Magnetic tags trick robot touch sensors into dropping or crushing objects

Researchers demonstrated a new attack, dubbed PhantomTac, that uses tiny, battery-free magnetic tags to manipulate the readings of robotic touch sensors. The tags, only 0.5 mm thick and 1.44 cm² in area, can be hidden under a barcode or QR-code sticker. When a robot’s magnetic tactile sensor detects the tag’s field, it misinterprets the signal as a change in contact force, causing the robot to release or over-grip an item.

The study, authored by Binbin Zhang, Yifeng Zhang, Xiaomeng Chen, and Dongyao Chen, was posted on October 19 and accepted for the ACM Conference on Computer and Communications Security (CCS) 2026, scheduled for November 15–19 in The Hague. The paper details how the tags exploit a limitation in magnetic tactile sensing: a nearby magnet can create a localized field that the sensor interprets as pressure.

How the attack works

Magnetic tactile sensors measure normal and shear force by tracking changes in a magnetic field as their sensing layer bends or compresses. PhantomTac places a small magnet close to the sensor surface. The resulting field is not fully cancelled by the sensor’s internal shielding, so the control system receives a false pressure reading. If the sensor reports higher pressure than the robot is actually applying, the robot may move before it has a firm grip and drop the item. If the sensor reports lower pressure, the robot may keep squeezing until the object is damaged.

The researchers tested several PaXini sensor models and a XELA uSPa 46 sensor, but only PaXini hardware was used in robotic demonstrations. A Dofbot-Pi4B robotic arm equipped with a magnetic touch sensor was used to show that the tags could make its gripper drop or crush an object. In grasping tests, the attack succeeded more than 95 % of the time when the contact position was within 8 mm of the intended tag placement.

Real-world implications

The attack requires only brief access to an object or its packaging and knowledge of the target sensor model. No software or control-system access is needed. Any robot that relies on magnetic tactile sensing for grip control—industrial pick-and-place arms, humanoid robots, or automated warehouses—could be vulnerable if its sensors are not shielded against external magnetic fields.

Because the tags are passive and battery-free, they can be deployed covertly. A malicious actor could attach a tag to a high-value component, then have a robot drop or crush it during handling. The damage could range from lost inventory to costly equipment failure.

What to verify

  1. Sensor model – Identify whether your robots use PaXini, XELA, or other magnetic tactile sensors.
  2. Shielding – Check if the sensor firmware or hardware includes magnetic shielding or filtering for external fields.
  3. Packaging – Inspect barcodes, QR codes, or stickers for hidden magnetic elements.
  4. Control logic – Verify that grip force thresholds are not solely based on raw sensor output; consider adding hysteresis or cross-checking with other sensors.

Immediate protective action

  • Shield sensors – Install magnetic shielding or redesign the sensor housing to block external fields.
  • Validate packaging – Run a magnetic field scan on all incoming items before they enter the robot’s workspace.
  • Update firmware – Implement firmware checks that flag sudden, large changes in pressure that could indicate a magnetic disturbance.
  • Audit supply chain – Ensure that suppliers cannot tamper with packaging or add magnetic tags.

Further reading

The PhantomTac paper also discusses how false shear-force readings can weaken a robot’s grip, potentially causing objects to slip. For a related example of magnetic interference in security devices, see the SPECTER USB Leak Shows Air-Gapped PCs Can Exfiltrate Secrets.

PhantomTac highlights a new vector for physical-layer attacks on automation. As robotics adoption grows, manufacturers must consider magnetic interference as a legitimate threat vector and incorporate mitigation into both hardware and software designs.

Explore more on this topic