Privacy in the Age of Ubiquitous Computing Jason I. Hong Scott Lederer Jennifer Ng Anind K. Dey James A. Landay G r o u p f o r User Interface Research University of California Berkeley
Mar The Origins of Ubiquitous Computing What’s wrong with Personal Computers? –Too complex and hard to use –Too demanding of attention –Too isolating from other people –Too dominating of our desktops and our lives Ubiquitous Computing Project at Xerox PARC –Advances in wireless networking, sensors, devices –Observations of how people use tools in practice –Make computers a natural part of everyday interactions
Mar The Origins of Ubiquitous Computing
Mar Emerging Examples of Ubicomp Never Get Lost Find Friends Emergency Response
Mar “But What About My Privacy?” Never Get Lost –You walk past a restaurant and your cellphone rings with the specials of the day Find Friends –“Family is already very close to you, so if they’re checking up on you…sort of already smothering and this is one step further.” –“[It] could tell when you were in the bathroom, when you left the unit, and how long and where you ate your lunch. EXACTLY what you are afraid of.” Emergency Response –“I don’t see how a government or an organization will not come up with an excuse to use [location info] for another purpose.” Flood of Location-Based Spam Never Hide From Friends and Co-Workers Constant Surveillance
Mar Our Research in Ubicomp Privacy Fundamental Tension –Ubiquitous Computing can be used for great benefit –Ubiquitous Computing can be used for great harm –Privacy may be greatest barrier to long-term success Our approach –What are the privacy concerns in ubicomp? –How can we design better user interfaces? –Are there better ways of building privacy-sensitive apps? Privacy is an area easy to make mistakes in –Will discuss lessons learned throughout
Mar What is Privacy? Lots of perspectives on privacy –US Constitution, UN Decl. Human Rights, Hippocratic Oath –Influenced by Legal, Market, Social, and Technical forces Privacy is not just Orwell –From “Big Brother” to “Little Sisters” –Media sensationalization of worst-case scenarios Privacy is not just computer security –Adversaries? Friends, family, co-workers, acquaintances –Anonymity? Friends already know your identity –Secrecy? We share personal info with friends all the time –Damage? Risk may be undesired social obligations We are approaching privacy from an HCI perspective
Mar An HCI Perspective on Privacy “The problem, while often couched in terms of privacy, is really one of control. If the computational system is invisible as well as extensive, it becomes hard to know: – what is controlling what – what is connected to what – where information is flowing – how it is being used – what is broken (vs what is working correctly)” The Origins of Ubiquitous Computing Research at PARC in the Late 1980s Weiser, Gold, Brown Make it easy to share: the right information with the right people (or service) at the right time
Mar What are End-User Privacy Needs? Lots of speculation about privacy, little data out there Analyzed survey of 130 people on ubicomp privacy prefs Analyzed nurse message board on locator systems – Examined papers describing usage of ubicomp systems Examined existing and proposed privacy protection laws –EU Directive, Location Privacy Act 2001, Wireless Privacy Act 2004 Interviewed 20 people on various location-based services –Did not mention the word “privacy” unless they did first
Mar End-User Privacy Needs Value proposition Simple and appropriate control and feedback Plausible deniability Limited retention of data Decentralized architectures Special exceptions for emergencies Alice’s Location Bob’s Location
Mar How to Design for Privacy? What are good privacy-sensitive user interfaces? –Knowing what is needed does not say how to do it well
Mar Five Pitfalls for Designers Understanding Obscuring potential information flow Obscuring actual information flow Action Configuration over action Lacking coarse-grained control Inhibiting established practices
Mar #1 – Obscuring Potential Flow Users can make informed use of a system only when they understand the scope of its privacy implications
Mar #2 – Obscuring Actual Flow Users should understand what information is being disclosed to whom Who is querying my location? How often? Requestor informed of disclosure Requestee sees each request
Mar #3 – Configuration Over Action Designs should not require excessive configuration to manage privacy –“Right” configuration hard to predict in advance –Make privacy a natural part of the interaction flow
Mar #4 – Lacking Coarse-Grain Control Designs should not forego an obvious, top-level mechanism for halting and resuming disclosure “[T]raveling employees may want their bosses to be able to locate them during the day but not after 5 p.m. Others may want to receive coupons from coffee shops before 9 a.m. on weekdays but not on weekends when they sleep in. Some may want their friends alerted only when they are within one mile, but not 10 miles.” Protecting the Cellphone User's Right to Hide NYTimes Feb Did I set it right? How do I know?
