Surveillance Camera System Balancing Privacy Protection and
Effective Surveillance Image Use
Kent Kobayashi
1
, Masaki Inamura
2
, Kitahiro Kaneda
3
, Keiichi Iwamura
1
and Isao Echizen
4
1
Tokyo University of Science, Niijuku, Tokyo, Japan
2
Tokyo Denki University, Ishizaka, Hiki-gun, Hatoyama-machi, Saitama, Japan
3
Osaka Prefecture University, Sakai-shi, Osaka, Japan
4
National Institute of Informatics, Chiyoda, Hitotsubashi, Tokyo, Japan
Keywords: Privacy, Surveillance Camera, Reversible Mosaic.
Abstract: Privacy protection has been attracting considerable attention in recent years. Several instances of surveillance
video recordings of famous people in public stores being uploaded to the Internet have been reported. Such
instances of privacy infringement have become increasingly concerning. A simple solution to this problem is
to obscure the facial features of individuals being recorded in surveillance camera systems. However, in some
cases where surveillance camera recordings are required, such as criminal investigations, the solution fails.
Therefore, we propose a new surveillance camera system that balances the requirements of privacy protection
and those of cases in which unobscured images are required. Further, we present the protocol of the proposed
system and evaluate the security of the system against attacks.
1 INTRODUCTION
In recent years, with the increasing popularity of
social networking sites (SNSs), opportunities for
sharing various types of media over the Internet, such
as videos and images, have increased. However, this
has led to a greater need for securing and protecting
the privacy and personal information of individuals.
Privacy refers to “private aaffairs, private life, and
personal secrets” and “the right to not infringe them.”
In addition, privacy includes “the right to control
one’s information.” Further, personal information
refers to “personally identifiable information,” such
as name, address, date of birth, and bio-information.
Surveillance cameras have been increasingly
installed in public spaces worldwide for various
purposes, such as traffic monitoring, security, post-
incident analysis, and so on. However, laws
governing the collection and use of private
information, such as facial images captured through
surveillance camera systems, have not been
established in many countries including Japan.
Moreover, in most cases, people being monitored by
surveillance camera systems have not been granted
adequate rights over the use and distribution of their
recorded facial images and information. For example,
several incidents wherein surveillance camera
recordings of a famous person visiting convenience
stores or video rental shops being uploaded to SNSs
by store employees have been reported.
From the perspective of privacy protection, the
person being photographed or video recorded must be
granted the right to control and manage the image or
video. However, surveillance camera images are
often used in criminal investigations. To this end, if
individuals are granted the right to control their
recorded information, criminal investigations might
be obstructed.
Therefore, in this paper, we propose a
surveillance camera system that balances the
requirements of both privacy protection and criminal
investigations. This feature is achieved by combining
“a group signature technique” and “a reversible
mosaic technique that employ reversible watermarks”
in the proposed system. Moreover, we evaluate the
security of our system against attacks.
The remainder of this paper is organized as
follows. In Section 2, we explain privacy
infringement issues associated with surveillance
camera systems. In Section 3, we discuss related
studies, namely, “Short Group Signatures” (Boneh,
2004), a group signature technique, and “Recoverable
original video for mosaic system,” (Kusama, 2015) a
reversible mosaic technique using reversible
watermarks. In Section 4, we explain our proposed
41
Kobayashi K., Inamura M., Kaneda K., Iwamura K. and Echizen I..
Surveillance Camera System Balancing Privacy Protection and Effective Surveillance Image Use.
DOI: 10.5220/0005511100410048
In Proceedings of the 12th International Conference on e-Business (ICE-B-2015), pages 41-48
ISBN: 978-989-758-113-7
Copyright
c
2015 SCITEPRESS (Science and Technology Publications, Lda.)
system, and in Section 5, we discuss some attacks
against the proposed system and security against
those attacks. We conclude the paper in Section 6.
2 PRIVACY INFRINGEMENT
ISSUES ASSOCIATED WITH
SURVEILLANCE CAMERA
Privacy has been interpreted as the “right to not
publish personal information without good reason.”
However, in general, privacy also includes “the right
of a person to control his/her information” (Westin,
1967)
However, it is difficult to say that conventional
surveillance camera systems provide privacy to
individuals who are photographed by a system. In fact,
there are cases in which a person is identified through
the output of surveillance camera videos, and
therefore, privacy is not ensured. Furthermore, such
leaks can be prevented if individuals are granted the
right to control their recorded videos and images.
Writefix posted a privacy concern on a surveillance
camera. The essay suggests the importance of
balancing the need for respecting validity and
personal privacy in surveillance camera security.
Therefore, it is important to provide privacy on
surveillance camera systems. To protect the right of
privacy on a surveillance camera system, individuals
need to have the right to control their personal
information (face information).
Considering the threat to privacy from current
surveillance systems, it is important to develop a
surveillance system that ensures the privacy of
individuals. To ensure privacy protection, individuals
being recorded must have the right to control their
recorded facial images. On the other hand,
surveillance system recordings are important in
criminal investigations. Therefore, it is not desirable
for a perpetrator in a crime to have the right to control
his or her recorded facial information.
Consequently, a surveillance system that achieves
both “privacy protection” and “effective use of
surveillance image” is required. Leaks of surveillance
camera images bring to light the dire need for privacy
protection in the current information society.
3 RELATED STUDIES
In this section, we introduce the key technologies
used in the proposed system. In Section 3.1, we
explain “Short Group Signatures” (Boneh, 2004), and
in Section 3.2, “Recoverable original video for
mosaic system” (Kusama, 2015) is explained.
3.1 Short Group Signatures
Short Group Signatures (Boneh, 2004), a group
signature technique, has the following three features.
1. Only group members can produce signatures.
2. A verifier can verify a signature, but cannot
identify the signer.
3. Only the Certification Authority can identify a
signer.
Short Group Signatures is the technique used in the
proposed system to confirm the identity of a recorded
person.
3.1.1 Bilinear Group
A bilinear group has the following features.
1.
and
are two cyclic groups of prime order p.
2.
and
are generators of
and
,
respectively
3. is a computable isomorphism from
to
,
with
=
.
4. is a computable map, and :
×
→
with
the following properties.
Bilinearity: for all ∈
,
and
a,b , e
,
=
,

