Digital Neighborhood Watch:
Investigating the Sharing of
Sensed Data amongst Neighbors
A.J. Brush
Jaeyeon Jung
Ratul Mahajan
Frank Martinez
Inspiration
http://spotcrime.com/wa/seattle
Digital Neighborhood Watch
+
Digital Neighborhood Watch
1) Providing digital evidence of incidents
2) Increasing visibility of neighborhood activity
3) Automatically detecting and alerting for
suspicious events
Social Feasibility?
Does the perceived value of DNW
outweigh the privacy and security risks
of sharing camera data?
Related Work Highlights
Community mailing lists facilitate interaction and
collective action
“Netville” & Blacksburg Electronic Village Studies
Technology that supports community policing
should facilitate communication and discussion
Lewis & Lewis, CHI 2012
Value of human monitoring of CCTV to detect incidents
and challenges around managing digital evidence
Tullio et. al, CHI 2010
Social Feasibility?
Does the perceived value of DNW
outweigh the privacy and security risks
of sharing camera data?
Method
Interviews with 11 households:
• Block watch/Neighborhood
association members
• 5 in High-crime neighborhood
• 6 with security systems
Cameras @ house
Share with Neighbors?
High-crime
neighborhood
In Region
Sharing Chart
Sharing options
Low-resolution
“Always”, “Sometimes” (by request or under specific circumstances), “Never”
Sharing options
“Always”, “Sometimes” (by request or under specific circumstances), “Never”
Limitations
• Participants reactions could be speculation
– Interface mock-ups, maps of their own house,
showed data from existing surveillance cameras
– Recruited households in neighborhood groups
• Relatively small numbers of households in a
few neighborhoods
Household Sensing Systems Desired
• All households except one interested in cameras
• 4 outdoor cameras (median):
Front yard, back door, side yards
• Audio data for gunshots, glass breaking (9 HH)
• Security and non-security benefits
Trust Not Proximity for Sharing Existence
Hypothesis: People would treat households in
view of cameras differently
Wrong!
Neighbors grouped based on trust
Existence of cameras shared with trusted neighbors (9HH)
5 households renamed proximate neighbors to “trusted
neighbors” in chart
Concerns of perceived spying,
dangerous households.
Reluctance to Share Field of View
Divulging Field of View shows
neighbors the blind spots
No household would share
field of view with
neighborhood group
Foreground only was
introduced to address this
concern
Reluctance to Share Field of View
Divulging Field of View shows
neighbors the blind spots
No household would share
field of view with
neighborhood group
Foreground only was
introduced to address this
concern
we want to tell that we have a
camera, we just don’t want to
tell them where it is. And this
way we could show them the
guy in the yellow shirt, but
they wouldn’t know [the FoV]”
Digital Neighborhood Watch
1) Providing digital evidence of incidents
2) Increasing visibility of neighborhood activity
3) Automatically detecting and alerting for
suspicious events
Share Evidence With Reason
All ten households would share for evidence
! Good reason required
“if it was a murder, whether I
didn’t like you or not, I’d
probably still share that
information”
One more comfortable with police than
neighbors
“I feel more comfortable sharing
the images with the police
department because they’re the
ones who need to take the action.”
Limited Willingness to Share Access
• Three households while
on vacation
• Two would trade data
• “Public-only” view was
more acceptable to a few
households
– Three would always share
with trusted neighbors
and law enforcement
Cautious Interest in
Suspicious Activity Detection
• Five households (4 high-crime)
would share events on ongoing
basis for pattern detection
• Seven share high-fidelity for
specific requests (police looking
for car, child)
• Lots of concerns about how it
would actually work, false
alerts…
Neighborhood Types vs.
Acceptability of DNW
• Broad similarity in attitudes we found in high and
low crime neighborhoods.
– equally willing to share data for evidence
– interest in sensing audio was present in both types
• Unique to high crime neighborhood:
– willingness to share events with a detection service on
an ongoing basis (4 of the 5 households from the high
crime neighborhood were interested)
– hidden cameras
Complex Neighborhood Trust
• Let neighbors build their own sub-networks
• Increased concern about renters
Implications for other
neighborhood sharing systems?
• Other neighborhood data could be shared
– water, electricity, commute patterns, …
• Our findings suggest avenues of exploration:
– Share based on specific events rather than
ongoing anonymized information
– Look for abstraction that preserve appropriate
levels of privacy.
Work in Progress
• Social Feasibility
– Cautiously optimistic, particularly with
processed views
• Technical Feasibility
– What can be recognized from video or audio
semi-reliably?
– Soon running in our houses
More generally, this is one possible
application on
Thanks.
A.J. Brush
Jaeyeon Jung
Ratul Mahajan
Frank Martinez