Typology of a Mobile Robot's Internal Data and the Threats to Their Security

Abstract

In this work, the internal data of an autonomous mobile robot is considered an asset to be protected. We propose a classification of data by processing level: raw sensor data (LiDAR, cameras, IMU), odometry and localization data, mapping data (occupancy grids, feature maps), ROS/ROS 2 logs and recordings, control commands, and telemetry. For each level, characteristic message types, communication channels, and storage locations are identified. Three groups of threats are analyzed: attacks on sensor data integrity (spoofing), data interception via insecure DDS middleware, and unauthorized data extraction in cases of physical access or robot compromise. The analysis shows significant risks associated with the propagation of distortions across data levels and with the fact that default ROS 2 security mechanisms are disabled. The proposed typology can serve as a foundation for formulating data protection requirements for mobile robotic platforms. Existing security tools are examined (SROS2, network segmentation, restricting automatic DDS discovery, and data-at-rest encryption), along with their limitations: cryptography secures the transmission channel but not the physical measurements; therefore, mitigating sensor spoofing requires cross-verification of data and anomaly detection.