CyLab Robotics Security and Privacy Initiative

Research Challenges

Through the multi-year initiative, CyLab intends to develop a suite of novel foundations and technologies that address the following research challenges:

Foundational Concepts: Defining Threats and Approaches

Threat Models

In addition to attackers against traditional computing infrastructure, specialized threat models for robotics systems are needed because they frequently interact with the physical world and humans in the environment, and sometimes operate under real-time constraints. RSPI addresses these specialized threat models and offers strategies to help mitigate threats.

Systems Approaches to Security & Privacy in Robotics

Robotic systems involve the integration of complex mechanical, electrical, and software components within dynamic and uncertain environments. These integrations can create emergent security and privacy challenges that cannot always be anticipated in advance. RSPI explores the role of system-wide strategies for improving security and privacy in robotics, including assurance cases, risk assessment methodologies, and security- and privacy-by-design principles.

Core Components and Supply Chain: From Hardware to Software

Sensing and Hardware Privacy/Security

Undermining the security of a sensor or another component can render the data from that sensor useless or even undermine all downstream decisions from that component. On the flip side, the earlier in the pipeline privacy can be addressed, this can potentially aid confidentiality and security down the line. RSPI examines attacks and defenses on security and privacy for sensors and other hardware components.

Supply Chain Security re: Software and Hardware

Robotics systems contain complex software components and hardware components, which may come from a large number of sources, leaving a large attack surface for attackers to modify the component, causing security, privacy, or safety issues. RSPI researchers explore challenges and potential solutions to the supply chain security in the context of robotic systems.

Privacy & Security in Robotics Software Ecosystems

Software frameworks, middlewares, and package ecosystems such as ROS 2 enable modular and reusable robotics development but also introduce security, privacy, and safety challenges. RSPI identitifies the safety, security, and privacy challenges that arise from robotics software and explores the space of potential solutions.

Safety x AI and Machine Learning

Powerful AI technologies such as LLMs are increasingly being deployed in robotic systems, but they also introduce new, unique challenges for security, safety, and privacy. RSPI identifies potential use cases and risks of using AI in robotic systems and develops approaches for mitigating those risks.

Evaluation and Testing

Evaluation and Testing Frameworks

RSPI appraises current and develops new approaches to evaluating the security and privacy of robotic systems, and identify gaps and challenges in evaluation metrics, benchmarks, verification methods, and simulation-based and field testing.

External Factors: The Human and Regulatory Context

Human Factors

Many an allegedly-secure system has broken down as a result of its failure to address the human element. RSPI explores the ways humans interact with robotics and AI that go beyond interactions with other automated systems, and seeks to identify key questions and answers regarding human factors in security and privacy of robotics.

Policy Considerations

What regulations, certifications, standards, or policies should (or do) apply to security and privacy aspects of robotics? RSPI addresses this question in the context of existing regulations covering some aspects of robotics (e.g., in the U.S., FAA for flying robots and FCC for communication, FTC actions addressing unfair or deceptive practices resulting from poor security or privacy, or regulations governing medical devices and health data).

Become a sponsor

Interested in getting involved or sponsoring this initiative? Contact Michael Lisanti, Senior Director of Partnerships, at 412-268-1870 or mlisanti@andrew.cmu.edu.