PaperlayerView exact revision on arXiv

Source-linked AI summary

GhostTac: Manipulating Tactile Sensors without Physical Contact

Kun Wang, Xuancun Lu, Ruochen Zhou, Kai Wang, Tongjun Ye, Yihao Shao, Chen Yan, Xiaoyu Ji, Wenyuan Xu

arXiv:2608.20817v1cs.CRcs.RO

TL;DR

Physical-layer security for tactile sensors remains little studied despite their central role in robotic perception and interaction. GhostTac addresses this gap with a contactless EMI attack that converts interference into persistent measurement deviations and enables fine-grained control. Across 15 sensors from 8 manufacturers, it achieved 100% success in tested manipulations and affected robotic grasping, slip detection, and material classification.

  • Problem

    The physical-layer security of tactile sensing remains unexplored despite tactile feedback’s central role in robotic perception, grasping, assembly, and safe interaction.

  • Method

    GhostTac uses carefully synchronized EMI to exploit sensing-array coupling, nonlinear rectification, and amplification, producing controllable DC offsets or denial of service without contact.

  • Results

    100% success was achieved across 15 tactile sensors from 8 manufacturers for manipulating force, location, interference width, and dynamic contact patterns, with impacts demonstrated in three robotic tasks.

  • Takeaways & Limitations

    GhostTac reveals a physical attack vector that can corrupt tactile feedback and induce harmful robotic behaviors, while the paper proposes hardware- and software-based countermeasures.

  • Takeaways & Limitations

    GhostTac cannot reliably place controllable interference at arbitrary locations, creating a scope boundary that pattern-based detection may exploit.

Abstract

from arXiv · show

Tactile sensors are integral to modern robotic systems, enabling robots to perceive and interact with the physical environment through tactile feedback. However, the physical-layer security of tactile sensors has received little attention. We present GhostTac, the first contactless attack, to the best of our knowledge, that manipulates tactile sensing through electromagnetic interference (EMI). GhostTac exploits nonlinear rectification and limited-bandwidth amplification, converting carefully crafted EMI signals into persistent DC offsets that bypass onboard filtering and induce stable measurement deviations. It enables fine-grained, controllable manipulation of sensor outputs by shaping the spatial distribution and magnitude of interference at targeted locations. Such manipulation can induce harmful robot behaviors, including excessive force that may damage objects or injure people. We evaluate GhostTac on 10 sensor modules and two dexterous hands, covering 15 tactile sensors of different types, and demonstrate consistent effectiveness across all tested devices. Three case studies involving tactile grasping, slip detection, and material classification further illustrate its practical impact on real robotic tasks. These findings reveal a new physical attack vector against tactile sensing in robotic systems.

1 Introduction

GhostTac exposes an unexplored physical-layer vulnerability in tactile sensing by using contactless EMI to manipulate measurements and induce denial of service. The attack enables controllable interference and produces harmful effects in robotic grasping and other tactile-driven tasks.

  • Tactile sensing supports force perception, material inference, stable grasping, precise assembly, and safe physical interaction, yet its physical-layer security remains unexplored.
  • GhostTac uses pre-designed EMI signals to manipulate tactile readings remotely or induce a DoS condition without physical contact.
  • The attack addresses signal injection by analyzing sensor architectures and locating EMI coupling points in sensing arrays and communication channels.
  • GhostTac synchronizes malicious signals with sequential tactile-array excitation to control where, when, and how strongly interference is injected.
  • 100% success was achieved across 15 tactile sensors from 8 manufacturers for manipulating force magnitude, location, interference width, and dynamic contact patterns.
  • The evaluation includes grasping, slip detection, and material-classification case studies demonstrating unintended and harmful robotic behaviors.

2 Background

Tactile sensors convert pressure changes into electrical signals through sensing arrays, amplification, digitization, and MCU processing. Resistive sensing and sequential scanning connect local pressure changes to measured voltage and contact location.