Nondegeneracy: e
,
=,

3.1.2 Algorithm
KeyGen(n).
The algorithm for Short Group Signatures takes a
parameter n, the size of the group, as input and
proceeds as follows. Select a generator
in
uniformly at random, and let
=
. Select
h
←
∖1
and ξ
←
, and let u,v
such
that u
=v
=h. Select γ
←
, and let w=
.
Using γ, generate for each user i, 1≪i≪n, a
strong Diffie–Hellman (SDH) tuple
,
: select
←
, and let

=
.
The group public key is gpk =
,
,,,ℎ,
.
The private key of the group manager key is gmk =
ξ
, and a group member i’s key is gsk
i
=
,
. No party is allowed to possess γ; it is known
only to the private-key issuer.
Sign(gpk, gsk[i], M).
Given a group public key as gpk =
,
,,,ℎ,
, a member’s key as gsk
i
=
ICE-B2015-InternationalConferenceone-Business
42
,
, and a message M, compute the signature as
follows.
1. Compute the values
,
,
,R
,R
,R
,R
,R
.
δ
=
,δ
=
=u
,
=u
,
=
h

R
=
,R
=
,R
=

,
R
=

R
=
,
∙
ℎ,


∙
ℎ,


α,β
←
,r
,r
,r
,r
,r
∈
2. Compute a challenge c using the hash function as
c=H
,
,
,
,R
,R
,R
,R
,R
∈
3. Using c, construct the values:
=r
,
=r

=r

=r
δ
,
=r
δ
=r
,
=r

=r

=r
δ
,
=r
δ
4. Output the signature σ, computed as
σ=
,
,
,,
,
,
,
,
Verify(gpk, M, σ).
Given a message M, a group signature σ, and
group public key gpk =
,
,,,ℎ,
, , verify
that σ is a valid signature as follows:
1. Re-drive R
,R
,R
,R
, and R
as follow:
2. R
=

,R
=

R
=
,
∙
ℎ,


∙
ℎ,


∙
,
,
R
=

,R
=

3. Check as follow:
c=H
,
,
,
,R
,R
,R
,R
,R
Accept if this check succeeds, else reject.
Open(gpk, gmsk, M, σ).
This algorithm is used for tracing a signer. Group
public key gpk =
,
,,,ℎ,
and group
manager’s key gmk =
ξ
are used together with
a message M and a signature σ=

