Turning misconceptions of AI into dynamic security solutions
- By Bill Brennan
- Apr 01, 2021
There is no doubt that Artificial Intelligence (AI) is
here to stay, as new AI-driven solutions are pervasive
across every category of professional and consumer
electronics today. This is especially true in
the professional security arena – both in terms of
physical and cyber security – as exciting new AI-driven solutions
promise to move the industry from passive monitoring to predictive
analysis modes of operation. The ability to autonomously
collect and analyze vast amounts of data from holistically integrated
security systems also opens-up a myriad of new business
intelligence applications across the enterprise.
Yet, despite the pervasiveness of AI technology, it is still widely
misunderstood, and that is understandable after hearing comments
from some of the most respected technology gurus of our
time who portend that AI will be death of civilization or believe
that sci-fimovies portray an accurate picture of the future. It is important
to separate media grandstanding and fiction from the real
science behind AI and the pragmatic benefits it promises to deliver.
For the purposes of this discussion, let’s focus on the latter purely
from the perspective of the professional security market.
Video surveillance, access control and other related physical
security systems have progressed in quantum leaps over the past
several years. It was not all that long ago that we were all looking
at grainy B&W images from analog CCTV cameras, and access
control was relegated to keys and guards who manually logged
individuals entry/egress activities with a ballpoint pen.
The truth of the matter is that the seemingly rapid evolution
from analog to digital to networked, and now intelligent state of
technology, has actually been in the works for quite some time – actually
since the 1940s. This is when research first began on developing
computational models that would eventually evolve into artifi-
cial neural networks, which provide the foundation and processing
power for AI and machine learning technology as we know it today.
According to Wikipedia, artificial neural networks, otherwise
referred to as connectionist systems, are computing systems that are
inspired by biological neural networks like those in human brains.
Such systems have the capacity to “learn” to perform tasks by considering
examples, generally without being programmed with taskspeci
fic rules. This is essentially how the term “artificial intelligence”
and the alternate term for AI, “machine learning”, came into being
as a means of describing a new generation of smart machines and
devices capable of making autonomous decisions.
Many of today’s new physical security solutions actually live up
to the long-standing expectations of AI with the ability to autonomously
learn and analyze situations and then make decisions based
on the data. These new solutions employ one or more of three different
electronic learning techniques: Supervised learning. Where
continuous feedback is presented to formulate decisions.
Unsupervised learning. Where decisions are based on estimation
of statistical data received by an input source or device,
Reinforcement learning. Where rules are adjusted based on a
combination of estimating data inputs and pre-established rules
to reach the desired results.
There are several relatively new, yet highly familiar professional,
security solutions trending today that adhere to these AI models,
including intelligent AI surveillance cameras, dynamic and
adaptive facial recognition, video redaction, and physical and
software enabled IoT sensor technologies. When used in combination,
the aforementioned solutions transform traditional security
monitoring and recording systems into sources of actionable
intelligence enabling new forms of predictive analysis.
Unfortunately, it often takes a cataclysmic event to identify
the need for new and innovative security solutions that migrate
into the mainstream thereafter. The ongoing pandemic is no exception.
New integrated AI-driven security and health safety solutions
have made their way to market to help address current
challenges and compliance mandates.
- Automatically identifying, authenticating and tracking individuals
upon entering a facility or campus.
- Automatically identify individuals who are not wearing face
masks in environments where masks are required.
- Automatically detecting and identifying individuals with elevated
- Automatically restricting access or redirecting individuals who
may require additional screening or attention.
- Redacting innocent individuals’ faces from video that is being
used to apprehend suspects or as video evidence.
- Detecting anomalies in the environment or standard operations
that pose a potential threat or liability.
Although we now perceive these intelligent integrated solutions
as specific measures to combat the pandemic, we should
anticipate that they will become permanent criteria for safe and
secure facilities and healthy buildings moving forward. A perfect
example of this was the widespread adoption of metal detectors
and body scanners following 9/11.
This level of integrated intelligence represents the tip of the iceberg
in relation to the wide range of new and emerging professional security
applications for AI-driven software and hardware solutions. This
is relevant to new and relatively inexpensive sensors and analytics that
can easily supplement traditional surveillance cameras and software
solutions such as VMS and access control management platforms.
The additional data from new sensor solutions further increases
situational awareness by providing the ability to detect potentially
threatening anomalies. Examples include everything from audio analytics
to detect and identify screams, gunshots, explosions and glass
breaking to environmental sensors that recognize noxious fumes and
gasses, smoke and fire, and feel vibrations such as
explosions or earthquakes.
As we constantly hear on the news these
days, it is important that we “follow the science”
as it pertains to AI – which is a win-win for security
professionals, users and integrators alike.
This article originally appeared in the April 2021 issue of Security Today.