GhostTac: Contactless Manipulation of Robot Tactile Sensors via Electromagnetic Interference

GhostTac proposes a contactless attack that exploits the nonlinear rectification and limited-bandwidth amplification effects of electromagnetic interference (EMI) to convert carefully crafted EMI signals into a persistent DC offset, bypassing on-board filtering and causing stable measurement deviations. The authors verified the attack's consistency across 15 tactile sensors of different types on 10 sensor modules and 2 dexterous hands, and demonstrated real-world impact through three case studies: tactile grasping, slip detection, and material classification. For example, it could cause a domestic robot to exert excessive force. The work has been accepted to ACM CCS 2026.