92
Internet of Things (IoT)
networks data is much more challenging than anonymizing relational data due to many
problems, some of which are listed below.
• It is much more challenging to model the background knowledge of adversaries
and attacks about social network data than that about relational data. In a social
network many pieces of information can be used to identify individuals, such as
labels of vertices and edges, neighborhood graphs, induced subgraphs, and their
combinations. So, all these make modeling social networks difficult.
• It is also challenging to measure the information loss in anonymizing relational
data. Unlike for relational data for which the sum of the information loss for tuples
solves the purpose, it is hard to compare two social networks by comparing their
vertices and edges individually.
• Anonymizing a social network is much more difficult since changing labels of
vertices and edges may affect the neighborhoods of other vertices, and removing
or adding vertices and edges may affect other vertices as well as the properties of
the network.
To protect the privacy satisfactorily, the models make guarantee that any individual cannot be identified correctly in the anonymized social network with a probability higher
than 1/k, where k is user-specified parameter in the k-anonymity model (Tripathy and
Panda 2010). An adversary with the knowledge of neighborhood of a vertex cannot identify any individual from this anonymous graph with a confidence greater than 1/k.
Furthermore, Tripathy et al. (2011a, 2011b, 2011c, 2011d) improved the clustering stage of
the One Pass K-Means (OKA) algorithm, clustering Minimum Mean Roughness (MMeR)
algorithm, and introduced l-diversity algorithms and variances using soft computing
techniques like rough set theory. Such algorithms satisfy the privacy of individual nodes
to the extent of anonymity in public social networks. These k-anonymity and l-diversity
algorithms have the highest possibility of convergence of integration in the SIoT security
threats.
5.5.3 Research Opportunities in SIoT
Social IoT is going to experience massive deployment of sensors and consumer objects which
would affect the potential quantity of interconnected links between these devices (Ortiz et al.
2014; Alam et al. 2015). Many of these devices will be able to establish links and communicate
with other devices on their own in an unpredictable and dynamic manner. Therefore, existing tools, methods, and strategies associated with IoT security may need new consideration.
Many SIoT deployments will consist of collections of identical or near-identical devices. This
homogeneity magnifies the potential impact of any single security vulnerability by the sheer
number of devices that all have the same characteristics. Some IoT devices are likely to be
deployed in places where physical security is difficult or impossible to achieve. Attackers
may have direct physical access to IoT devices or wirelessly hack such devices. Anti-tamper
features and other design innovations will ensure security in SIoT. The social network-based
collaborative approach will be an effective solution to industry, government, and public
authorities to secure the Internet and cyberspace, including the SIoT.
We list out some questionnaires related to the security and probable threats and attacks
in IoT/SIoT in Table 5.3. The list may be helpful to the students, academicians, and researchers who have an interest in this topic for further research investigations.
Internet of Things (IoT)
networks data is much more challenging than anonymizing relational data due to many
problems, some of which are listed below.
• It is much more challenging to model the background knowledge of adversaries
and attacks about social network data than that about relational data. In a social
network many pieces of information can be used to identify individuals, such as
labels of vertices and edges, neighborhood graphs, induced subgraphs, and their
combinations. So, all these make modeling social networks difficult.
• It is also challenging to measure the information loss in anonymizing relational
data. Unlike for relational data for which the sum of the information loss for tuples
solves the purpose, it is hard to compare two social networks by comparing their
vertices and edges individually.
• Anonymizing a social network is much more difficult since changing labels of
vertices and edges may affect the neighborhoods of other vertices, and removing
or adding vertices and edges may affect other vertices as well as the properties of
the network.
To protect the privacy satisfactorily, the models make guarantee that any individual cannot be identified correctly in the anonymized social network with a probability higher
than 1/k, where k is user-specified parameter in the k-anonymity model (Tripathy and
Panda 2010). An adversary with the knowledge of neighborhood of a vertex cannot identify any individual from this anonymous graph with a confidence greater than 1/k.
Furthermore, Tripathy et al. (2011a, 2011b, 2011c, 2011d) improved the clustering stage of
the One Pass K-Means (OKA) algorithm, clustering Minimum Mean Roughness (MMeR)
algorithm, and introduced l-diversity algorithms and variances using soft computing
techniques like rough set theory. Such algorithms satisfy the privacy of individual nodes
to the extent of anonymity in public social networks. These k-anonymity and l-diversity
algorithms have the highest possibility of convergence of integration in the SIoT security
threats.
5.5.3 Research Opportunities in SIoT
Social IoT is going to experience massive deployment of sensors and consumer objects which
would affect the potential quantity of interconnected links between these devices (Ortiz et al.
2014; Alam et al. 2015). Many of these devices will be able to establish links and communicate
with other devices on their own in an unpredictable and dynamic manner. Therefore, existing tools, methods, and strategies associated with IoT security may need new consideration.
Many SIoT deployments will consist of collections of identical or near-identical devices. This
homogeneity magnifies the potential impact of any single security vulnerability by the sheer
number of devices that all have the same characteristics. Some IoT devices are likely to be
deployed in places where physical security is difficult or impossible to achieve. Attackers
may have direct physical access to IoT devices or wirelessly hack such devices. Anti-tamper
features and other design innovations will ensure security in SIoT. The social network-based
collaborative approach will be an effective solution to industry, government, and public
authorities to secure the Internet and cyberspace, including the SIoT.
We list out some questionnaires related to the security and probable threats and attacks
in IoT/SIoT in Table 5.3. The list may be helpful to the students, academicians, and researchers who have an interest in this topic for further research investigations.