine – magnifies geospatial variables • Ex: Continuous heart monitoring • Frequent vs. outlier • Outlier population / Adverse Drug Events • Sporadic events (e.g., heart conditions) • Location vs. trajectory • Location@time referenced with conditions • Conditions inferred from trajectory and georeferenced data • Correlation between individuals based on colocation (not necessarily in time) • Data quality • Exact? • Probabilistic? • Generalization of truth? (Don’t tell them what the real data is) • Define policy before technology hits the market Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Privacy and Web 2.0 • Change in sensitivity? • What does privacy mean when people volunteer/publish data? • (Particularly mobile/georeferenced data) • Interplay of privacy and trust • Do people know what they are giving up? • Inference • Archival • Psychological privacy vs. quantifiable risk • Context for privacy • How does integration of other data with location affect privacy? • Anonymity in the presence of external information? Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Seminar Proceedings Killer App • Traffic Data • Health Data Research Web 2.0 outline • Kinds of geospatial self-published data • Uses • Risks / (Mis)uses <above 1 axis, below 2nd axis> • What do we do about this? • Education • Regulation • Policy • Technology • Risk Assessment Research Agenda Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Other “next steps” Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Seminar Proceedings • Killer Apps. for anonymized data • Description • Data needs • Anonymity/privacy • Traffic Data • Health • Privacy in Web 2.0 • What is self-published geospatial data? • Uses/value? • Privacy concerns: • Risk • Perceptions • Recommendations Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Data RepresentationEnable use of existing tools? • Identical to real data • Reconstruct representative trajectories (Saygin, Nergiz, Atzori GIS’08) • Region bounds • Region distributions (PDF) Seminar 08471: Geographic Privacy-Aware Knowledge Discovery and Delivery
Context for Privacy Discussion Topic / Working Group Seminar 08471