  • Commercial tactile sensors commonly use cover, RX-electrode, sensitive, TX-electrode, and support layers to transduce pressure distributions into electrical signals.
  • The sensor architecture contains a sensing array, amplifying circuits, an ADC, and an MCU, with an m×n grid formed by m RXs and n TXs.
  • The acquisition pipeline sequentially excites TXs, amplifies RX signals, digitizes voltages, and processes them into force distribution and per-unit tactile data.
  • Resistive Sensing: In resistive sensing, pressure decreases a unit’s resistance R11; applying excitation Vi1 produces amplifier output Vo1 whose magnitude increases as pressure rises.
  • Scan Driving Methods: Sequential scanning activates one TX at a time, allowing the system to identify contact location from the active TX-RX intersection.

3 Threat Model

GhostTac assumes a contactless attacker targets tactile-equipped devices to manipulate or deny sensor measurements. The model grants prior knowledge of an identical COTS device and proximity for covert EM emission, but excludes physical tampering.

  • The attacker seeks to manipulate or DoS tactile measurements contactlessly, causing grasping failures or material misclassification.
  • The victim is a tactile-equipped device requiring precise control, such as a robotic arm used for domestic services or patient care.
  • The attacker is assumed to obtain an identical COTS device and reverse engineer parameters needed for signal optimization.
  • The attacker can approach to emit EM signals or pre-place a malicious device nearby, but cannot physically touch or tamper with the victim device.

4 Principle of GhostTac Attack

GhostTac exploits EMI coupling and nonlinear signal processing in tactile sensors to create stable, frequency-dependent measurement deviations or communication disruption. Experiments validate DC-offset generation through rectification and amplification, including polarity changes with frequency.

  • The security analysis examines EMI coupling, measurement interference, denial of service, and physical validation in tactile sensors.The sensor architecture includes sensing, amplification, conversion, and control stages that provide multiple potential coupling targets.
  • EMI coupling and injection points: 701 MHz injection at the sensing array caused stable positive measurement interference, while 1380 MHz communication-line injection caused data loss and signal fluctuations.The sensing-array effect indicates direct electrode coupling, whereas communication-line interference produced sensor DoS.
  • Nonlinear rectification: High-frequency EMI is coupled into op-amp circuitry, where transistor nonlinearities rectify it into a DC component that can shift sensor measurements.The second-order term produces a DC component proportional to V_x^2, and the op-amp processes the resulting signal.
  • Frequency-dependent response: The frequency-dependent phase factor determines offset polarity, with different circuit regimes producing opposite responses.The cited analysis links polarity to cos(2ϕ(f)) and distinguishes resonant and capacitive behavior.
  • Nonlinear rectification: +0.13 V and -0.27 V DC offsets appeared under 266 MHz and 612 MHz DPI signals, respectively, validating frequency-dependent rectification.The measured offsets were superimposed on the high-frequency signal after extraneous AC components were filtered.
  • Bandwidth-limited amplification: Limited op-amp bandwidth attenuates high-frequency signals while amplifying the DC component, allowing the offset to interfere with sensor measurements.Output capacitance shunts more high-frequency current as frequency increases, lowering effective voltage gain.

5.1 Efficient Signal Injection

Efficient injection requires identifying susceptible sensor components and tuning EMI conditions to produce measurable effects. Experiments show frequency-specific offsets, communication loss, and persistent hardware failure modes.

  • Attack workflow: GhostTac’s workflow optimizes carrier frequency and power while shaping baseband signals to control static and dynamic tactile effects.The workflow combines these signals through modulation before electromagnetic transmission.
  • Frequency tuning: The op-amp’s limited bandwidth can suppress the AC component while preserving an amplified DC offset in sensor measurements.This filtering behavior explains why tuning the interference frequency can produce a measurable deviation without a prominent AC signal.
  • Frequency tuning: Specific EMI frequencies produced a positive deviation at 262 MHz and a negative deviation at 127 MHz, while the corresponding AC signal stayed below the noise floor.These observations were obtained in an inverting-amplifier test using a 2 V V_pp sinusoidal input.
  • Communication disruption: 1380 MHz interference caused persistent communication loss through corrupted TXD data, with recovery after a software reboot.The result indicates a protocol-level DoS mechanism rather than physical hardware damage in that sensor module.

