1. SEE IT, SHAKE IT, SET IT
privacy awareness and control for mobile applications
Arosha K. Bandara
The Open University, UK
Mobile East Conference
June 2012
2. RESEARCH CONTEXT
• EPSRC Funded PRiMMA Project:
Privacy Rights Management for Mobile Applications
• Collaboration between
The Open University and Imperial College London
• Contributions include methodologies for understanding privacy
requirements, machine learning techniques, architectures for
privacy aware social networks and design of real-time feedback
mechanisms for privacy awareness and control.
http://primma.open.ac.uk
3. RESEARCH TEAM
• Bashar Nuseibeh • Morris Sloman
• Yvonne Rogers • Alessandra Russo
• Clara Mancini • Emil Lupu
• Arosha K. Bandara • Naranker Dulay
• Blaine Price • Domenico Corapi
• Lukasz Jedrejcyzk • Ryan Wishart
• Keerthi Thomas
• Adam Joinson
4. PRIVACY THEORY
• Bi-directionality (Altmann) Status
Update
• Output: sharing information Location
with others
• Input: sensing activity of
others, previous experience,
etc. 2
1
Photographs
5. PRIVACY THEORY
• Social translucence
(Erickson and Kellog)
• Visibility
• Awareness
• Accountability
• Enforces social norms.
7. RESEARCH CHALLENGES
• Understand people, their
behaviour and requirements.
• Translate this understanding
into solutions.
8. RESEARCH CHALLENGES
• Understand people, their
behaviour and requirements.
• Translate this understanding
into solutions.
• Evaluate solutions ‘in the wild’
9. UNDERSTANDING PEOPLE !"#$%&'($)*+",#-'./$"01
• Investigating mobile privacy is
difficult because ...
... privacy is sensitive and
depends on socio-cultural
context.
... mobility introduces
contextual shifts and logistical
obstacles.
!"##$%&'()*%+,-$".'$%'/)($0"'%"#1)-2$%&
10. UNDERSTANDING PEOPLE
• It is also difficult ...
... for people to articulate
subtle concerns and
preferences.
... for researchers to observe
contextualised behaviour.
11. EXPERIENCE SAMPLING ++
• We address these challenges
by combining a variety of
complementary, indirect
methods:
• Experience sampling
enhanced with memory
phrase.
• Individual, in-depth
deferred contextual
interviews.
12. EXPERIENCE SAMPLING ++
• We address these challenges
by combining a variety of
complementary, indirect
methods:
• Experience sampling
enhanced with memory
phrase.
• Individual, in-depth
deferred contextual
interviews.
13. BUDDY TRACKER
1. Location Contextual
Updates
Real-time
Learning
Alice Engine
3. Notification
2. Location Request
1.
Location
Updates
Bob
17. SEE IT: REAL-TIME FEEDBACK
• Study 1
• Two families with mixture of Week 1 Week 2
relationships. 58% 24%
• Conducted over 3 weeks, with
simple real-time feedback Week 3
introduced in final week. 18%
• Quantitative data from server
logs and qualitative data from
ESM and post-study Location Request
interviews. Frequency
18. SEE IT: REAL-TIME FEEDBACK
• Study 2
Phase 2
• 3 week study with 15 7
participants.
• Context-aware real-time Phase 1
feedback with machine 42
learning in final week.
• Quantitative data from server
logs and qualitative data from
ESM and post-study Frequency of
interviews. ‘intrusive’ feedback events
23. PRIVACY-SHAKE
Study 3 - User evaluation
Male Female
100
75
% Success
50
25
Initialise 0
Increase Privacy
Reduce Privacy
Privacy control task
24. SEE IT, SHAKE IT, SET IT
• Context-aware real-time feedback supports bi-directionality
and social translucence in location sharing applications.
• Machine learning techniques make awareness less intrusive,
leading to greater acceptance of technology.
• Intuitive control mechanisms can be used for privacy control
actions.
• Further work is required to investigate alternative privacy
control interactions - e.g., multi-touch gestures.
25. SEE IT, SHAKE IT, SET IT
privacy awareness and control for mobile applications
Arosha K. Bandara
The Open University, UK
a.k.bandara@open.ac.uk - @arosha
http://primma.open.ac.uk