AI Trustworthiness Challenges and Opportunities Related to IIoT
not be readily accessible, so a model and
sensitivity analysis may be needed. This
suggests that a “black box” mechanism for
recording sensor data leading to a decision
may be needed (or real time network
transmission of the data) in order to use a
model to validate and establish confidence
in the system. Data Security plays a central and enabling
role in the Data Protection strategy 16 of
organizations. AI can be applied to IoT data
(in-motion, at-rest, in-use) to assess
infringements to design objectives of
security and power the notification
processes to HIL so that remediation
processes can be applied.
AI systems can do much to enhance the
safety of systems by improving decisions and
solutions to problems through the analysis
of more data and more complex data than
people can handle in a limited time or with
limited resources. An example is an airplane
in disrupted mode where an “experienced
pilot” does not have a solution. In this case
an AI system backed with a data store of
similar cases can prove invaluable. AI-augmented cyber-defense capabilities for
IoT systems can be superior to traditional
rules-based cybersecurity. However, AI can
also present new opportunities for cyber-
attackers to carry out attacks at greater
scale. It is therefore important for
developers and operators of IoT systems to
consider the wider scope of IoT
Trustworthiness, when they reflect on cyber
threats and the application of AI-powered
cybersecurity tools to their IoT systems. This
is especially important considering the
interdependencies that exist between
security, safety, reliability, resilience and
privacy; and cyber threats can permeate the
whole IoT system.
AI AND S ECURITY
Similar to the concerns regarding safety and
the potential side-effects of faults and
errors, cyber security issues may adversely
affect IoT systems. In the case of cyber
security, there would be malicious intent to
compromise the systems, and AI may be
leveraged to find vulnerabilities in these
systems and enable such attacks to be
launched. On the other hand, the same AI
techniques may be used to defend the
systems by identifying such attacks and
mitigating
them
with
appropriate
countermeasures and controls. This battle
for security using AI for both weaponization
and defense is likely going to escalate over
time.
16
AI can also be used to improve situational
awareness, detect system vulnerabilities,
detect attacks in progress and help with
forensic analysis.
AI AND P RIVACY
Privacy concerns are not new to IoT.
Protecting Personal Data is central to the
privacy strategy. However, in many cases, it
is not a particular leak of Personal Data that
causes the privacy violation but rather the
aggregation of a number of different,
seemingly unrelated pieces of information
Refer to the IIC Data Protection Best Practices white paper
IIC Journal of Innovation
- 86 -