Preventing Camera Recording by Designing a Capture-Resistant Environment.

Full text

Page 1

Preventing Camera Recording by Designing a Capture-
Resistant Environment
Khai N. Truong, Shwetak N. Patel, Jay W. Summet and Gregory D. Abowd
College of Computing & GVU Center
Georgia Institute of Technology
Atlanta, GA 30324-0280 USA
{khai, shwetak, summetj, abowd}@cc.gatech.edu
Abstract. With the ubiquity of camera phones, it is now possible to capture
digital still and moving images anywhere, raising a legitimate concern for many
organizations and individuals. Although legal and social boundaries can curb
the capture of sensitive information, it sometimes is neither practical nor
desirable to follow the option of confiscating the capture device from an
individual. We present the design and proof of concept implementation of a
capture-resistant environment that prevents the recording of still and moving
images without requiring any cooperation on the part of the capturing device or
its operator. Our solution involves a tracking system that uses computer vision
for locating any number of retro-reflective CCD or CMOS camera sensors in a
protected area. A pulsing light is then directed at the lens, distorting any
imagery the camera records. Although the directed light interferes with the
camera's operation, it can be designed to minimally impact the sight of other
humans in the environment.
1 Introduction and Motivation
By the last quarter of 2004, approximately 75% percent of mobile phones in Japan
were camera phones; it is expected this number will saturate at 75%-85% in 2005.1
By 2006, more than 80 percent of mobile phones shipped in the United States and
Western Europe will have cameras.2 Camera phones, and related consumer
technologies, make it easy to capture still and moving images anywhere, creating a
legitimate concern among those who wish to retain some level of privacy or secrecy.
Companies concerned that camera phones compromise the security of their
intellectual property often ban such devices from their facilities. These confiscation
practices, however, are not always desirable or practical. Although some legal
controls and social boundaries may curb inappropriate capture behaviors [2, 3, 8], we
believe technological solutions can safeguard against undesired recording without
requiring confiscation by an authority or cooperation by the public at large.

1 Source: Eurotechnology, Japan K.K., Camera phones: disruptive innovation for imaging,
Market and trend report, 5th Version of October 11, 2004
2 Source: Gartner Inc. http://www3.gartner.com/press_releases/pr3mar2004a.html

Page 2

2 Khai N. Truong, Shwetak N. Patel, Jay W. Summet and Gregory D. Abowd
Previous work addresses this challenge by disabling recording features in the
cameras [5, 6, 9]. In this paper, we present an alternative that requires neither
instrumentation nor control of the recording device. Instead, we present a technique
for safeguarding the environment itself against recording, creating a so-called
“capture-resistant” environment. Our system detects camera phones in the
environment and emits a strong localized light beam at each device to neutralize it
from capturing. Although our approach does have limitations, its main strength is that
it requires no cooperation on the part of the camera or its owner and it minimally
disturbs the natural viewing experience by the human eye.
2 Related Work
Technical solutions have been proposed to prevent or to react to undesired camera
capture. Most of these solutions require some sort of instrumentation of a capture
device. For example, solutions, such as Safe Haven, leverage the short-range wireless
capability available on camera phones (such as Bluetooth or WiFi) to allow the
environment to notify the device that the space does not allow photography or other
forms of recording [5, 6, 9]. There are many drawbacks to this solution. First, it
assumes that the user of the camera would install and use special software on the
device and that she would abide by the environmental constraints. Hewlett-Packard
has proposed a paparazzi-proof camera [7] that automatically modifies images when
it receives commands from a remote device. This camera includes a facial recognition
feature that selectively blurs certain parts of an image. Other approaches also require
different forms of cooperation on the part of the camera or its operator. The Cloak
system addresses privacy concerns with surveillance cameras by having users carry a
“privacy enabling device” (PED) [1]. This device informs the environment that any
footage of the carrier of this device must be sanitized at a later time. A solution called
“Eagle Eye” couples a light sensor to a flash unit [4]. When a flash of light is
detected, this small wearable device instantaneously flashes back. This technique
obscures a portion of the photographic image, similar to the approach described in
this paper. However, Eagle Eye only works against still, flash photography.
We take a significantly different approach from these previous solutions in the
design of our capture-resistant environment. First, rather than requiring users to trust
cameras to sanitize images after the recording has occurred, we actively impede
recording at the point of capture, as with Eagle Eye. Second, unlike many previous
solutions, our system does not rely on any cooperation or instrumentation on the part
of the capture devices or the people operating them. Finally, our solution addresses
video capture in addition to still imagery.
We initially focused on protecting stationary regions of an environment, such as a
wall. Surfaces in an environment can be covered to prevent capture, but then the
surfaces would not be visible. There are numerous commercially-available retro-
reflective sprays and shields that can also be placed over a surface to reflect light and
flashes in a manner that prevents recording; for example, these products are intended
to prevent traffic cameras from capturing the license plate of a car running a red light.
These solutions create glares that impact visibility from the human eye as well as the

Page 3

Preventing Camera Recording by Designing a Capture-Resistant Environment 3
camera sensor. Our solution minimally impacts what an observer in the environment
sees while still preventing a camera from being able to record.
3 Design Goals for a Capture-Resistant Environment
Our primary goal in addressing this problem was to design an environment that
prevents certain portions of that space from being captured by mobile phones that
include a CCD or CMOS camera.3 This motivation, and review of past related work,
highlights the major design goals for building a capture-resistant environment. These
are:
•
elimination of the need for cooperation or control of the recording devices
before, during or after capture;
•
prevention of the capture of both still images and video; and
•
minimal impact to the view of the environment by the naked human eye.

