Page 181 - Cyber Terrorism and Extremism as Threat to Critical Infrastructure Protection

GRAEME BALLARD: IF THE FACE FITS: IS IT POSSIBLE FOR ARTIFICIAL INTELLIGENCE TO ACCURATELY PREDICT THREATS TO
PROTECT CRITICAL INFRASTRUCTURE?

implementation. Change has always been a continuous process but, after engaging ML, the
process of change will occur at a faster pace. Political and economic systems must adapt to this
8
– preferably proactively rather than reactively . As I hope will be clear from the discussion so
far, the relationship between ML and human processes will be symbiotic – each will influence
the other. Under these conditions, the ML tool itself must be considered as highly active and
will need to be constantly modified, curated and updated in order to remain relevant and
accurate – not just technologically, but also in terms of data management and its continued
ability to identify threats in a world that is constantly changing.

By definition, therefore, data-sharing is a necessity that will make a positive contribution to the
identification of threats and those tasked with maintaining security. It must become the norm –
both in attitude and in practice. The more frictionless movement of data that can be achieved,
the more accurate and trusted any ML tool will be. Data must be acquired and moved from as
varied and wide a source(s) as possible. This will mean not only sharing data between nuclear
power plants within the same company, but data between competing companies; perhaps also
with different industries and, almost certainly, across geographical and political boundaries.

In order to facilitate such frictionless data-sharing, it is imperative that changes to both legal
and commercial practices must be found. This will likely require a shift in the psychological
attitude of both our politicians and our business leaders, and also significant (and potentially
disruptive) changes to business models, accounting practices and the security services
themselves. This must all be achieved without itself creating a security risk, and achieved
within the law, ensuring that what Lepine (2014) describes as the social contract remains
intact.

6 Conclusion

The use of ML to identify threats to protect critical infrastructure within the energy sector is a
tantalising prospect and a project that my consortium partners have been actively undertaking
for the last 18 months. There is little question that the correct and careful implementation of
such a technology will add value to the security of nuclear (or other) energy production. The
ultimate success of any implementation, however, is dependent on six factors:
• Finding good consortium partners who truly understand the idiosyncrasies and
complications of researching human subjects.
• Finding the appropriate amount of funding to be able to account for the idiosyncrasies and
complications of researching human subjects.
• If points one and two can be addressed, the next factor of success is the gathering of
good data and ensuring certain methodological issues are addressed when designing the
tool. These human methodological issues are no longer mere academic niceties, but are
essential to the accurate and meaningful implementation of the technology.
• Trust in the technology. The technology must be human-centric. It must add value in the
widest possible terms. For example, ML has the potential to not merely identify threats
in terms of preventing catastrophe, it has a role to play in human resource management,
ensuring the wellbeing of workers in terms of their physical and psychological health.
These factors should be considered not simply because of their relevance in preventing a
disaster, but in wider acceptance of the results from the tool.

8 Once more, the lessons from social media v legacy media should be learned.

181