Affine Transformation-based Perfectly Undetectable False Data Injection Attacks on Remote Manipulator Kinematic Control with Attack Detector

TL;DR

The paper addresses the risk of perfectly undetectable false data injection attacks on remote robotic manipulators by modeling affine transformations on both observables and control commands. It derives conditions (Theorem 1) under which the attacked and nominal plant dynamics are equivalent, enabling undetectable manipulation of joint trajectories, and further shows (Theorem 2) that adaptive detectors can be evaded with a specific pair of affine scalings. Experimental validation on a $6$-DOF FANUC manipulator using Jacobian-transpose velocity control demonstrates three attack scenarios—scaling, reflection, and shear—that yield observed trajectories indistinguishable from nominal while altering the true motion. The results highlight a concrete vulnerability in standard linear joint-velocity control and motivate the development of defense strategies and extensions to nonlinear plants. Overall, the work provides a formal framework and empirical evidence for coordinated, perfectly undetectable FDIA in remote manipulator control, with implications for cyber-physical security in Industry 4.0.

Abstract

This paper demonstrates the viability of perfectly undetectable affine transformation attacks against robotic manipulators where intelligent attackers can inject multiplicative and additive false data while remaining completely hidden from system users. The attacker can implement these communication line attacks by satisfying three Conditions presented in this work. These claims are experimentally validated on a FANUC 6 degree of freedom manipulator by comparing a nominal (non-attacked) trial and a detectable attack case against three perfectly undetectable trajectory attack Scenarios: scaling, reflection, and shearing. The results show similar observed end effector error for the attack Scenarios and the nominal case, indicating that the perfectly undetectable affine transformation attack method keeps the attacker perfectly hidden while enabling them to attack manipulator trajectories.

Figures (8)

• Figure 1: Conceptual diagram of false data injection attack (FDIA) on remote manipulator: (a) Attacked networked control system with coordinated FDIA on the commands and observables. (b) Plant dynamics as perceived by the controller, indistinguishable from the nominal plant behavior and thus undetectable.
• Figure 2: FDIA on manipulator kinematic control: Attack detector and control scheme adopted from Zhang23.
• Figure 3: Nominal Trial Results
• Figure 4: Detectable Attack Results
• Figure 5: Scenario 1: Scaling Attack Results