SWE & Research Intern
Stanford GPS Lab · Stanford, CA · Jun — Sep 2024
GPS signal authentication is a hard problem because the spectrum is RF-public and signals can be spoofed. The Stanford GPS Lab works on techniques that let receivers verify a signal really came from the claimed satellite — important for navigation in adversarial environments (military, aviation, financial timestamping).
My work centered on rebuilding the lab's authentication codebase with stronger cryptographic guarantees. The previous version used 16-bit HMAC tags. I increased that to 28–37 bits while keeping the bitrate budget realistic, by aggregating authentication across multiple signal frames and proving the resulting per-frame security guarantee.
I also implemented an error-correction layer that tolerates up to 40% message loss — crucial because in real conditions GPS receivers regularly lose packets to multipath, jamming, or weak-signal environments. The system degrades gracefully instead of dropping authentication entirely under packet loss.
On the engineering side: brought the codebase to 100% test coverage, containerized the development environment so new contributors can be productive on day one instead of fighting MATLAB and toolbox version mismatches, and refactored the simulation harness to be configurable from a single YAML rather than hand-edited scripts.
Highlights
- 01Increased HMAC strength 16 → 28–37 bits via aggregated authentication across signal frames
- 02Error-correction layer tolerating up to 40% message loss without dropping authentication
- 03Reached 100% test coverage on the authentication codebase
- 04Containerized the development environment to remove MATLAB toolbox setup friction
- 05Refactored simulation harness to be YAML-driven rather than hand-edited