21-08-2012, 02:25 PM
Security and Privacy in Distributed Smart Cameras
[attachment=32293]
ABSTRACT
Distributed smart camera systems are becoming
increasingly important in a wide range of applications. As they
are often deployed in public space and/or our personal
environment, they increasingly access and manipulate sensitive
or private information. Their architectures need to address
security and privacy issues appropriately, considering them
from the inception of the overall system structure. In this
paper, we present security and privacy issues of distributed
smart camera systems. We describe security requirements,
possible attacks, and common risks, analyzing issues at the
node and at the network level and presenting available
solutions. Although security issues of distributed smart
cameras are analogous to networked embedded systems and
sensor networks, emphasis is given to special requirements of
smart camera networks, including privacy and continuous realtime
operation.
INTRODUCTION
Distributed smart cameras are used in surveillance
applications, which, in several environments, have strict
security as well as privacy protection requirements. As a
result, security of the smart camera system itself becomes a
critical factor in the design of the overall architecture.
Distributed smart camera systems constitute distributed
systems, which are networks of computationally strong
sensors, the cameras. Thus, security design of smart
camera systems takes into account both node-level security
and network-level security. Security requirements at both
node and network level should provide adequate data
protection.
DISTRIBUTED SMART CAMERAS
ARCHITECTURE
A detailed security analysis of smart camera systems
requires specification of their main architectural characteristics
[1], especially those relating to security. A
distributed smart camera system is typically composed of
the nodes, i.e., the smart cameras themselves,
interconnected through a network infrastructure. Depending
on the exact architecture, the system can be completely
distributed or allow for a certain amount of central control.
Considering the typical architecture for a distributed smart
camera system, we present our analysis based on a
complete system model that is described in detail below
and pictured in Fig. 1.
Camera Software
Three main software tasks are implemented in a smart
camera node: i) the processing task, ii) the communication
task, and iii) the control task. The processing task includes
all digital signal-processing functions, which form the
heart of the system, and is assigned to the processing unit.
This task can be highly distributed with many cooperating
smart camera nodes, since pairs or groups of nodes can be
working on overlapping areas or coprocess the same task.
The communication task includes all processes related to
the nodes communicating among themselves or with a
central control system; this task is assigned to the
communication unit. The control task consists of the core
software of the node, which starts and synchronizes the
other tasks; it can be assigned to either the processing or
the communication unit. In our model, we will assume
that the control tasks are assigned to the communication
unit, which can be implemented using general-purpose
processors, leaving the application computational part,
i.e., the pure DSP hardware, to the processing unit. This
approach for the software task distribution is reasonable
and supported by the fact that the control and communication
tasks are of a less distributed natureVactually, not
distributed at all in most cases. This leads to a processing
unit that performs distributed tasks and a communication
unit that performs local tasks.
Network
The network architecture of a distributed smart camera
system may employ a wide range of interconnection
technologies, topologies, protocols, etc. In general, a smart
camera network has many similarities to an ad hoc
network. However, it is often not necessary for all nodes
to be able to communicate with all other nodes, especially
in environments where the number of nodes becomes very
large and the area of coverage very wide. Pairs or groups of
nodes can be formed in order to comonitor or coprocess.
As a result of its operation, a smart camera network is
typically assumed to operate in an unattended, public, and
possibly hostile environment.
Tamper Resistance, Physical Attacks, and
Side-Channel Attacks
Tamper resistance is the ability of a device to remain
secure even if found in the hands of attackers [6], as
mentioned previously. In general, this is very demanding
and costly, and thus it is not recommended as an objective
for all smart camera nodes. However, it can be used for
certain special nodes, such as a base station [7], that should
be as secure as possible. Although nodes of importance,
such as base stations, should, in general, be in controlled
environments which are secure against tampering, we will
assume that there is a need for a special node to exist in an
insecure environment. As explained in Section V, this
special node could aggregate data or assist in encryption
key distribution. There exist several methods to tamper
with a reachable or captured system.
Strategies for Key Management
Encrypted communication between nodes requires
assignment of appropriate encryption keys to nodes for
encrypted communication and authorization procedures
[18]. The key distribution strategy is a crucial first step
because it forms the basis for secure communication. Once
keys have been securely set up, communications secrecy is
largely accomplished. The employed encryption method
determines not only the level of security but also the
communication and processing overhead; thus, it is
necessary to follow an efficient strategy. Public key
cryptography [19] is an effective, although costly, method
to encrypt communication between nodes, as it is
computationally demanding. Considering conventional
embedded system capabilities, we believe that public key
cryptography is beyond the processing abilities of smart
camera nodes, especially for large networks that require a
large number of keys, when using one key for each node.
CONCLUSION AND FUTURE
RESEARCH DIRECTIONS
Security for smart camera networks is a crucial property,
considering their typical applications, which include
sensitive and private information. Importantly, security
must be considered from the first stages of system
architecture and design and cannot be easily added to
previously designed systems and protocols. In this paper,
we surveyed a wide range of mechanisms and securityenabled
protocols that need to be included in distributed
smart camera systems.