,
,
,,
,
,
,
,
to trace.
First, verify the signature. Second, recover the
member’s key
as
∙

=


∙

=
The group manager can then identify the signer
from a group member index.
3.2 Recoverable Original Video for
Mosaic System
We employ “Recoverable original video for mosaic
system” (Kusama, 2015) as the method for
controlling an individual’s facial information. This
method provides the reversible mosaic on a video of
MPEG2. In that paper, implementation evaluation is
carried out on MPEG2 videos. We briefly explain the
reversible mosaic generation method of “Recoverable
original video for mosaic system” in this section.
3.2.1 Method for Generating the Mosaic
To obtain the JPEG image, the original image is
divided into blocks that are converted into discrete
cosine transformations after quantization. The n×n
blocks around the DC component are set to zero (as
shown in Figure 1), resulting in distortion, and the
distortion acts as a mosaic. In addition, keep the value
of the quantization output before the change, and
remove the mosaic by replacing locations that are
zero.
Figure 1: Mosaic method (n = 3).
3.2.2 Method for Generating Reversible
Mosaic
As shown in Figure 2, the face section of the data to
be hidden is extracted first. The extracted face data
are divided into 8 × 8 pixel blocks. Discrete cosine
transformations and quantization are performed on
the blocks, as in the case of JPEG. In the blocks, all
the values of low-frequency n × n pixels, except for
Figure 2: Generating a reversible mosaic.
SurveillanceCameraSystemBalancingPrivacyProtectionandEffectiveSurveillanceImageUse
43
the DC element of the quantization output, are
replaced with zeroes after the original values are
stored. Further, the encrypted original values are
encrypted in the DCT domain of the mosaic image
using the reversible watermark (Xuan, 2007). Finally,
a compressed JPEG stream can be obtained after
encoding is performed.
3.2.3 Method for Removing a Reversible
Mosaic
Figure 3 shows decoding process. Through this
process, we obtain the value of the quantization
output, which includes the watermarked information.
The watermarked information is extracted by
decoding the reversible watermark; then, we decrypt
the face data of each block. Finally, we obtain the
original image after reversing the data of the low-
frequency n × n matrix expect for the DC element.
Figure 3: Removing the reversible mosaic.
4 PROPOSED SYSTEM
4.1 Outline of the Proposed System
The objectives of the proposed system for balancing
the requirements of both privacy protection and crime
investigation are as follows.
1. A person who wants to conceal his or her recorded
face information can do so.
2. The identity of a person who wants to conceal his
or her own face information mustn’t be divulged,
unless that person performs a criminal act.
3. In a crime investigation, the police department
must be given access to unobscured face
information, if required
Objective 1 considers the people’s right to control
their own recorded face information. People who care
about their privacy can conceal their face
information, whereas the face information of those
who are not concerned about privacy is not concealed.
A surveillant can identify a shoplifting culprit in real
time, if the culprit is in the latter category. However,
if the culprit is in the former, we can identify him or
her through Objective 2, i.e., the system must have a
mechanism to identify a person whose information is
concealed, if the need arises, such as in the case of a
criminal act. Further, the police department must
receive unobscured face information related to a
criminal investigation, so that the investigation can be
conducted in a conventional manner; this is
considered in Objective 3. Each element of the
proposed system is shown in Figure 4.
Figure 4: Elements of the proposed system.
In the following, we explain each element in
Figure 4. User H is a person who has an interest in
securing privacy and wants to conceal his/her face
information. In contrast, user O is a person who is not
concerned about privacy. The surveillant is a person
who monitors the recordings of the surveillance
cameras on the monitors in real time. The police
department is the element that requires unobscured
images for crime investigations. The Certification
Authority (CA) manages the personal information of
user H securely and generates secret keys for each
user H. These secret keys preserve the anonymity of
user H using Short Group Signatures. The
surveillance camera sends the image to the server.
The access points transfer data between user H and
the server. The signature of user H is generated in the
device such as smart phone owned by user H.
Moreover, the access points collect the location data
of the device and send them to the server. The server
verifies the signature from the access points to
