what-when-how
In Depth Tutorials and Information
may emerge in case they extend to the lower layer of STNs such as the communica-
tion and physical layers. here is no clear deinition of trust in the STNs. hus, it
is even more difficult to attach the concept of trust to a physical device. Moreover,
the metrics of trust include some subjective elements in its social layer while in
its device layer the trust should be more objective by eliminating some subjective
part. For example, the device trust value can be associated with the reliability of
the device. here must be some kind of interfaces between the layers to translate
the trust upwards or downwards. hese interfaces are very important, just like the
protocol stack, to ensure the trust model in each layer works independently and
cooperatively. Besides, although these trust models have provided some strategies
for resilience to social attacks, these strategies cannot extend well to the device
layer. We must consider the requirements of the device layer to detect device fail-
ures and physical attacks.
According to the characteristics of centralized and distributed trust models we
can suggest a mixed system that makes use of the advantages of centralized sys-
tems to counterbalance the shortcomings of distributed systems and vice versa. For
example, an STN can be separated into several cliques. In a clique there is a leader
managing to compute the trust of each member and communicate to other leaders
of different cliques. Besides, the calculation of trust among cliques is distributed.
his scheme may be effective if the network is divided appropriately according to
sociotechnical relationships. his topology is also easy to scale and maintain.
None of the trust models we mentioned above considers all the factors affect-
ing the computation of trust. here should be a general trust computation model,
especially in a distributed environment. Although it is very hard to achieve this
goal, we can emphasize several factors at the expense of others, according to the
requirements in a specific STN environment. As for the topology mentioned above,
we can adopt different trust computation schemes in each clique based on their
characteristics. A simplified model may be used in a more trustworthy environment
while we should exploit a complex trust model with strategies limiting the mali-
cious behaviors of nodes.
Because trust computation is an iterative process, converge time plays an impor-
tant role in trust modeling. It is meaningless for the trust model if the calculation
cannot follow the interaction of the nodes. However, most of the trust models
above do not take time complexity into account. It is a challenge to balance the
effectiveness and complexity.
According to the strategies we mentioned in the trust models, they can limit
the malicious behaviors to some extent. However, some shortcomings which can be
exploited by malicious nodes still exist in these strategies. We should enhance the
strategies after future research. Based on the characteristics of STN there must be
some nodes with high centrality, which means that the nodes have many links. he
trust and the security of such nodes should be investigated. In the common sense,
the node that has the highest trust value should be popular in the transactions
among nodes. Other nodes should be prone to interact with it and make efforts to
