Automating Critical Environments

Improving security and efficiency through AI should not be understated

• By Brent Boekestein
• Mar 01, 2019

Are you tired of hearing about Artificial Intelligence (AI) yet? Did people in the 19th century grow tired of hearing about the wonders of the combustible engine, and how it would surely revolutionize modern life… until it actually did? One can imagine that they, those skeptical of the combustion engine’s positive effects, soon experienced the force-multiplying benefits of traveling from point A to point B more efficiently—and with wind in their hair to boot. The goal of AI is not to replace humans, but to enhance human agency, helping us to accomplish more, better and faster.

While, admittedly, the fervor over AI can be a bit much, but the more reasonable and measurable expectations around improving security, safety and efficiency through AI should not be understated. It strikes me as odd that as we rush to replace traditional vehicles with autonomous ones, there is a reluctance to automate the critical spaces and environments that we find ourselves spending most of our waking time within—work, education, and the various enterprise campus or facility settings.

On average, people spend 100 hours a year commuting to and from work, compared to nearly 2,000 hours spent at work. Is it possible that the way we currently perform important tasks within critical environments will, once AI solutions have been implemented, make about as much sense as choosing to ride horseback from San Jose to San Francisco rather than driving (or being driven there autonomously)?

Where to Begin with AI

As key stakeholders look to implement AI into their workflow, many find that security and safety is a commonsense place to begin. There are three areas currently saturated by humans engaged in manual processes that, when augmented by AI, provide immediate results and quantifiable data from which to further justify large-scale AIcentric deployments. The first, and most visible form is that of a security guard observing streaming security monitors. While the optics of a security guard attentively staring at multiple screens may convey a sense of security, and provide a nominal amount of enhanced perimeter security, in reality a human’s ability to perform this mundane, but critical, task is severely limited.

A study done by the UK Police Scientific and Development Branch (PSDB) concluded that a person tasked with monitoring multiple video screens, with the goal of identifying a person holding an umbrella on a busy street, had accuracy scores of 85 percent, 74 percent, 58 percent and 53 percent when viewing one, four, six and nine monitors, respectively. The study also found that the person was extremely less-likely to detect an object of interest in the background area of the field of view.

Even more alarming is how humans fare when tasked with manually reviewing collected video in post-event situations. While each investigator in the areas of public safety and law enforcement spends an average of 1.5 hours/week reviewing video, which equals a whole month of their annual salary, it’s been found that after only 22 minutes of sustained video review humans lose 95 percent of our visual acuity. Basically, a literal red herring could pass by the screen and a human wouldn’t notice. When corporate security professionals are attempting to resolve employee, guest, or vendor related incidents, this fact alone could seriously hinder the time-to-resolution in critical situations.

Finally, access management is the next logical place to implement AI-powered solutions. The threats to a physical place in human form are various. People are fired from or quit their jobs every day, and sometimes the parties are not parting ways amicably. As we’ve seen far too often, those individuals may wish to return to their place of employment with ill-intent. Maybe it isn’t a former employee, but a third-party vendor, a scorned lover, or a rogue individual with imagined political differences to avenge—the list can go on. When the termination is complete, HR then sends a notice to the security team, who then creates a physical BOLO report and... will keep a vigilant eye out for that person?

Augmented Intelligence

Security cameras, whether fixed or mobile, augmented with AIpowered solutions can provide the crucial first step in mitigating overwhelming physical risks and financial losses that accompany unwanted access by bad actors, through biometric verification, like face recognition, coupled with badge-access safety protocols. In addition to keeping critical assets, like employees, safe while inside the walls of the premises, when someone is leaving work late at night, their face should trigger an auto-response from security to provide offer an escort to the person’s car. It should be automatic, effortless. Similar to driving from point A to point B.

It is far too easy to bypass security measures at points of entry and egress by either swiping a stolen ID badge, waiting for a door to be opened and slipping in, or using a more sociable technique— simply greeting the familiar security guard like one wasn’t just fired the day before. Instead, AI augments a security team’s effectiveness through automated Watchlists, BOLOs and Blocklists that work 24/7 and never tire or look away from the monitor. There is no judgement call made on the part of a security guard, because the person was nice, the AI only knows to do what it’s told: alert security when the person face is detected.

Securing entryways through AI-powered solutions like video analytics gives the security team an unparalleled control over access points. Currently, the manual process involved in each of these scenarios is fraught with inefficiencies. Even something as seemingly harmless as tailgating, or piggybacking (the act of entering a secure entryway without badging-in correctly) can lead to obvious safety concerns and, according to one survey of enterprise security executives, cost the company anywhere from $150,000 to “too high to measure” in losses.

Frictionless, face recognition-powered access management not only speeds the process, but eliminates the vast majority of accessrelated security and safety issues.

Where to Begin?

Returning to the autonomous vehicle comparison, the parallels in production and safety, were we to augment human agency and automate our critical environments through AI, are myriad. According to research done by Business Insider, if 90 percent of cars on roads were autonomous, accidents would plummet from 6 million to 1.3 million, annually. Thanks to better, more efficient driving, CO2 emissions would decrease by 300 million tons per year. And finally, commuters would get to their destinations faster because of less traffic congestions, resulting in 1 billion hours, every day, that could be repurposed for higher-level, more productive activities.

Safer, more productive, and a higher quality of life—for 100 hours of the year. By automating critical spaces with AI, stakeholders in charge of these spaces will reap the same benefits and produce the same types of results.

So how does one dive in? Start small and with reasonable expectations. As covered, deploying AI-powered solutions to better secure perimeters, improve investigations, and shore up access management security protocols provides immediate results and data from which to evaluate the efficacy of the investment. AI-powered solutions should not just be another line item on the budget, but be a force multiplier for security, safety, and productivity now and in the future. What follows are three considerations to keep in mind when evaluating solutions to best fit your needs:

Extensibility of tech. Will the solution work with all of the components of a security stack? In the case of video analytics, one must ensure that the tech will work on the different types of cameras currently in use, plus any cameras that may be added to the stack—drones, wearables, dash cams, etc. As mobile cameras become the new norm, being able to extract actionable intelligence from every sensor available is a no-brainer.

Control over the AI. Many AI deployments fall flat in the long run, and end up causing more user headaches than triumphs, as a result of the limitations of out-of-the-box AI-solutions. While the standard offerings may be powerful, it is important for each customer to be able to create case-specific algorithms that truly meet the varied needs you may encounter. Solutions that are powered by deep learning AI are able to offer the customer unparalleled control over their outcomes.

Policy and privacy. Protecting the privacy and rights of those affected by the AI-solution is of utmost importance to all key stakeholders. A company should have in place a use-policy for biometric data that can be easily accessed by all employees. In addition, partnering with companies that are committed to training their algorithms with unbiased data sets, and that have clear trust and privacy policies in place, will ease the transition to automating the environments that people inhabit on a daily basis.

By adding AI to security and safety processes, it is possible to automate the critical environments of our workplaces, schools, and facilities while simultaneously increasing existing human agency in order to get more done, more efficiently, and with better results.

This article originally appeared in the March 2019 issue of Security Today.