Mar #5 – Inhibiting Established Practices Designs should not inhibit users from transferring established social practices to emerging technologies Rather than getting an immediate ring, an answering machine comes on the line and says, "Lee has been motionless in a dim place with high ambient sound for the last 45 minutes. Continue with call or leave a message." 1. University and Ramona 2. Palo Alto 3. Custom… 9.Ignore for now
Mar How to Build Applications Better? Develop a toolkit to make it easier to build privacy- sensitive ubicomp apps –Prevent – Strong guarantees on your personal data –Avoid – Better user interfaces for managing privacy –Detect – Finding over-monitoring or accidental disclosures –Need all three for effective systems Key architectural points of Context Fabric –Locality –InfoSpace Diary –Access Descriptions –Privacy Tags Show time-based issues Applications tend to have one of three dominant interaction patterns Can do main work of privacy management before, during, or after Which to use depends on the app and community of users Add how the work on previous slides influenced this work Simpler UIs / Architecture Add more technical depth here Hitting this problem at the UI Level and the Systems Level Really need both, can’t do one without other Also, tried to get lots of low-fi user feedback throughout the process Add lo-fi prototype screenshots or mention it Describe distinction between push / pull Push is easier since you control when it’s sent out Need preview of why prevent / avoid / detect is right Vs just Prevent Quote: trust person or not Focus on alternatives and why those didn’t work
Mar Locality Keep personal data “close” to end-users –Move from centralized systems to decentralized ones –Capture, store, and process personal data on my computer PlaceLab ABC Add more placelab guts here Add slide on MiniGis here as well What are the alternatives here? Centralization was an issue early on Also talked to a lawyer about this (Deidre) Closer data is to you, more legal protection – Works indoors and in urban canyons – No special equipment – Privacy-sensitive – Rides the WiFi wave
Mar PlaceLab ~52000 Nodes (3Megs)
Mar PlaceLab ~500 Nodes
Mar Locality MiniGis Server for processing location locally Country Name= United States Region Name= California City Name= Berkeley ZIP Code= Place Name= Soda Hall Lat Lon= ,
Mar Locality MiniGis Server data sources USGS State Gazetteer –Names in USA –2m records ~650 megs –States, Cities, Places “Places” hardest to get –Airports & schools useful, “hammocks”, “lava”, “quicksand” less so –2 undergrads scouring Berkeley –Research opportunity here in open, distributed naming of places GEOnet Names Server –Names outside USA –5.5m records ~700megs –Regions, Cities, Places
Mar InfoSpace Diary InfoSpace stores your personal information –Static info, like name and phone# –Dynamic info, like current location and activity Runs on your personal device or on a trusted service –Local sources (ex. PlaceLab) can update dynamic info –Can choose to expose different parts to different people & services –Can also see who can see what about you Describe guts of InfoSpace more Describe Tuples more Operators too?