confirm that the person is a legitimate user H.
Furthermore, the server generates a removable
mosaic on the face of user H, to obscure that person’s
identity. The server manages the data used for
generating the mosaic. Finally, the server sends a
modified image (one with the mosaic) to the
surveillant.
ICE-B2015-InternationalConferenceone-Business
44
4.2 Proposed System Configuration
In Figure 5, we present the system model of the
proposed system and explain each element in the
model.
Figure 5: System model.
4.2.1 Photographed People
We classify the photographed people into two types
(user H and user O). Each user has the following
features.
User H
1. Wants to hide face information.
2. Has a mobile device to establish legitimacy and
communicate with access points.
3. Has registered personal information with the CA
and has obtained a key for producing short group
signatures.
User O
1. Low interest in concealing face information.
2. Does not have a key for producing short group
signatures.
4.2.2 Surveillant
The surveillant has the following features.
1. Monitors the images sent from the surveillance
camera via the server.
2. Can operate only on the monitoring server and
cannot modify the functions of the server.
4.2.3 Surveillance Camera
Here, we assume network cameras to be the
surveillance cameras, as some network cameras can
output encrypted images. A surveillance camera has
the following features.
1. Encrypts and sends images to the server.
2. Is tamper-resistant against function changes.
4.2.4 Access Points
Two or more access points are installed and have the
following features.
1. They are located at the site where the surveillance
cameras are installed.
2. They communicate with user H’s mobile device
and the server.
3. They collect information for specifying user H’s
device position and send it to the server.
4. They are tamper-resistant against function
changes.
4.2.5 Server
The server has the following features.
1. It is managed by a trusted administrator.
2. It communicates with user H’s device via access
points.
3. It verifies whether the person is user H without
identifying the person.
4. It obtains the information for specifying user H’s
device position using the TDOA (Time Difference
of Arrival) method.
5. It estimates the location of user H’s face in the
image.
6. It generates a reversible mosaic for the user’s face.
Feature 3 listed above is achieved through short
group signatures.
Reversible mosaic refers to the concept
introduced in “Recoverable original video for mosaic
system”.
The TDOA method in feature 4 estimates the
position of the mobile devices (Cong, 2004).
Face estimation is achieved through facial
recognition technology for the person with the
devices.
4.2.6 Certification Authority (CA)
The CA has the following features.
The CA has the following features.
1. It manages the personal information of user H
securely.
2. It generates the sign key (for generating short
group signatures) and provides the key to each user
H.
3. It generates the verify key (for verifying the short
group signatures), and publicizes it.
4. It provides the information of user H and removing
mosaic to the police department that has a warrant
SurveillanceCameraSystemBalancingPrivacyProtectionandEffectiveSurveillanceImageUse
45
from the court with regard to a criminal
investigation.
4.2.7 Timestamp Server
We use the inbuilt timestamp server. The timestamp
server sends timestamp data to the server and the CA.
4.2.8 Police Department
The Police department demands the recording of a
surveillance camera by producing a warrant with
regard to a criminal investigation; the police
department receives the mosaicked surveillance
camera recording and information for removing the
mosaic from the CA.
4.3 Communication Protocol
In this section, we present the protocol of the
proposed system
4.3.1 Preconfigured
We present the protocol for short group signatures
between the people who wish to be user H and the CA.
This protocol is executed before entering the visibility
range of the surveillance cameras.
Step 1: The people who want to be user H register
their personal information with the CA.
Step 2: The CA confirms the personal information
and provides sign key (gsk) and ID to user H.
Step 3: The CA publishes the verify key (gpk) and
manages the secret key (gmk) for short group
signatures.
4.3.2 Reversible Mosaic Generation
The reversible mosaic image generation protocol is
executed when user H enters the visibility range of
the surveillance camera. This system does not
conduct mosaic generation for User O.
Step 1: The server sends the server ID (IDserver)
periodically via the access points.
Step 2: User H Receives IDserver on his mobile
device, and generates random number β.
Step 3: User H’s mobile device calculates the
following value k with IDserver, β, and user H’s
ID (ID).
k=H
 
