A Street That Keeps Watch
When one house is in trouble, the street should know in seconds. All it shares is the alarm.
When one house is in trouble, the street should know in seconds. All it shares is the alarm.
Most of us find out about a break-in on our street the next day.
It happens at two in the morning. A window breaks three doors down. The owners are away, or asleep upstairs. Two houses across the road have cameras pointed at the street, and both record the whole thing. Nobody watches the footage until the police ask for it, days later.
Fire is worse. A smoke alarm in an empty house is very loud, and it can be heard about as far as the front fence. By the time a neighbour notices smoke, the house is often lost.
The pieces to do better are already on the street. Most houses have a Wi-Fi router. Many have a camera, a smoke alarm and a doorbell that knows when someone is at the door. What is missing is a way for those houses to warn each other.
So consider a street that keeps watch for itself. When one house is in trouble, every house around it knows within a second.
Two things make this work: an agent in every house, and an alarm channel between the houses.
The house agent. In the near future, I expect most homes to have a software agent that looks after the house: it knows who is home, watches the sensors, locks the doors and manages the power. It already has the information to tell a real emergency from a cat on the windowsill. Give it one more job: when something is wrong, warn the neighbours.
The alarm channel. Every home router gets a small, separate channel that does one thing: pass alarms to the routers around it. It is not part of your home network and it carries nothing else. Routers reach about five neighbours each, about 40 metres away through two walls. An alarm hops router to router, as the plate shows, and reaches every house within 80 metres in well under a second.
The alarm does not use the internet. That matters, because the internet is the first thing to fail. A careful burglar cuts the line. A fire burns through it. A storm takes down the exchange. The alarm channel still works, as long as the routers have a little power.
A street alarm is only useful if people do not learn to ignore it. Car alarms are ignored because they go off for no reason. So the agents need to be careful about who they wake, and when. There are only two kinds of alarm.
Anything less urgent than that is not the street’s business, and the channel does not carry it.
This is the part that matters most, because nobody wants strangers on their Wi-Fi. They would be right to refuse.
If the owner of a camera chooses, they can share a short clip with the police afterwards. That is a separate decision, made by a person, not by the system.
Rough figures, for a typical Australian suburb.
| Quantity | Value |
|---|---|
| Width of a house block | ≈ 15–18 m |
| Router-to-router range, 2.4 GHz, through two walls | ≈ 30–50 m |
| Neighbours in range of each router | ≈ 5 |
| Houses within 80 m of an alarm | ≈ 12–15 |
| Hops to reach them | ≈ 3–6 |
| Size of one alarm | a few hundred bytes |
| Delay per hop for a small message | a few ms |
| Time to alert the whole 80 m | under 1 s |
| Router power draw | ≈ 5–10 W |
| Battery for a night without power | ≈ 100 Wh |
| Time to discover a 2 am break-in today | often 5 hours or more |
The numbers are not the hard part. An alarm is a tiny message, and a few hops between home routers carry it almost instantly. The hard parts are human.
The word “network”. Tell people their router will join a network with the neighbours, and many will say no. They are right to be wary. The answer has to be the design itself: a channel that carries only alarms, cannot reach the home side, is opt-in, and is open for anyone to check. It should be described the way it is used, as a street alarm, not as networking.
False alarms. One false alarm a week, and the street stops listening. The agents must be strict: two sensors agreeing, and the owner given a chance to cancel.
Suspicion. A street that watches itself can turn on the people who look different from its residents. The system should report events, such as a window breaking, never people who simply walk past. No faces, no profiles, and no list of “suspicious” visitors.
Safety. The alarm must never send people towards danger. A woken neighbour should turn on their lights, lock their doors and call for help, not go outside to confront anyone. The agents can say exactly that.
Trust and spoofing. A fake alarm could empty a street or send a family running into the cold. Every alarm must be signed by a real house on the street, and an agent that raises false alarms must be ignored by the others.
Who is included. Not everyone has a smart home or wants one. An older neighbour living alone may need this most. The simplest member should be a plug-in alarm with a big button and a battery, not a phone app.
Privacy and the law. Recording the street, and sharing what was recorded, is regulated, and the rules differ between states and countries. Keeping the video inside each house avoids most of the problem, but not all of it.
Emergency services. An automated call to 000 from a street of agents is not something emergency services should take on trust. The link to them needs to be agreed with them, probably through the monitored alarm companies that already do this for individual houses.
What I add is the combination: agents that already know the house, a channel that carries nothing but alarms, and a warning that reaches the people close enough to help.
A good neighbour has always been the best alarm. This just gives them better hearing.