Home Reviews About
Twenty of Time

Building a privacy-first smart home with open-source tools

The lights blink on as you walk through the door, the thermostat has already dialled the temperature down a few degrees, and a small speaker announces the arrival of guests you had been expecting all afternoon. None of this required anything more than a passing thought. Connected homes have become that ordinary in Australia, where JB Hi-Fi, Bunnings and a long tail of online stores push smart speakers, cameras and energy monitors into millions of living rooms. The trade behind the convenience, however, is rarely discussed on the shop floor. Every motion sensor, cloud bulb and voice assistant quietly ships behavioural data to servers on the other side of the world, then pays third parties to make sense of it for advertisers.

This guide is for readers who would rather keep that information inside their own four walls. It assumes a typical Australian household, with the National Broadband Network feeding Wi-Fi to a router somewhere near the television, and walks through how to build a smart home on open-source foundations. Nothing here relies on a subscription, an account or a vendor's good faith. By the end, the lights should still come on when you walk through the door, but the data trails that follow you around should stop at your router.

Why privacy is the actual product

Surveillance is not a side effect of the smart home economy; it is the actual business model. Devices are often sold at a loss because the vendor plans to monetise the data flow long after the box has been plugged in. Friso van Dijk has argued the same point more broadly across the internet in his surveillance business model essay, and the logic applies just as forcefully to the kettle that knows when you boil it. Australian law does not yet provide a sturdy shield against this. The Privacy Act 1988 and the Australian Privacy Principles govern how "personal information" is handled, but metadata about daily routines sits in a much greyer zone, and the Notifiable Data Breaches scheme only catches disclosures that companies bother to report.

A privacy-first home flips the default. Instead of devices shipping raw events to a remote cloud, every reading stays on hardware you control and software you can audit. Open-source projects publish their code, accept community review and let you verify what is and is not being transmitted. That is not a guarantee that nothing leaks, but it is the only meaningful starting point for a household that wants to know what its own technology is up to.

Picking hardware you can actually buy here

Open-source software is most useful when it runs on hardware that you can swap, repair and replace without a three-week courier wait. Australians have a healthy kit of local options for this. Single-board machines such as the Raspberry Pi 4 and 5, the Orange Pi 5 Plus and the Odroid M1 sit on shelves at Core Electronics in Newcastle, at Altronics stores from Perth to Hobart, and through the usual secondary-market channels in Brisbane and Sydney. Zigbee and Matter coordinators from brands like Sonoff, MOES and Aqara can be ordered through Amazon Australia, Bunnings or specialist outlets such as Rob's World in Queensland without paying international shipping twice over.

For end-device switching and sensing, the trick is to choose components that talk standardised protocols rather than cloud-to-cloud glue. Zigbee and Z-Wave products, after a brief pairing step, report to a local coordinator and never touch a vendor server by default. Wi-Fi sensors based on the ESP32 chip are also worth considering, particularly for the keen hobbyist, and projects shared on international hubs like the Ortolandia community walk through flashing them with Tasmota, ESPHome or a fork of your choice. Treat any device that genuinely requires a vendor app and account as a bad fit for a privacy-first build, no matter how many five-star reviews it has on a local shopping portal.

Software foundations and where the data lands

Hardware without software is just expensive plastic. The centrepiece of almost every open-source smart home today is Home Assistant, a community-maintained platform written in Python that runs comfortably on a Raspberry Pi 4 or a small x86 box from brands such as Protectli. It speaks hundreds of integrations natively, so a Zigbee bulb paired through a coordinator appears as a configurable light without ever leaving the home network. Companion projects such as ESPHome, Node-RED and Frigate for camera processing extend the platform further, and the surrounding community offers dashboards, voice assistants and energy monitors that plug straight in. Newcomers often pick up solid first-build checklists from home-focused resources such as Easy Organized Home, which covers the everyday side of coordinating devices around actual routines rather than chasing abstract privacy scores.

Solar households in particular will find rich territory here. With Australian rooftop solar now generating more than a tenth of national electricity, monitoring feeds from inverters such as Fronius, Enphase, GoodWe and SolarEdge can flow into Home Assistant and pair with smart plugs, battery status and tariff data from retailers like AGL, Origin or EnergyAustralia. Enthusiasts in regional New South Wales and South Australia have been quick to combine that information with weather feeds from the Bureau of Meteorology, letting blinds, pool pumps and reverse-cycle air-conditioners react to free solar surplus rather than grid imports. None of it requires handing hourly production data to a vendor portal.

Locking down the network and the NBN

A private home is only as secure as the gateway behind the Wi-Fi password. In Australia that gateway is usually a hybrid modem supplied by the ISP, with models from Telstra, TPG, Optus and Aussie Broadband dominating retail shelves. These units are functional but rarely inspire confidence in security research, and several have shipped with cloud-management accounts turned on by default. The first step in any privacy-first build is to switch the modem to bridge mode where possible, connect a separate router running something like OpenWrt or pfSense, and let it handle DHCP, firewalling and DNS. Using a local DNS resolver such as AdGuard Home or Pi-hole prunes advertising, tracking and the bulk of telemetry that some smart-home devices still attempt.

Travel and mobile backups matter too. Anyone catching the Sydney-to-Melbourne XPT or the morning service from Frankston will recognise the convenience of station Wi-Fi, and the risks are laid out in a train station Wi-Fi analysis elsewhere on this site. Bringing a small travel router with a built-in VPN client, or tethering through a phone on a Telstra or Vodafone pass, keeps smart-home configuration traffic off shared hotspots. The same principle applies at home, where every smart device should ideally speak only to the local coordinator and lose outbound internet access entirely if its protocol does not need it.

A realistic rollout you can finish in a weekend

Most readers do not need to rip and replace on day one. A workable rollout moves room by room and ends with a household where every device either has a clear local path or has been unplugged and replaced. The summary below maps the most common open-source platforms against what they actually require to install and run on Australian networks.

Platform Core hardware Cloud dependency Local control Typical install effort
Home Assistant Raspberry Pi 4/5 or low-power x86 box Optional, used only for remote access Full, with automations and dashboards One to three hours for a first install
OpenHAB Raspberry Pi, Linux server or Docker host Optional Full, with a rule-based engine Two to four hours for first-time users
ESPHome ESP32 or ESP8266 boards flashed from a PC None once flashed Device-level, integrated through Home Assistant About twenty minutes per device
Tasmota ESP-based devices reflashed over Wi-Fi None once flashed Device-level, MQTT or HTTP integration Ten to thirty minutes per device
Frigate Coral USB or Hailo-8 accelerator None for detection Local recording and motion indexing Half a day including storage layout

A rough Australian weekend plan looks like this. Day one: install Home Assistant on a Raspberry Pi 4 from a kit bought at Core Electronics, plug a Zigbee coordinator into it and pair every bulb and sensor that already sits on standard Zigbee. Day two: flash any Wi-Fi devices with ESPHome or Tasmota, retire anything that refuses to be reflashed, and set up AdGuard Home on the network. Day three: automate the obvious routines, monitor the logs for a week, and only then expose the lot to the outside world through a reverse proxy such as Caddy with proper TLS. By the time a heatwave rolls in and the system actually earns its keep by trimming an AGL bill, the household will own a smart home rather than simply subscribe to one.

The underlying argument here is not anti-technology but pro-sovereignty. A smart home built on open-source foundations answers to its occupants rather than to a vendor, an advertiser or a foreign cloud. Readers who want to follow the broader set of essays on this site can begin at the about page. When the technology is auditable and the network is local, the household finally owns the data its routines generate.