Step 4: User H’s mobile device produces a short
group signature (σ) for the value k by gsk and
sends σ, k, and β to the server via the access
points. The short group signature is calculated as
follows.
σ=
,
,
,,
,
,
,
,
Step 5: After receiving each datum, the server verifies
the signature σ. If the signature is the same as the
previous signature, the server does not verify the
signature.
Step 6: If the signature is valid, the server generates a
reversible mosaic on the face of user H with a
mosaic key (mk) using the technique shown in 3.2.
The mosaic key consists of timestamp Ts and σ and
encrypted by the server secret key sk as follows.
mk = Enc
H
 σ
Step 7: The server sends the mosaicked image to the
surveillant. The server stores σ , k , and the
mosaicked image, and deletes mkafter generating
the reversible mosaic.
4.3.3 Removing Mosaic Generation
This protocol stated below is used for removing a
reversible mosaic in the case of a crime investigation.
Step 1: The police department obtains a search
warrant from a court to view surveillance camera
recordings. The search warrant is presented to the
CA.
Step 2: The CA requests the server to send the mosaic
image of the investigation subject, σ and k.
Step 3: The CA sends the timestamp Ts to the server.
(Timestamp Ts corresponds to the recording time.)
Step 4: The server restores the mosaic key mk from
the Ts and sends mk and the mosaicked image to
the police department.
Step 5: The police department requests the CA to
provide identification information regarding user
H, if necessary.
Step 6: Using the secret key gmk, the CA identifies
the user H from σ and sends the personal
information of the identified user H to the police
department if requested.
5 CONSIDERATION
5.1 Achieving the Objective
The following are the objectives of the proposed
system.
ICE-B2015-InternationalConferenceone-Business
46
1. A person who wants to conceal his or her recorded
face information can do so.
2. The identity of a person who wants to conceal the
person’s own face information mustn’t be
divulged, unless that person performs a criminal
act.
3. In a crime investigation, the police department
must be given access to unobscured face
information, if required.
Objective 1: A person who wants to conceal his or
her recorded face information registers his/her
personal information with the CA. The person can
become user H after registering. The face
information of user H is concealed by the
reversible mosaic according to the “reversible
mosaic generation protocol.” User H
communicates with the server via the access points
and sends the information for reversible mosaic
generation in Steps 1-3 of the reversible mosaic
generation protocol. The face information of the
user H is concealed with Step 6 of the protocol.
Accordingly, we achieve Objective 1.
Objective 2: The data that user H sends are only
signature σ , calculated value k, and random
number . The server cannot identify the user H
from these data. Objective 2 is achieved, because
it is not possible to identify user H unless the CA
uses the gmk and identifies the user H from that
user’s signature σ.
Objective 3: With Steps 1-4 of “remove the
reversible mosaic protocol,” the police department
can restore the face information from the
mosaicked surveillance camera image. Objective 3
is achieved with the “remove the reversible mosaic
protocol.”
5.2 Attacks against Proposed System
We consider some attacks against the proposed
system in this section. We assume that the attacker
has the following agenda.
1. Impersonate user H.
2. Disrupt communication between the surveillance
camera, the access point, and the server.
3. Tamper with or intercept the image stored on the
server.
Concrete attacks based on Purpose 1
Attack 1: The attacker intercepts the data that are sent
by user H and resends them to the server.
Attack 2: The attacker uses the device of user H
illegally.
Concrete attacks based on Purpose 2
Attack 3: The attacker blocks the communication
between the surveillance camera, the access points,
and the server.
Attack 4: The attacker falsifies the data that are sent
by user H, the access point, and the surveillance
camera.
Concrete attack based on Purpose 3
Attack 5: The attacker tampers with or intercepts the
images stored on the server.
We consider the security for the aforementioned
five attacks in Section 5.3.
5.3 Security for Proposed System
Attack 1: This attack can be prevented by step 5 of
the “generation reversible mosaic” protocol (i.e.,
the server verifies the same signature only once).
The signature for verifying user H consists of a
calculated number. This number consists of a
random number and two IDs (the server and user
H). The attacker can create a valid signature on
another random number, if he or she knows the
sign key. It is possible to prevent this attack by
securely managing the sign key of user H.
Attack 2: There are several ways to bypass security
on a user’s device. These are not considered. The
