Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Adversarial Machine Learning Threats to Spacecraft

Rajiv Thummala, Shristi Sharma, Matteo Calabrese, Gregory Falco

TL;DR

The paper addresses AML threats to autonomous spacecraft by introducing a threat taxonomy tailored to space systems and validating it through simulations on NASA's Core Flight System (cFS) and On-board Artificial Intelligence Research (OnAIR). It demonstrates two practical AML attacks—poisoning the GNC model and evading on-board computer vision—highlighting significant degradations in navigation accuracy and landing-safety assessments. The findings underscore the urgent need to integrate AML-focused defenses into spacecraft design, operation, and ground segments, including data integrity, monitoring, and cross-domain collaboration for resilient autonomous space operations. The work lays a foundational framework for threat assessment and mitigation, with future directions toward higher-fidelity digital twins and edge-learning defenses.

Abstract

Spacecraft are among the earliest autonomous systems. Their ability to function without a human in the loop have afforded some of humanity's grandest achievements. As reliance on autonomy grows, space vehicles will become increasingly vulnerable to attacks designed to disrupt autonomous processes-especially probabilistic ones based on machine learning. This paper aims to elucidate and demonstrate the threats that adversarial machine learning (AML) capabilities pose to spacecraft. First, an AML threat taxonomy for spacecraft is introduced. Next, we demonstrate the execution of AML attacks against spacecraft through experimental simulations using NASA's Core Flight System (cFS) and NASA's On-board Artificial Intelligence Research (OnAIR) Platform. Our findings highlight the imperative for incorporating AML-focused security measures in spacecraft that engage autonomy.

Adversarial Machine Learning Threats to Spacecraft

TL;DR

The paper addresses AML threats to autonomous spacecraft by introducing a threat taxonomy tailored to space systems and validating it through simulations on NASA's Core Flight System (cFS) and On-board Artificial Intelligence Research (OnAIR). It demonstrates two practical AML attacks—poisoning the GNC model and evading on-board computer vision—highlighting significant degradations in navigation accuracy and landing-safety assessments. The findings underscore the urgent need to integrate AML-focused defenses into spacecraft design, operation, and ground segments, including data integrity, monitoring, and cross-domain collaboration for resilient autonomous space operations. The work lays a foundational framework for threat assessment and mitigation, with future directions toward higher-fidelity digital twins and edge-learning defenses.

Abstract

Spacecraft are among the earliest autonomous systems. Their ability to function without a human in the loop have afforded some of humanity's grandest achievements. As reliance on autonomy grows, space vehicles will become increasingly vulnerable to attacks designed to disrupt autonomous processes-especially probabilistic ones based on machine learning. This paper aims to elucidate and demonstrate the threats that adversarial machine learning (AML) capabilities pose to spacecraft. First, an AML threat taxonomy for spacecraft is introduced. Next, we demonstrate the execution of AML attacks against spacecraft through experimental simulations using NASA's Core Flight System (cFS) and NASA's On-board Artificial Intelligence Research (OnAIR) Platform. Our findings highlight the imperative for incorporating AML-focused security measures in spacecraft that engage autonomy.
Paper Structure (27 sections, 3 equations, 6 figures, 3 tables)

This paper contains 27 sections, 3 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Background and Prior Art
  3. Adversarial Machine Learning Overview
  4. Classifications of Adversarial Machine Learning
  5. White Box Attacks
  6. Black Box Attacks
  7. Transfer Attacks
  8. Existing Studies
  9. AML Threat Taxonomy for Spacecraft
  10. Parameters
  11. Mission Objectives and Operational Context
  12. Resource Constraints
  13. Learning Architecture, Method, and Stage
  14. Storage Architecture
  15. C&DH Accessibility (Uplink/Write Access)
  16. ...and 12 more sections

Figures (6)

  • Figure 1: AML Threat Decomposition for Spacecraft
  • Figure 2: Perseverance Rover Terrain Relative Navigation System Attack Overview
  • Figure 3: Baseline Performance for Nominal GNC Algorithm
  • Figure 4: Performance of Poisoned GNC Algorithm
  • Figure 5: Performance of Nominal Terrain Navigation System
  • ...and 1 more figures