Connected-device security coverage tends to arrive in one of two unhelpful shapes. Either everything is fine, or someone is watching your children through the baby monitor. Neither is a good guide to what actually happens.
What actually happens is more mundane and more common: a cheap device with software nobody maintains sits on your network for six years, and eventually that becomes somebody's problem. Usually not yours — which is precisely why it goes unnoticed.
Why these devices are weak
A smart plug is not a plug. It is a small computer running a stripped-down Linux, a network stack, and often a web server, sold for eleven pounds with a margin of pennies.
That price constrains everything:
- No security budget. There is no team reviewing the firmware, because there is no money for one.
- No update mechanism, or one nobody uses. Many devices can only be updated through an app the owner opened once.
- Reused components. Whole categories of device are built on the same reference design from the same chip vendor. One flaw in that design affects dozens of brands under different names.
- A support window measured in months. The product is discontinued; the device stays plugged in for a decade.
- Default credentials. Less common than it was, but still present in the cheapest tier.
The result is a class of device that is permanently connected, rarely patched, and effectively invisible.
The four things that realistically go wrong
1. It joins a botnet
This is by far the most common outcome, and it is undramatic. Automated scanners sweep the internet continuously for devices with known flaws or default logins. When one is found, it is enrolled into a network of compromised devices and used to attack other people.
You will probably never notice. The symptoms are a slightly slower connection and, occasionally, your address appearing on an abuse list. The device keeps working perfectly, because breaking it would draw attention.
2. It becomes the way in
More concerning. A compromised device on your network is a foothold behind your router, on the same flat network as your laptop, your file storage and your phone. Most home networks trust everything inside them completely, which means a plug in the kitchen can reach your work laptop.
This is the reason for network separation, covered below. It is the single highest-value action in this article.
3. It leaks more than you expected
Not through compromise — through normal operation. Many devices report far more than their function requires: connection times, room occupancy patterns, precise usage schedules, sometimes audio snippets for "quality purposes". This is disclosed in a privacy policy nobody read, and it is legal.
The practical risk is aggregation. Occupancy patterns from a thermostat, combined with location from a phone app and purchase history, is a detailed picture of a household held by companies you did not choose.
4. The company disappears
The under-appreciated failure. When a vendor shuts down its cloud service, cloud-dependent devices stop working. Locks that will not open from the app, cameras with no feed, thermostats reduced to manual dials. This has happened repeatedly and there is no consumer protection that meaningfully addresses it.
Before purchase, ask one question: does this device still perform its core function if the manufacturer's servers vanish tomorrow? If the answer is no, you are renting it, and the lease has no stated end date.
What to do, in order of value
1. Put untrusted devices on their own network
This is the big one, and it is far easier than it used to be. Most current routers support either a guest network or several separate wireless networks.
Create a second network. Put every connected device on it — plugs, bulbs, cameras, televisions, speakers, appliances. Keep laptops, phones and storage on the main one. If the router offers client isolation for that network, enable it.
Now a compromised bulb can reach the internet and nothing else. This does not require you to trust any individual device, which is the point: you cannot audit a bulb, so build a network that does not need you to.
2. Change the router's admin password and update it
The router is the one device where compromise is catastrophic, because it sees all your traffic. Change the administration password from the default, apply firmware updates, and turn off remote administration from the internet unless you specifically need it.
3. Turn off UPnP
UPnP lets a device on your network ask the router to open a port to the internet, without asking you. It exists for convenience and it undoes a great deal of your firewall. Some games and consoles want it; if you can live without it, disable it.
4. Never expose a camera directly to the internet
If you want remote access to a camera, use the manufacturer's cloud (accepting the privacy trade) or a VPN into your own network. Port-forwarding a camera to the open internet puts it in front of every scanner on earth, and search engines exist specifically to index those.
5. Audit what is actually connected
Open the router's client list. Most people find something they had forgotten: an old tablet, a previous tenant's device, a smart speaker in a spare room. Remove what you do not use. Every device you unplug is one you never have to patch.
6. Update what you can, once a quarter
Fifteen minutes with the apps you never open. Not thrilling, but firmware updates are where the actual fixes live.
A rough risk ranking
| Device | Why it matters | Priority |
|---|---|---|
| Router | Sees everything, controls everything | Highest |
| Cameras and microphones | Direct privacy impact if compromised | High |
| Smart locks and garage doors | Physical access | High |
| Network storage | Your data, often unencrypted | High |
| Televisions and streaming boxes | Extensive tracking, long support gaps | Medium |
| Plugs, bulbs, sensors | Low individual impact, common botnet target | Medium |
| Kettles, scales, novelty items | Minimal function, permanent exposure | Reconsider owning |
Buying better
For anything you keep for years, the questions worth asking before you pay:
- Does it work without the vendor's cloud?
- Is there a stated security support period, in writing?
- Does it use a common local protocol, or only the manufacturer's app?
- Has the vendor shipped firmware updates for its older products?
- Is there a physical control that works when the software does not?
A device that passes all five costs more. It is also the one still working, and still patched, in 2032.
The proportionate summary
Nobody is targeting you personally through your light bulbs. The realistic risks are that a cheap device is quietly conscripted into attacking strangers, that it is used as a stepping stone to something of yours that does matter, and that the whole product line stops working when a company loses interest.
All three are substantially mitigated by the same two decisions: separate the network, and prefer devices that work without a cloud. Do those and the remaining risk is genuinely small. Skip them and the odds are simply that the consequences will not be visible to you.