Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study evaluates the latency-accuracy trade-offs of various TinyML-compatible classical models—specifically Random Forest, Logistic Regression, SVM, and MLP—for detecting cyber-RF threats in autonomous spacecraft using the SPARTA attack model. The authors conduct a thorough theoretical analysis alongside empirical measurements, revealing that Logistic Regression can achieve microsecond-level inference with only a 1% accuracy drop compared to Random Forest. The findings highlight the potential for enhancing spacecraft cybersecurity through advanced feature encoders and multi-timescale learning architectures, paving the way for more robust onboard threat detection systems.
Logistic Regression can achieve rapid inference for cyber threat detection in spacecraft with minimal accuracy loss, setting a new standard for TinyML applications in space.
Autonomous spacecraft require rapid, lightweight, and reliable onboard detection of cyber-RF threats. Using the SPARTA attack model, we analyze the latency-accuracy trade-offs of TinyML-compatible classical models -- Random Forest, Logistic Regression, SVM, and MLP -- for detecting uplink jamming, Fake-NR spoofing, payload manipulation, ground-segment compromise, and unauthorized command injection. We present a physics-informed theoretical analysis of each model's computational complexity, VC dimension, Lipschitz continuity, and latency scaling, supported by empirical measurements on adversarial RF spectrograms generated via BandErasure, FakeNR, and NoiseBurst corruption modes. Results show that Logistic Regression achieves microsecond-level inference with only a 1\% accuracy drop relative to Random Forest, making it an effective TinyML baseline for onboard autonomy. The study also identifies opportunities for advancing spacecraft cybersecurity through richer feature encoders and multi-timescale learning architectures, building on recent progress in edge intelligence and trustworthy AI.