Real-Time Video Anonymization at 30 FPS on a $35 Computer

Most privacy pipelines I encountered before building PrivacyGuard shared the same assumption: you have a server. They pipe video frames to the cloud, process them there, and return anonymized output. This works well in San Francisco. It works poorly in Beirut, where the internet goes out without warning, and it works not at all in environments where the entire point is that sensitive data must never leave the premises. The constraint I was designing for was different: detect and anonymize faces, persons, and license plates entirely on-device, in real time, on hardware that costs $35. Here is what that actually required, and what I learned building it. Why edge matters for privacy The framing of "edge AI" usually emphasizes latency or cost. For privacy use cases, the constraint is more fundamental. If you pipe a frame of a hospital corridor to a cloud API, you have already violated the premise. The sensitive data left the facility. The compliance posture is gone before the anonymization