Is this project an undergraduate, graduate, or faculty project?
Undergraduate
Project Type
individual
Campus
Daytona Beach
Authors' Class Standing
Sean McConoughey, Freshman
Lead Presenter's Name
Sean McConoughey
Lead Presenter's College
DB College of Engineering
Faculty Mentor Name
Cagri Kilic
Abstract
SHIELD: Semantic Heuristic Intelligence for Ensuring Legitimate Direction
As the automation of robotic systems becomes increasingly prevalent, the risk of malicious hijacking grows in parallel. While robotic systems maintain human oversight for critical operations, they remain vulnerable to communication channel attacks that can compromise mission integrity. Physical components like firmware and sensors offer inherent security through their specialized nature, but the network communication interfaces between robots and operators present significant attack vectors that could lead to mission compromise, physical damage, or asset loss.
This research implements and evaluates a mission assurance framework for ROS2-based autonomous systems. The framework employs a local Large Language Model (LLM) to establish baseline mission parameters, continuously monitor command structures, and detect anomalous instructions that deviate from authorized mission profiles.
Our implementation deploys this framework on a physical robot running ROS2 and custom Python 3 modules. The system captures initial mission parameters as a semantic reference model, then continuously validates incoming commands against this model to identify potential attacks. Upon detecting unauthorized command patterns, the system triggers a fail-safe protocol, suspending operations until secure communication is re-established through predetermined authentication methods or, alternatively, continuing to execute the mission based on predetermined settings.
We evaluate the framework through a series of controlled communication-channel attacks, measuring detection accuracy, false positive rates, and mission completion metrics. Results demonstrate the practical viability of semantic-based mission assurance mechanisms in preserving operational integrity during active communication compromise attempts.
This work contributes to the growing field of cyber-resilient autonomous robotics by providing an implementable approach to mission assurance that balances operational flexibility with security requirements.
Did this research project receive funding support (Spark, SURF, Research Abroad, Student Internal Grants, Collaborative, Climbing, or Ignite Grants) from the Office of Undergraduate Research?
No
SHIELD: Semantic Heuristic Intelligence for Ensuring Legitimate Direction
SHIELD: Semantic Heuristic Intelligence for Ensuring Legitimate Direction
As the automation of robotic systems becomes increasingly prevalent, the risk of malicious hijacking grows in parallel. While robotic systems maintain human oversight for critical operations, they remain vulnerable to communication channel attacks that can compromise mission integrity. Physical components like firmware and sensors offer inherent security through their specialized nature, but the network communication interfaces between robots and operators present significant attack vectors that could lead to mission compromise, physical damage, or asset loss.
This research implements and evaluates a mission assurance framework for ROS2-based autonomous systems. The framework employs a local Large Language Model (LLM) to establish baseline mission parameters, continuously monitor command structures, and detect anomalous instructions that deviate from authorized mission profiles.
Our implementation deploys this framework on a physical robot running ROS2 and custom Python 3 modules. The system captures initial mission parameters as a semantic reference model, then continuously validates incoming commands against this model to identify potential attacks. Upon detecting unauthorized command patterns, the system triggers a fail-safe protocol, suspending operations until secure communication is re-established through predetermined authentication methods or, alternatively, continuing to execute the mission based on predetermined settings.
We evaluate the framework through a series of controlled communication-channel attacks, measuring detection accuracy, false positive rates, and mission completion metrics. Results demonstrate the practical viability of semantic-based mission assurance mechanisms in preserving operational integrity during active communication compromise attempts.
This work contributes to the growing field of cyber-resilient autonomous robotics by providing an implementable approach to mission assurance that balances operational flexibility with security requirements.