5 GhostTac Design

GhostTac combines frequency-selective EMI coupling with amplitude-modulated, scan-synchronized interference to control tactile measurements in force, location, width, and time. Its signal design exploits sensor architecture and resonance to create predictable static and dynamic perturbations.

  • GhostTac jointly designs carrier and baseband signals to regulate perceived force, target locations, adjust affected width, and synthesize dynamic patterns.These control dimensions are composed for grasping, slip detection, and material-classification attacks.
  • Carrier design: Sensor wiring can couple strongly when carrier frequency matches electrical characteristics, with strong coupling typically occurring within c/50l≤f≤c/2l.The exact resonant frequency also depends on conductor geometry, material properties, and surroundings.
  • Carrier design: 1400 MHz produced the strongest positive interference, while 1.25–1.27 GHz produced the largest negative interference in a sensor whose baseline averaged 200 g.The carrier sweep covered 1.10–1.50 GHz in 10 MHz increments.
  • Carrier design: Different sensor resonances enable selective targeting, while multi-frequency signals can interfere with multiple sensors simultaneously.Sensor resonance varies with sensor type, placement, and array geometry.
  • Signal power: Weight deviation increased approximately proportionally with transmission power, and output above 2 dBm drove the tactile sensor into DoS.The power sweep used a 1443 MHz carrier from −15 dBm to 10 dBm.
  • Baseband design: AM was adopted because its induced sensor offset is approximately proportional to injected EMI amplitude, unlike the less predictable responses of FM and PM.The carrier targets the victim resonance, while the baseband regulates the sensing perturbation.
  • Controllable manipulation: Four baseband parameters—frequency, amplitude, temporal offset, and duration—support static force, location, and width control plus dynamic slipping effects.Static waveforms included sine, ramp, and pulse envelopes whose outputs closely tracked the injected waveforms.
  • Controllable manipulation: Synchronizing interference with excitation timing allows GhostTac to inject false measurements into specific sensing positions.The attacker monitors excitation leakage and varies latency relative to target TX events.

6 Evaluation

GhostTac was evaluated across commercial tactile sensors, dexterous hands, attack geometries, and three robotic tasks. The experiments found broad manipulation capability, distance- and power-dependent attack performance, and harmful task-level effects including grasp failure and slip misclassification.

  • Overall evaluation: GhostTac was evaluated on 15 commercial tactile sensors from 8 manufacturers, including 10 sensor modules and 5 sensors from 2 dexterous hands.The evaluation used commercial off-the-shelf devices and summarized overall performance in Table 1.
  • Overall evaluation: All tested sensors exhibited EMI vulnerability, while 13 of 15 were susceptible to DoS attacks.The evaluation swept 700 MHz–1.6 GHz and recorded measurement deviation, DoS success, and manipulation types.
  • Overall evaluation: Static and dynamic effects were achieved across all tested sensor arrays, excluding single-unit sensors where fine-grained spatial manipulation is inapplicable.A 1325 MHz EMI signal also triggered a DoS state in one evaluation.
  • Distance and angle: Manipulation deviation decreased monotonically with distance across all tested power levels, while higher power extended the effective attack range.DoS success likewise decreased with distance and lower power, but remained achievable up to 40 cm at the highest power.
  • Distance and angle: Force manipulation succeeded within a ±60° vertical angular window, and horizontal injection produced a 310 g deviation even at 90°.These results indicate that precise aiming was not required in the tested setup.
  • Long-range attack: Under line-of-sight conditions, a high-gain polarized log-periodic antenna enabled bottle-squeezing attacks up to 3 m with 426 g deviation per sensing unit.Total deviation scaled with the number of contact units.
  • Grasping: In grasping, negative interference caused excessive force and object damage, whereas positive interference reduced grip force and caused object drops.The attacks were conducted along the grasping path using tactile feedback control.
  • Slip detection: Slip attacks reliably suppressed real-slip detection or generated false slips, with both scenarios repeated 10 times consistently.The detector was driven between stable-grip and slipping classifications by injected attack patterns.

7 Discussion

The discussion outlines hardware and software defenses against GhostTac while identifying spatial, sensor-type, alignment, and system-level boundaries on its effectiveness.

