On a remote Australian mine site, whether a trespasser gets in comes down to something a fence cannot do: watch the boundary at the moment someone crosses it.

It is two in the morning on a mine site three hundred kilometres from the nearest town, and a set of headlights has just gone dark outside the fence. The perimeter here is textbook. Full-height fencing along the whole boundary, signage up, the line properly marked and properly maintained. On paper the site is secure, and the paper is the only thing anyone will check before the incident. Someone walks the fence in the dark until they reach the stretch that backs onto an old rehabilitation pit, where the ground is soft, the lighting stops, and no vehicle has passed in days. They are over it in under a minute.

The fence did exactly what it was built to do. It marked the line. What it could not do was notice the line being crossed, and on a boundary that runs for kilometres, nobody else noticed either. A fence marks a boundary. It does not watch one. On a small yard those are close enough to the same thing, because someone is usually near enough to see. On a remote site they are entirely separate jobs, and almost everything that goes wrong with trespass lives in the space between them.

So when a site has a trespass problem, the reflex is to spend more on the marker. A taller fence, a second line of it, heavier gauge, more signage. Every dollar of that goes into the job the fence already does well, and none of it into the job it cannot do at all. The boundary ends up better marked and no better watched. And the person coming over the back stretch at two in the morning does not care how tall the fence is, because they worked out the site was asleep before they ever touched it. Which, as the research turns out, is exactly how the decision gets made.

Nobody crosses a fence because it looked short

The most useful place to start is not the perimeter but the people who cross it, because how they behave is firmer ground than what a fence is assumed to prevent. The Australian Institute of Criminology ran a long program, Drug Use Monitoring in Australia, that interviewed offenders directly about how they choose a target and how they get in. Two of its findings land hard on the fence-height reflex.

The first is how they choose. Offenders weighed up whether a place was worth entering mainly on visible signs of inactivity, not on how strong the perimeter looked. An empty, quiet, still site reads as low risk. A tall fence around an empty, quiet, still site reads as low risk with a tall fence in front of it. The barrier is not the thing they are solving for. Emptiness is. The second is how fast they move once they have chosen. Most got inside within about five minutes, more or less regardless of the barrier in front of them. Hold that against a patrol route on a large site, where the gap between one pass of a given stretch and the next is measured in hours rather than minutes, and the mismatch is plain. The barrier buys seconds. The intruder needs five minutes.

A fair caution on that research: it is general offender-behaviour work out of Western Australia, not a study of mine sites specifically, and it should be read that way. What it describes is how people who break into places tend to think, and there is no obvious reason a remote pit sits outside that pattern. Read at face value, it undercuts the most common answer to trespass, which is to make the marker bigger.

The same research is just as clear about what does send people away, and it is not the height of anything. Offenders reported being deterred by signs of active presence and response. An alarm that actually sounds. A dog. Lights and movement that say someone is awake and paying attention. What stopped them was the prospect of a reaction while they were still standing there, and a wall, however tall, supplies no reaction at all. There is a sharper edge to this for anyone about to buy hardware: fake and non-functional deterrents did not work. A dummy camera, an alarm housing with nothing wired into it, a sign that says the site is patrolled by a company that stopped patrolling it last year. Offenders could tell, and a hollow deterrent reads as worse than none, because it marks the site as somewhere that gestures at security instead of doing it.

Now describe an unwatched fence in that same language. It announces a boundary, then stays completely silent when the boundary is crossed. It promises a consequence it has no way to deliver. In the terms this research uses, a marked-but-unwatched perimeter sits a good deal closer to the dummy camera than to the dog. The line is drawn. Nothing answers when someone steps over it.

A patrol is somewhere. The perimeter is everywhere.

This is where the mining context stops being general and gets specific, and it is on the public record. In 2023 the Queensland Parliament’s Transport and Resources Committee took evidence that split resource sites cleanly into two kinds. Active, staffed sites tend to lose the most to people who are already inside, employees and contractors. Remote care-and-maintenance sites, the minimally staffed, low-activity ones, get opportunistic trespass and theft from outsiders instead. That second description is a working mine’s own quiet corners word for word: the inactive pits, the rehabilitation zones, the laydown yards and back-of-site access tracks that are between uses. The parts of the operation with the least going on are the parts most exposed to someone wandering in, and by no coincidence they are the same parts a patrol schedule reaches last and least.

Put real distances on it. A remote site’s perimeter can run for kilometres, and a patrol is one vehicle in one place at a time. When it is parked at the main gate, it is fifteen or twenty minutes from the far boundary. The offender research says five minutes is enough. There is no schedule that closes a twenty-minute gap against a five-minute problem, not because the patrol is doing anything wrong, but because one moving point cannot be on every stretch of a long line at once. That is geometry, not effort.

