Inspiration
There's been a lot of discourse about automated license plate technology used by government contractors. Cases have emerged of misuse by law enforcement, and the public has a deep mistrust of these companies to handle individual privacy. I wholeheartedly agree with individual privacy first, yet wanted to explore automated safety technology.
What it does
Herd is a privacy-first Automatic License Plate Reader (ALPR) middleware that eliminates mass surveillance while preserving the critical crime-solving benefits of vehicle tracking. Instead of uploading unencrypted video feeds and maintaining searchable location maps of innocent drivers in a cloud database, Herd processes license plate scans locally on edge cameras. Plates that do not match an active hotlist (such as stolen vehicles or Amber Alerts) are dropped from RAM immediately. When a match occurs, Herd logs the query through a cryptographically signed audit, ensuring law enforcement gets actionable leads without exposing everyday citizens to dragnet surveillance in addition to allowing scrutiny into each query made.
How I built it
Python and OpenCV running locally on a Raspberry Pi simulates an edge node (Herd camera) to detect vehicles and parse license plates entirely in RAM. SHA-256 hash chains are used to create zero-knowledge style local hotlist matching. Raw video feeds and non-matching plate data are instantly scrubbed at the edge node. FastAPI backend enforces the privacy policy by only unlocking search queries if it matches details accompanied by a valid case ID or warrant.
I also use an SQLite ledger that signs every query with a private key associated with law enforcement to prevent undocumented searches.
Challenges we ran into
Running real-time character recognition on low-power hardware to simulate real cameras was difficult. Higher scrutiny for privacy also had the risk of increasing latency.
Accomplishments that we're proud of
I'm proud of demonstrating that police can catch suspects without keeping a 30-day searchable movement history of mostly innocent citizens. Automated accountability also ensures that abuse will be minimized or immediately caught out via the secure ledger.
What we learned
Every design decision in a project meant to impact real time individual privacy and security had to be examined closely to ensure privacy standards were never compromised.
What's next for Herd
This was a fun project for my first time at Hack the North. I could see this concept being scaled to the real government contractors that are doing similar, lesser-trusted things.
Log in or sign up for Devpost to join the conversation.