International Association for Cryptologic Research

International Association
for Cryptologic Research

Transactions on Cryptographic Hardware and Embedded Systems, Volume 2024

Laser-Based Command Injection Attacks on Voice-Controlled Microphone Arrays


README

Silent-Attack_LCMA (Laser Commands for Microphone Array)

This repository contains some supported files and demo videos associated with the paper Laser-Based Command Injection Attacks on Voice-Controlled Microphone Arrays (CHES 2024). Due to confidentiality requirements from our partner companies, we have only disclosed a portion of the LCMA's attack effects in this repository.

Description of the Attack

The LCMA attack marks a breakthrough in light-based interference with voice-controlled devices, surpassing former methods (LightCommands) limited to single-microphone targets. This attack manipulates devices equipped with microphone arrays and sensor fusion by synchronously targeting all apertures with the PWM (Pulse Width Modulation) algorithm. Implemented on a single MCU (Microcontroller Unit), it precisely controls multiple lasers to emit modulated light signals, mimicking audio commands.

The LCMA attack streamlines the complex process of compromising voice-controlled systems by utilizing BLE (Bluetooth Low Energy) for remote configuration. Pre-prepared audio (in Testing Audio/ folder) is transmitted to LCMA's control module. This module then digitizes the audio and outputs a PWM (Pulse Width Modulation) control signal to drive the laser transmitters, which emit the modulated light signals.
This procedure has been validated across various types of microphone arrays and voice-controlled devices to ensure its effectiveness and versatility. The audio received by the microphones (in Results/ folder), along with successful demonstration videos (see Demo Videos for details). Additionally, we've compiled a list of commands that LCMA can inject and some practical tips for deploying the attack (in the Tips/ folder).

Test Audio Dataset

The folder contains a comprehensive set of audio files used for injecting into different Voice-Controlled systems, such as Amazon Alex, Apple Siri, Google, Tesla, Car-Mounted Electric Kettle XiaoLi, Tmall, Vendor-A XiaoYi, Weltmeister and Xiaomi XiaoAi.

Results Audio Dataset

This folder contains part output audio files from different types of microphones and devices subjected to LCMA laser injection.

Injection Tips

This folder contains some helpful tips for effectively utilizing LCMA in various scenarios and applications.

Demo Videos

Here is our part collection of demonstration videos showcasing LCMA's practical applications and real-world functionality. You can access them at https://youtube.com/playlist?list=PLcPx5A14HPkpxNluy0s-akC_hrHMw0Rz0&si=f3ES1a2e5UaZvtrn.

Contact

Our team can be contacted at zhugejw@tsinghua.edu.cn.

Licensing

This work is licensed under a Creative Commons Attribution 4.0 International License.

Creative Commons License