The scale behind this is not a rounding error. The same inquiry heard that more than 500 remote resources and gas sites were hit by theft across a single twelve-month period, with more than 20 sites struck on a single day at the peak. Numbers like that are not a run of bad luck at a few unlucky sites. They are what it looks like when a whole category of boundary is marked and not watched, and a whole population of offenders has worked that out. There is a pressure signal from overseas worth noting for context, not as a local statistic: plant and equipment theft has climbed sharply in other markets on the back of demand for stolen machinery, and the United Kingdom recorded a 200% rise after the war in Ukraine began. That is not evidence the same multiple has landed in Australia, and it should not be dressed up as one. It is a reading of the direction of pressure remote Australian sites should be planning against.

One more figure changes how you should read every number a site keeps on itself. Western Australia has reported 73% under-reporting of metal theft, with most victims lodging a report only when an insurance claim forced the paperwork. Whatever your own incident log says happened along the boundary last year, the honest reading is that more happened than the log admits. The quiet stretches are quiet in the records too.

What it costs once the crossing finishes

Be concrete about what is on the other side of that unwatched fence, because trespass is not an abstraction about lines and permissions. It is someone loading the movable, sellable parts of your site into a ute. What actually walks off remote sites is small, portable and easy to sell: copper cabling, solar panels, batteries, fuel, the guts of a switchboard. The haul trucks stay where they are. The valuable, movable infrastructure does not. And once it is gone, it is gone: under 15% of stolen equipment is ever recovered. Any security plan that quietly leans on getting things back after the fact has already lost, because the numbers on recovery are close to hopeless.

The bill for letting it finish is not theoretical either. A single copper-theft event at GrainCorp’s Kupunn site did roughly $100,000 in damage and cost the site four to eight weeks of lost production. Per incident, direct losses of this kind tend to land somewhere between $18,000 and $100,000 before the downtime is counted at all. Every one of those figures started as exactly the crossing described at the top of this piece: one person, one unwatched stretch, under a minute to get over.

Catch it at the line

That undefended moment, the crossing itself, is precisely where almost all site security has nothing to offer. Fencing and signage work before it, to lower the odds of an attempt. Cameras reviewed later and reports filed the next morning work after it, to describe what already happened. The one window that decides the outcome, the minute someone is coming over a dark stretch of boundary, is the window nothing is built for. Spectur is built for that window and very little else.

The sequence is short, and it happens at the point of crossing. Spectur’s AI detects the intrusion in real time, the moment someone breaches the line. An on-site audible and visual deterrent fires straight away, right there at the boundary, so the active response the offender research says works is present at the one second it counts. And a real-time alert reaches the site manager while the person is still on the ground, not in a report read over coffee the next day. That supplies the active presence offenders back away from, at the exact spot and second an intrusion starts, on the stretches of perimeter no patrol was ever going to be standing on. The boundary stops being a marker that goes quiet at the worst possible moment and starts being a line that answers back. There is an illustrative example of what this can do at scale: across 70 remote sites, a Gallagher and OETC deployment brought unauthorised entries down to zero. That is an international utilities program, not an Australian mining one, and it belongs here as a proof of concept rather than a like-for-like mine result, but it shows that monitoring many remote perimeters in real time, all at once, is achievable and not theoretical.

None of this says the fence was a waste. It marks the boundary and adds a few seconds of physical delay, and those seconds are worth having when there is finally a response on its way to use them. Keep the fence, keep the patrol. Their weakness was never that they are useless, it is that they work at the edges of the problem and go quiet in the middle of it. Spectur is the part that makes the boundary watched instead of only marked, and it earns its place most on the long, dark, low-traffic stretches a patrol schedule can never reach every hour. There is a longer game in it too. Offenders choose targets on how a site reads, and a boundary that answers an intrusion while the intruder is still on it re-rates the site in the exact terms the AIC research uses, from a quiet, worth-a-look target into one that pushes back. It also changes what the site can show an insurer: not a completed-loss register describing what already went wrong, but a log of intrusions caught and answered in progress, which over successive renewals is a pattern an underwriter can actually price in your favour.

Back at the fence, at two in the morning

Go back to that dark stretch of boundary and the person coming over it in under a minute. The fence was never the thing that failed. It did precisely what a fence does, and a taller one would have done precisely the same thing a little slower. The question that actually decided how that night ended was simpler and harder than fence height: at the moment the line was crossed, was anything watching it.

The evidence points one way from every angle. People decide to cross on how empty a site looks, not on how high the fence is. What turns them back is an active response in the moment, not a taller barrier. A perimeter of any real size has long stretches a patrol simply cannot be standing on, and those stretches are where the crossing happens. None of that is fixed by more fence, and all of it is the same underlying gap: a line that is marked and not watched. So here is the reframe the data keeps arriving at. The remote sites getting on top of trespass right now are not the ones with the newest or the tallest fence. They are the ones where every metre of that fence is actually being watched at the moment it matters, and where the boundary has something to say the instant someone crosses it.