Mar Confab Architecture InfoSpace Diary InfoSpace Diary LocNamePlaceLabTourguideFind FriendMiniGis How to control when and how much personal info is disclosed? Request
Mar Access Notifications One possibility: “[T]raveling employees may want their bosses to be able to locate them during the day but not after 5 p.m. Others may want to receive coupons from coffee shops before 9 a.m. on weekdays but not on weekends when they sleep in. Some may want their friends alerted only when they are within one mile, but not 10 miles.” Problems: –People are not good at defining rules beforehand –Tradeoff between fine-grained control and understandability
Mar Observations on Setting Preferences Who is requesting information is most important factor –“Either I trust someone with my information or I don't – it doesn't depend on where I am.” Time is an essential aspect for maintaining control –Access described in terms of “always”, “never”, or “work hours” –“Work people can know my information during work hours. Home/SO people can know my information always.” Can set prefs before, during, or after a request –Before case can lead to configuration pitfall –During case easier to understand, but can overwhelm –After case easy to setup, but can lead to accidental disclosures
Mar Access Notifications Describe more rationale here Evaluation of this? Add slide on LOC Explain design rationale Time came from interviews / surveys What are the alternatives? My original version had lots of fine-grained mechanisms IP Address Location Time Problem was that from interviews, people didn’t seem to do things that way “I trust person” or “I trust service” between “these times” So I fell into the pitfall too! Second iteration, organized things by People and Services Two different kinds of Access Descriptions Persons and Services Push and Pull as well Push not too hard (greater control, end-user initiated) Pull harder, since not initiated by end-user
Mar Access Notifications Initial Evaluations –Tested with 4 people for understandability and reactions –Location-enhanced messenger, tourguide, emergency response
Mar Access Notifications Initial Evaluations –Tested with 4 people for understandability and reactions –Location-enhanced messenger, tourguide, emergency response Results –Distinction between Push vs Pull, Continuous vs Discrete “Giving a GPS location once or twice does not provide enough information for an invasion of privacy… [but] if GPS location is shared every 2 seconds, there is a potential for an invasion of privacy.” “No need for continuous update of location. Only in a race or a marathon (where staying on track is essential) would continuous update be helpful.”
Mar Access Notifications PlaceBar Continuous Discrete Push??? PullAccess Notifications
Mar Confab Architecture InfoSpace Diary InfoSpace Diary LocNamePlaceLab Tourguide Find Friend MiniGis How to control what happens to your info once it leaves your InfoSpace? Access Pull Push
Mar Privacy Tags Digital Rights Management for Privacy –Like adding note to , “Please don’t forward” –Notify address- –Time to live- 5 days –Max number of sightings- last 5 sightings of my location Libraries for making it easy for app developers
Mar Analysis Prevent –Capture and process personal information locally –PlaceLab, MiniGis –Minimizes risk of mission creep (ex. SSNs) Avoid –Interfaces for helping people make good decisions –Access Notifications / PlaceBar Detect –Finding cases of over-monitoring –Access Notifications –Privacy Tags (processed on requestor’s side)
Mar Implementation Confab, PlaceLab, MiniGis –Java 1.5, Tomcat Web Server, MySql, Jaxen XPath Data –WiFi from wigle.net and undergrads –MiniGis from USGS, GeoNET, and undergrads –~35 megs of data (30 megs of place data) #Classes Lines of code Confab PlaceLab MiniGis Shared Libs
Mar Putting it Together Lemming Location-enhanced Messenger Describe Rationale as to why location + messenger? Show more guts Describe LOC
Mar Putting it Together BEARS Emergency Response Server Field studies and interviews with firefighters [CHI2004] Finding victims in a building –“You bet we’d definitely want that.” –“It would help to know what floor they are on.” But emergencies are rare –How to balance privacy constraints with utility when needed?
Mar Putting it Together BEARS Emergency Response Server Trusted third party (MedicAlert++) Data Sharer Location Building BEARS Service Link 1 2 Trusted BEARS Third- Party Trusted BEARS Third- Party Location 3 4 Medic Alert++ Medic Alert++ Loc “ABC” On Emergency
Mar Requirements Check Value proposition Simple and appropriate control and feedback –Access Notifications (pull) and PlaceBar (push) Plausible deniability –No action, “Ignore for now”, and “Never Allow” appear same Limited retention of data –Privacy Tags Decentralized architectures –Capture and process information locally Special exceptions for emergencies
Mar Contributions Investigated ubicomp privacy from many perspectives –What are end-user needs? How to design? How to build? Context Fabric architecture for privacy-aware apps –Prevent / Avoid / Detect –Suggests a way of architecting privacy-sensitive ubicomp Services on devices, local processing, presentation to end-users –Evaluation with two applications –Starting deployment of Lemming instant messenger “Use technology correctly to enhance life. It is important that people have a choice in how much information can be disclosed. Then the technology is useful.”
Jason I. Hong G r o u p f o r User Interface Research University of California Berkeley Thanks to: DARPA Expeditions NSF ITR PARC Intel Fellowship Siebel Systems Fellowship