In addition, this approach should allow for two interesting improvements:
•
the ability to allow authorized cameras to record; and
•
the possibility of making mobile entities (e.g., people) similarly capture-
resistant.

Our design uses a combination of computer vision and projection, described in the
next section, to actively search for cameras and systematically block them from
recording clear pictures, as opposed to relying on removal or alteration of content
later. We envision uses of our system for situations like conferences, tradeshows and
museums. For example, some artists want to prevent people taking pictures of their
artwork. Similarly, companies may want to limit capture of early prototypes and
designs on display in research laboratories or exhibits at tradeshows.
4 A Capture-Resistant Environment
In this section, we present our capture-resistant environment, which consists of two
components. First, a camera detector actively tracks CCD sensors in the environment.
When the system detects a camera’s CCD sensor, the second system component, the
camera neutralizer, directs a localized beam of light at each camera’s lens to obstruct
its view of the scene. For each component, we describe the theory of operation and
our proof of concept implementation. We then critically evaluate the limitations of
this prototype, distinguishing the theoretical limits from the current engineering
limitations of our specific implementation and discuss how we can extend our system.

3 CCD and CMOS cameras both use semi-conductor based sensors. Our approach works
against both types. We will refer to this category of cameras as “CCD cameras” throughout
the rest of the paper.

Page 4

4 Khai N. Truong, Shwetak N. Patel, Jay W. Summet and Gregory D. Abowd


(top view) (side view)

Figure 1. System diagram. When a user introduces a camera into the capture-resistant
environment, a camera detector component locates the device within its field of view and the
camera neutralizer component emits a localized light beam (yellow) at the camera to block the
camera’s view of a portion of the surface the system attempts to guard from capture. The red
bar indicates the protected surface. The blue indicates the field of view of the user’s camera.
The pink indicates the camera neutralizer’s field of influence. Dashed lines indicate the portion
of the protected surface that is affected by the neutralizer’s light beam.
4.1 Detecting Cameras in the Environment
CCD cameras have an optical property that produces well-defined light reflections.
By tracking these reflections, we can effectively locate and track cameras.
4.1.1

Our camera detector leverages the retro-reflective property of the CCD sensor found
on most consumer-level digital cameras. Retro-reflection causes light to reflect
directly back to its source, independent of its incident angle. This effect is often
noticed on photographs when the camera flash can make a subject’s eyes appear to
glow red, caused by the retro-reflective property of the retina at the back of the eye.
Commercial applications of retro-reflection include traffic signs and reflective clothes
commonly worn by road construction workers.
CCD sensors are mounted at the focal plane of the camera’s optical lens, making
them very effective retro-reflectors. Although many objects in the environment
exhibit this property, they are typically imperfect retro-reflectors and can be
distinguished from CCD cameras as we demonstrate in Section 5. By tracking these
retro-reflections we can detect and track cameras pointed at a given area.
Theory of Operation

Page 5

Preventing Camera Recording by Designing a Capture-Resistant Environment 5
4.1.2

To detect cameras in the environment, we used a Sony Digital HandyCam video
camera placed in NightShot mode. IR transmitters surround the lens and a narrow
bandpass IR filter covers the detector’s lens (Figure 3a). This instrumentation,
referred to as the detector, projects an IR light beam outwards from the camera and
detects any retro-reflective surfaces within the field of view. The specific placement
of the IR illuminator around the perimeter of the detector’s lens ensures a bright retro-
reflection from cameras within the field of view of the detector. The detected CCD
cameras can be pointed directly at it or tilted away at slight angles (which we
computed to be up to roughly ±20°). This retro-reflection appears as a bright white
circular speckle through the IR filtered camera (Figure 2).
We detect reflections by simply locating bright regions in the camera view above a
certain luminance threshold (Figure 2). By using a thresholding technique, there is no
limit to the number of the cameras that can be detected within the cross-section of the
camera detector. In the next section, we discuss the handling of both false positives
and false negatives.
Our system effectively tracks cameras at a rate of 15 Hz. A more powerful
computer could track at 30 Hz, however 15 Hz is sufficient because a user must hold
the average camera still for at least this period of time to avoid motion blur in her
picture.
The camera detector has approximately a 45° field of view. Reflections from
cameras of varying shapes and sizes can be detected up to 10 meters away. In our
proof of concept, at 5 meters away, the cross-section of the detector camera’s field of
view is roughly a 4m width x 3m height area. Although a zoom lens can be added to a
camera, we estimate that 5 meters is roughly the length of a reasonably-sized room.
Room sizes and walls naturally prevent people from recording our capture-resistant
environment from afar. Our current proof of concept only involves a single detector
unit. To ensure that we can detect cameras from all angles, we can measure the angle
at which users can approach the surface. Accordingly, we can determine how many
detector units we must use to cover that range. We can add additional detectors
throughout the environment to find cameras from farther away if needed.
Implementation

(a) (b)
Figure 2. On the left is an unprocessed IR view captured by our camera detector with plenty of
ambient light in the room. A person holds a camera phone pointed at a region in the
environment we want to protect from capture. On the right is the processed view. The camera is
detected by locating a bright white circular speckle.