LightTouch: Harnessing Laser-Based Signal Injection to Manipulate Optical Human-Computer Interfaces

This paper studies the capability of light-based signal injection attacks to remotely inject fake inputs to modern optical sensor-based input devices. We demonstrate how an attacker can successfully inject malicious Character User Interface (CUI) commands on different laser-projection keyboards and generate fake mouse-cursor movements in bending mouses, using invisible to human eyes, infrared lasers. Our proof-of-concept evaluation shows how the attack achieves 100% success rate on injecting basic keyboard operations up to 7 meters away from the victim input device and through glass windows, without tampering with the victim computer system, or needing a network connection to pursue the attack. This vulnerability allows the attacker not only to type unauthorized commands but also to prevent the unlocked victim PC or workstation from automatically going to sleep mode, thereby extending the time window for locally committing malicious activities, i.e., lunchtime attacks. Within a wide viewing angle of 30°–45° from the victim input device location, the attacker can continuously inject false movements and press at least 24 consecutive keys without any failure up to 5 meters away. We also verify the attack feasibility in realistic office environments where the laser beam is partially occluded. Our analysis shows the potential security risks of invisible light injection attacks, including providing preventive defense measures to limit exposure of optical input interfaces to such a threat. This work aims to help manufacturers address the vulnerability and build reliable Human-Computer Interfaces.