Trustworthiness Model Representation
Business goals ● A Retail store proprietor incorporates smart store analytics
to increase profits
● A system integrator provides the system and a Trust Service
to the proprietor
Core Requirements ● Capture and analyze video and sensor data for store,
product marketing
● Detect tampering and theft (of system and merchandise)
Key Trustworthiness
Characteristics ● Security
● Reliability
● Resilience
Attributes ● Camera position
● Camera and sensor reliability
● Application software
Table 3: Practical application of Trustworthiness in a Retail Store
the analytics data, gives the observer a
practical method to evaluate the system and
to apply the business model with a
measurable degree of confidence.
C ONCLUSIONS AND F URTHER W ORK
In this paper, we presented a practical
approach for a model representation of an
IIoT system’s trustworthiness. The Trust
Model is designed to flexibly address the
varying priorities of different types of IIoT
systems to provide the appropriate context
for the system. The component and system
Trust Scores that are computed by the
Model provide an easily interpreted value
that enables the model user to take the
appropriate actions to maintain the
trustworthiness of the system in operation.
We have not shared specific details on the
Trust Function algorithm or the specific
visualization of the model output as these
can be vendor-specific and proprietary.
In the use cases described above, the data
collected is not so unusual or unique in and
of itself. However, results of the analytics are
only as good as the video or sensor data that
is collected. The Trust Score generated by
the Trust System tackles the original
problem of an unreliable data source.
There is ideally a combination of multiple
cameras and sensors that provide enough
coverage. What if only subsets of the
cameras or sensors are operating correctly?
What If there is an intermittent connectivity
issue with one of the cameras? Either of
these conditions will result in a lower Trust
Score that is computed from the individual
attributes of the system. This results in
suboptimal data collection which results in
suboptimal data analytics and correlation.
The Trust System takes a set of expected and
observed attributes that describe the system
under observation and computes a Trust
Score. The Trust Score, when combined with
September 2018
42