Abstract

A satellite is a million dollar space asset that is moving at approximately 17,000 miles per hour. After being hacked, ground control could only find out that the satellite was hacked when it had orbited for hundreds of miles, at which point it would be too late. This delay is the main problem that is addressed in this paper. OrbitalMind AI is an independent, on-board security system. It tracks its own health, identifies cyber threats as they occur, disseminates threat information to other satellites in the same constellation and changes its orbit to avoid debris without needing to be instructed by Earth. We tested 12 simulated satellites with the help of real NORAD data, which resulted in the detection of anomalies with a precision of 97.2%; 93.5% of the simulated attacks were blocked and recovery from induced failure was achieved in less than 8 seconds. In all tests, the system was able to detect GPS spoofing, command injection and signal jamming and collisions with simulated debris.