[Paper Review] Light Commands: Laser-Based Audio Injection Attacks on Voice-Controllable Systems

Takeshi Sugawara, Benjamin Cyr+3 more|arXiv (Cornell University)|Jun 21, 2020

Digital Media Forensic Detection68 citations

TL;DR

The paper demonstrates that MEMS microphones can be covertly injected with audio via amplitude-modulated laser light, enabling remote command injection on popular voice-controllable systems up to 110 meters and across buildings, and discusses defenses.

ABSTRACT

We propose a new class of signal injection attacks on microphones by physically converting light to sound. We show how an attacker can inject arbitrary audio signals to a target microphone by aiming an amplitude-modulated light at the microphone's aperture. We then proceed to show how this effect leads to a remote voice-command injection attack on voice-controllable systems. Examining various products that use Amazon's Alexa, Apple's Siri, Facebook's Portal, and Google Assistant, we show how to use light to obtain control over these devices at distances up to 110 meters and from two separate buildings. Next, we show that user authentication on these devices is often lacking, allowing the attacker to use light-injected voice commands to unlock the target's smartlock-protected front doors, open garage doors, shop on e-commerce websites at the target's expense, or even unlock and start various vehicles connected to the target's Google account (e.g., Tesla and Ford). Finally, we conclude with possible software and hardware defenses against our attacks.

Motivation & Objective

• Show that MEMS microphones respond to light as sound and can be driven by amplitude-modulated lasers.
• Demonstrate remote injection of voice commands into popular VC systems at long distances.
• Assess security implications of light-based command injections on VC devices and linked smart systems.
• Propose software and hardware defenses against light-based signal injection attacks.

Proposed method

• Use amplitude modulation (AM) to encode audio into laser intensity and inject into MEMS microphone via its acoustic port.
• Characterize laser-to-light conversion and microphone frequency response across blue and red lasers (450 nm and 638 nm).
• Identify physical transduction mechanisms (photoelectric and photoacoustic) responsible for light-to-sound conversion in MEMS microphones.
• Test command injection on multiple VC devices (Alexa, Siri, Portal, Google Assistant) and measure success across varying laser power and distances up to 110 m.
• Evaluate attack feasibility with minimal hardware and discuss attacker’s cost and setup.
• Discuss countermeasures and potential defenses.

Experimental results

Research questions

• RQ1 Can commands be remotely and stealthily injected into a VC system from large distances?
• RQ2 What laser power and aiming conditions are required to inject commands into popular VC devices?
• RQ3 What are the security implications for devices and accounts linked to VC systems when exposed to light-based injections?
• RQ4 What hardware/software defenses can mitigate light-based signal injection attacks?

Key findings

• A 5 mW laser can control many VC devices; ~60 mW can affect phones and tablets.
• The attack demonstrated command injections at distances up to 110 meters and through closed glass windows.
• VC devices often lack proper user authentication or have flawed implementations enabling unauthorized commands.
• A cheap setup using commercially available lasers and drivers suffices for attacks under favorable conditions.
• Injection exploits likely involve both photoelectric effects in the ASIC and photoacoustic effects in the diaphragm; illumination through the acoustic port is effective.
• Possible countermeasures include hardware shielding, authentication enhancements, and monitoring of anomalous light-induced outputs.