argument that the attacker cannot use the device
unless it is unlocked does not have enough strength.
Please consider elaborating the suggestion. The CA
can identify the signer through step 6 of the
“remove the reverible mosaic” protocol in a crime
investigation. The user has registered personal
information including face information with the
CA. The police department can confirm the face in
the recording for a crime investigation after
removing the reversible mosaic. However, the user
might be suspected, if identification is performed
based on the signature; therefore, it is necessary to
confirm via a facial inspection in a crime
investigation.
Attack 3: Falsification of the surveillance camera and
the access points is prevented by the tamper-
resistant feature. The falsification of the server can
be prevented, if a legitimate administrator manages
the server. However, if the attacker routes the
communication link, this solution will still allow
you to send and receive the “hello message. The
surveillant can detect blocking of the
SurveillanceCameraSystemBalancingPrivacyProtectionandEffectiveSurveillanceImageUse
47
communication between the surveillance camera
and the server, because the image would not appear
on the monitor, if the communication were
blocked. The administrator can deal with this
attack by setting in place a mechanism for
reporting unsuccessful communication.
Attack 4: Falsification of the data sent to the server
by the surveillance camera can be prevented using
the encryption function of the surveillance camera.
To detect falsification of the data sent by the access
points, the server produces a signature of the data
of the server. Therefore, user H can detect the
falsification easily. On the other hand, detection of
falsified data sent by user H is easy. The server can
verify it by verifying the data. The surveillant must
deal with user H, if the server fails several times to
verify the signature of the same device.
Attack 5: Detection of falsification of the image
stored in the server can be achieved by producing
a signature on the image data where the server
stores the image. The key for this signature must be
securely managed by the administrator. This
signature must be decided at a predetermined time
interval (for example, each day). An alternate
method is that the administrator encrypts the image
using his secret key. In this manner, falsification
can be prevented.
6 CONCLUSIONS
In this paper, we proposed a new surveillance camera
system that balances the requirements of both privacy
protection and criminal investigations. Further, we
presented the communication protocol of the
proposed system. We also assumed possible attacks
to the system and presented security measures these
attacks.
REFERENCES
Warren, S., Brandeis, L., 1890. The Righ to Privacy.
Harvard Law Review. Vol.4, pp.193.
Westin, A., 1967. Privacy and Freedom. Washington and
Lee Law Review, vol.25, pp.197-170.
Murakami, Y., 2004. Privacy Issues in the Ubiquitous
Information Society and Law in Japan. Systems, Man
and Cybernetics, 2004 IEEE International Conference
on, vol.6, pp.5645-5650.
Xiaoyi, Y., Chinomi, K., Koshimizu, T., Nitta, N., Ito, Y.,
Babaguchi, N., 2008. Privacy protecting visual
processing for secure video surveillance. 15th IEEE
International Conference on, pp.1672-1675.
Chaum, D., Heyast, E., 1991. Group Signatures
Proceedings of Eurocrypt 1991, vol.547, pp.257-265.
Boneh, D, Boyen, X, Shacham, H., 2004. Short Group
Signatures. Advances in Cryptology -CRYPTO 2004-,
pp.41-55.
Kusama, Y., Kang, H., Iwamura, K., 2015. Recoverable
original video for mosaic system.(japanese).
Xuan, G., Shi, G., Ni, Z., Chai, , Z., Cui, X., Tong, X., 2007.
Reversible Data Hiding for JPEG Images Based on
Histogram Pairs. Image Analysis and Recognition,
pp.715-727.
Cong, L., Z., Zhuang, W., 2001. Non-line-of-sight error
mitigation in TDOA mobile location. GLOBECOM’01.
IEEE, vol.1, pp.680-684.
Gezici, Z., 2008. A Survey on Wireless Position Estimation.
Wireless Personal Communications, vol.44, pp.263-
282.
Comaniciu, D., Ramesh, V., 2000. Real-time tracking of
non-rigid objects using mean shift. Computer Vision
and Pattern Recognition, 2000. Proceedings. IEEE
Conference on, vol.2, pp.142-149.
Ross, D., Lim, J., Lin, R., Yang, M., 2008. Incremental
Learning for Robust Visual Tracking. International
Journal of Computer Vision, vol.77, pp.125-141.
Viola, P., Jones, M., 2004. Robust Real-Time Face
Detection. International Journal of Computer Vision,
vol.57, pp137-154.
Wright, J., Yang, Y., Ganesh, A., Sastry, S., Yi M., 2008.
Robust Face Recognition via Sparse Representation.
Pattern Analysis and Machine Intelligence, IEEE
Transaction on, vol.31, pp.210-227.
Andrew, S., 2009. Privacy Protection in Video Surveillance
System. Protecting Privacy in Video Surveillance,
pp.35-47.
Writefix. Security Cameras and Privacy.
http://writefix.com/?page_id=1584.
ICE-B2015-InternationalConferenceone-Business
48