  • Countermeasures: Countermeasures target electromagnetic robustness and tactile-sensing resilience through hardware and software defenses.Proposed directions include shielding, differential comparators, filtering, excitation changes, waveform randomization, and scan-timing randomization.
  • Limitations and Future Work: GhostTac cannot reliably place controllable interference at arbitrary locations because of driving schemes and wide electromagnetic coupling between receivers.Its controllable interference is restricted along transmitting electrodes, although the available location and intensity range remains sufficient for effective attacks.
  • Limitations and Future Work: The evaluation primarily covers piezoresistive and capacitive tactile sensors, leaving visuotactile and other multimodal sensors for future investigation.These tested sensor types dominate the current market.
  • Limitations and Future Work: Misalignment weakens electromagnetic coupling and degrades attack effects, although GhostTac remains effective within a susceptible spatial region around the target.Moving-grasping attacks can still induce observable weight deviations while the target remains inside that region.
  • System-Level Analysis: Sensor fusion may reduce attack effectiveness, but tasks dominated by tactile feedback remain vulnerable.Material recognition and manipulation in camera-blind regions are cited as examples where tactile measurements can dominate decisions.
  • System-Level Analysis: GhostTac can continuously mislead tactile-driven closed-loop control, while simple filtering is insufficient against carefully crafted continuous injections.Suggested safeguards include trusted contact verification, active physical consistency probing, and runtime monitoring of suspicious temporal patterns.
  • System-Level Analysis: Amplitude- or continuity-based anomaly detection may be unreliable because forged readings can remain within plausible ranges.Pattern-based detection could instead test whether activation patterns are physically consistent with natural contact geometry, including GhostTac’s TX-aligned constraint.

8 Related Work

Prior EMI attacks target many sensor modalities, while GhostTac extends this attack surface to tactile sensing with fine-grained, dynamic manipulation tailored to dense force measurements.

  • Related Work: EMI attacks have been reported against microphones, touchscreens, temperature sensors, LiDAR, and image sensors, with effects ranging from denial of service to signal manipulation.Prior work also affects PWM driving signals and can transmit electromagnetic signals indirectly through other components.
  • Novelty over prior work: Tactile sensors pose distinct challenges because they measure minute force variations at dense locations for complex robotic tasks.Unlike threshold-based touchscreens, tactile sensing requires manipulating force magnitude, polarity, location, and sensor selection.
  • Novelty over prior work: GhostTac introduces a parameterized modulation framework and dynamic adversarial patterns for tactile manipulation, including attacks on slip detection and material classification.The paper also analyzes nonlinear rectification and limited-bandwidth amplification through circuit-level analysis and physical validation.

9 Conclusion

GhostTac is presented as a contactless EMI attack that corrupts tactile measurements through coupling and nonlinear rectification, with validated effects across commercial sensors and dexterous hands.

  • Conclusion: GhostTac is the first contactless attack presented to manipulate tactile sensor measurements through EMI.It controls force magnitude and location or induces denial of service without requiring physical contact or a stationary target.
  • Conclusion: Coupling vulnerabilities and nonlinear rectification convert high-frequency EMI into DC offsets that corrupt force measurements.The underlying mechanism is analyzed theoretically and experimentally.
  • Conclusion: GhostTac is validated on 10 commercial-off-the-shelf sensor modules and two dexterous hands.The paper additionally proposes hardware- and software-based countermeasures that significantly reduce risk.

Ethical Considerations

The paper describes strict isolation and personnel-safety measures for investigating hardware vulnerabilities in a controlled environment.

  • Ethical Considerations: The experiments used isolation, supervision, logging, and personal electromagnetic shielding suits to protect personnel.These measures were intended to support controlled investigation without ignoring safety and ethical responsibility.

Open Science

The authors make research artifacts accessible to strengthen transparency and reproducibility, while acknowledging that reproducing the attacks requires specific hardware equipment.

  • The research artifacts are accessible through the GhostTac_CCS repository to support transparency and reproducibility.
  • Concrete implementation of the proposed attacks requires specific hardware equipment because the primary contribution concerns hardware vulnerabilities.
  • The authors provide additional experiment demonstrations intended to assist verification without the physical setup.
Loading 2608.20817v1…