The Smart Door Lock was a home-made IoT door lock I built in high school. It combined voice recognition with smartphone control so a spoken passphrase or a tap on my phone could release the lock. I don't use it these days, but it was one of the first projects where I had to make hardware, software and security decisions all fit together, and I learnt a lot from it.
What It Did
The idea was to replace a key with something more convenient without pretending convenience and security are the same thing. I designed the physical locking side as well as the software that sat behind it.
Voice access. The lock listened for a spoken passphrase and released when it heard the right one
Smartphone control. I could lock and unlock it from my phone
Status and logging. It kept track of the lock state and recorded access attempts
How It Worked
Hardware
A Raspberry Pi was the brain of the system. It handled the audio input, ran the passphrase check and switched the lock through a GPIO relay. Running the speech side locally on the Pi kept the important decisions on the device instead of depending on a remote server.A microphone captured audio for the voice side, and the relay meant the Pi only ever switched a signal rather than powering the lock directly. I also planned for the obvious failure cases, like keeping a physical key as a manual override in case the power or the Pi went down.
Software
The core logic was written in Python, using speech recognition libraries to turn audio into text and then compare it against the stored passphrase. Getting a recent enough Python onto the Pi meant compiling 3.11 from source, which I wrote up in Compiling Python 3.11 from Source on a Raspberry Pi.The service ran headless under tmux so it kept going after SSH disconnected, and a cron job started it again on boot. Small Bash scripts handled routine maintenance.
Security Considerations
A door lock is security-critical by definition, so a lot of the project was spent thinking about how it could be attacked rather than just how it should work.
Voice is a weak secret on its own. A spoken passphrase can be overheard or recorded, so I treated voice as a convenience layer rather than strong authentication
Network exposure. Anything reachable from a phone is reachable from somewhere else too, which pushed me to think about keeping control traffic on the local network where possible
Failure modes. Power cuts, a crashed script or a dead SD card all need a sensible answer, and for a door the answer has to include a manual way in
Logging. Recording access attempts made it much easier to see what the system was actually doing
Working through these properly is a big part of why I ended up studying security engineering.
Development Process
Looked at how existing smart locks approached the problem
Worked through a basic threat model before choosing components
Prototyped the hardware and proved the relay and lock worked before adding voice control
Tuned the voice recognition and the smartphone control, then tested the whole system together
Technology Stack
Hardware
Raspberry Pi
GPIO relay driving the lock
Microphone for voice input
Software
Python for the core logic
Speech recognition libraries
tmux, cron and Bash for running the service headless
What I Took Away From It
The lock worked well enough to prove the idea, and building it taught me far more than reading about IoT security would have.
Hardware and software integration, from GPIO wiring through to the service running on the Pi
How quickly convenience features turn into attack surface
Practical Linux skills on constrained hardware, including compiling Python from source and keeping a headless service alive
Why physical fallbacks matter for anything that controls access
It's retired now, but a lot of what I do in my home lab and at university traces back to this project.