[attachment=32293]
ABSTRACT
Distributed smart camera systems are becoming
increasingly important in a wide range of applications. As they
are often deployed in public space and/or our personal
environment, they increasingly access and manipulate sensitive
or private information. Their architectures need to address
security and privacy issues appropriately, considering them
from the inception of the overall system structure. In this
paper, we present security and privacy issues of distributed
smart camera systems. We describe security requirements,
possible attacks, and common risks, analyzing issues at the
node and at the network level and presenting available
solutions. Although security issues of distributed smart
cameras are analogous to networked embedded systems and
sensor networks, emphasis is given to special requirements of
smart camera networks, including privacy and continuous realtime
operation.
INTRODUCTION
Distributed smart cameras are used in surveillance
applications, which, in several environments, have strict
security as well as privacy protection requirements. As a
result, security of the smart camera system itself becomes a
critical factor in the design of the overall architecture.
Distributed smart camera systems constitute distributed
systems, which are networks of computationally strong
sensors, the cameras. Thus, security design of smart
camera systems takes into account both node-level security
and network-level security. Security requirements at both
node and network level should provide adequate data
protection.
DISTRIBUTED SMART CAMERAS
ARCHITECTURE
A detailed security analysis of smart camera systems
requires specification of their main architectural characteristics
[1], especially those relating to security. A
distributed smart camera system is typically composed of
the nodes, i.e., the smart cameras themselves,
interconnected through a network infrastructure. Depending
on the exact architecture, the system can be completely
distributed or allow for a certain amount of central control.
Considering the typical architecture for a distributed smart
camera system, we present our analysis based on a
complete system model that is described in detail below
and pictured in Fig. 1.
Camera Software
Three main software tasks are implemented in a smart
camera node: i) the processing task, ii) the communication
task, and iii) the control task. The processing task includes
all digital signal-processing functions, which form the
heart of the system, and is assigned to the processing unit.
This task can be highly distributed with many cooperating
smart camera nodes, since pairs or groups of nodes can be
working on overlapping areas or coprocess the same task.
The communication task includes all processes related to
the nodes communicating among themselves or with a
central control system; this task is assigned to the
communication unit. The control task consists of the core
software of the node, which starts and synchronizes the
other tasks; it can be assigned to either the processing or
the communication unit. In our model, we will assume
that the control tasks are assigned to the communication
unit, which can be implemented using general-purpose
processors, leaving the application computational part,
i.e., the pure DSP hardware, to the processing unit. This
approach for the software task distribution is reasonable
and supported by the fact that the control and communication
tasks are of a less distributed natureVactually, not
distributed at all in most cases. This leads to a processing
unit that performs distributed tasks and a communication
unit that performs local tasks.
Network
The network architecture of a distributed smart camera
system may employ a wide range of interconnection
technologies, topologies, protocols, etc. In general, a smart
camera network has many similarities to an ad hoc
network. However, it is often not necessary for all nodes
to be able to communicate with all other nodes, especially
in environments where the number of nodes becomes very
large and the area of coverage very wide. Pairs or groups of
nodes can be formed in order to comonitor or coprocess.
As a result of its operation, a smart camera network is
typically assumed to operate in an unattended, public, and
possibly hostile environment.
Tamper Resistance, Physical Attacks, and
Side-Channel Attacks
Tamper resistance is the ability of a device to remain
secure even if found in the hands of attackers [6], as
mentioned previously. In general, this is very demanding
and costly, and thus it is not recommended as an objective
for all smart camera nodes. However, it can be used for
certain special nodes, such as a base station [7], that should
be as secure as possible. Although nodes of importance,
such as base stations, should, in general, be in controlled
environments which are secure against tampering, we will
assume that there is a need for a special node to exist in an
insecure environment. As explained in Section V, this
special node could aggregate data or assist in encryption
key distribution. There exist several methods to tamper
with a reachable or captured system.
Strategies for Key Management
Encrypted communication between nodes requires
assignment of appropriate encryption keys to nodes for
encrypted communication and authorization procedures
[18]. The key distribution strategy is a crucial first step
because it forms the basis for secure communication. Once
keys have been securely set up, communications secrecy is
largely accomplished. The employed encryption method
determines not only the level of security but also the
communication and processing overhead; thus, it is
necessary to follow an efficient strategy. Public key
cryptography [19] is an effective, although costly, method
to encrypt communication between nodes, as it is
computationally demanding. Considering conventional
embedded system capabilities, we believe that public key
cryptography is beyond the processing abilities of smart
camera nodes, especially for large networks that require a
large number of keys, when using one key for each node.
CONCLUSION AND FUTURE
RESEARCH DIRECTIONS
Security for smart camera networks is a crucial property,
considering their typical applications, which include
sensitive and private information. Importantly, security
must be considered from the first stages of system
architecture and design and cannot be easily added to
previously designed systems and protocols. In this paper,
we surveyed a wide range of mechanisms and securityenabled
protocols that need to be included in distributed
smart camera systems.