Page 180 - Cyber Terrorism and Extremism as Threat to Critical Infrastructure Protection

SECTION II: CYBER TERRORISM AND SECURITY IMPLICATION FOR CRITICAL INFRASTRUCTURE PROTECTION

is a double edged sword because any error in the technology will have negative real-world
consequences that will emerge far faster, and be far more wide-reaching, than anything
7
legacy analysis could achieve . This is because, in a legacy world of relatively slow scientific
analysis, debate and falsification is a slow process, which gives people (both powerful and
not) time to adjust to developments and change. This will not be the case with a ML tool.
Analysis will be swift and debated only within the machine. It is vitally important, therefore,
to ensure ML uses “good” data by considering ALL aspects of “threat”, however it might be
defined, over time and across geographical and social boundaries. In doing this, the ML will
not only be more accurately and reliably utilised to identify threat, but it is more likely that
the machine judgement of threat will be trusted by all stakeholders, namely government(s),
people with economic self-interest, and the population at large.

In this sense, trust in the accuracy, legitimacy and fairness (whatever this is) resulting from
ML is the most important factor in assessing whether or not the execution of an ML tool is a
success that adds value to society. Significant changes and developments in business models
and the attitudes of governments and others in positions of power (and society at large) are
likely to be required in order for ML tools to be trusted enough to be effectively deployed.

5 Changes Required for the Successful Implementation
of ML to Identify Threat

Given everything that has been discussed in this paper, there are a number of changes that
need to occur for ML to be successfully or, more correctly, meaningfully deployed.

5.1 The Assumptions when Programming ML

When researching human subjects, it is vital to consider the human idiosyncrasies noted
by Blumer (1989), along with the (much) wider contexts in which all behaviour occurs –
including scientific analysis itself. Three fundamental assumptions must be at the core of ML
programming – the same assumptions for researching all human subjects (e.g. Denzin, 1989;
Blumer, 1989; Dunning, 1999):
• Reality is a social product.
• Humans can guide their own futures and that of others.
• Humans are social beings; they must interact with each other, and gain meaning and
insight in so doing.

Such assumptions can no longer be considered merely as academic niceties, of no value in the
real world. When dealing with ML, such considerations are the new real world.

5.2 Attitudes Towards Data, Privacy and Security

In the new real world, attitudes towards data, privacy and security must, probably and
necessarily, change – not only as a necessity for programming the ML, but as a result of its

7 As an example of what could happen, I cite current attempts to shoe-horn legacy business models and
approaches into social media technology. The superior new technology creates negative outcomes in models
that used to work well with legacy technology.

180