If you’ve watched a product launch from one of the technology giants such as Apple, Microsoft or Google in the past two years, you would be hard-pressed not to have heard of machine learning.
In addition to Artificial Reality or Augmented Reality, it’s currently one of the biggest buzzwords in the IT world.
Machine learning has the power in future to add value to the physical security function by providing greater insight into what is going on within our organisations.
What exactly is machine learning in physical security?
According to Wikipedia, Machine learning is a field of computer science where computers have the ability to learn without being given explicit instructions or programming.
In a very broad sense, this means that computers can think for themselves.
What can machine learning in physical security do for me?
Rather than this leading to computer operated drones replacing security staff in a George Orwell envisioned future, I see machine learning being able to interpret and analyse data contained within physical security systems.
Historical Pattern Analysis
Imagine an organisation where most employees start and end their workday consistently within a 1-2 hour time range each day.
All of a sudden, two employees start coming into work much earlier and then leaving at the same time as usual.
This could be completely normal and due to them having a seasonal increase in workload, or working on a project with a client in another time-zone which requires them to be in early.
This could also be an indicator that they are up to no good and are trying to avoid being watched.
Some organisations have very busy flows of people entering and leaving their facilities, especially organisations with large estates where everyone starts and leaves at a similar time.
This can have an impact on the local transport networks, and also on ensuring that people entering and leaving the site in a vehicle securely.
Imagine if that organisation had a theoretical maximum of four lanes connecting the site to the motorway.
When we know there will be a large number of employees entering the site, we can switch the signage to allow three lanes to be dedicated to inbound traffic, whilst leaving one lane available for outbound traffic.
Once the peak flow of employees is over, the lanes could switch to allow one lane for delivery vehicles supplying parts or materials to the facility, one lane for incoming visitors and one lane for employees, whilst leaving one lane spare for outgoing traffic.
By analysing stored data within the access control system with machine learning, we can accurately predict when there will be peak flows of traffic, and can use this data to drive the electronic signage to ensure that traffic can be managed throughout the day without impacting on the productivity of the site, or by causing large queues on public roads around the area.
Cleaners will typically check toilet facilities every two hours to ensure that toilet paper, soap and other supplies are re-stocked, and to ensure they are clean and ready for use.
In a large building or estate, this process can be very resource intensive, and therefore expensive.
It is possible that in a large building, some toilets will not have been used at all on a particular day, let alone since the last check 2 hours ago.
If we fitted the main entrance into the toilets with a door contact, wired into the access control system, we could then calculate how many times each toilet location had been used in any one day.
This information can then be used to allow the cleaners to prioritise which toilets they clean and check in which order.
Incorporating Internet of Things technologies, similar to Amazon Dash Buttons, where building users can press a button when an area is out of soap or toilet paper, cleaners could also top up supplies on demand.
Whilst most of these technologies and concepts are theoretically possible, they are yet to be introduced in a physical security system.
One thing is for sure, the future applications of machine learning is very exciting!