Paper Review - BackDoor - Making microphones hear inaudible sounds

This is a paper review for: Roy, Nirupam, Haitham Hassanieh, and Romit Roy Choudhury. "Backdoor: Making microphones hear inaudible sounds." Proceedings of the 15th Annual International Conference on Mobile Systems, Applications, and Services. 2017.

Summary

The authors showed that regular microphones could record audio waves outside the human hearing range as well as the microphone recording range. They made use of non-linearities in microphone diaphragms and power amplifiers to achieve this. The authors played audio signals through ultra-sound speakers, and the nonlinear amplifier creates a “shadow” that is in the microphone’s recording range and, therefore, can be recorded. This shadow, although inaudible, but carries data that enables acoustic communication. The authors evaluated BackDoor on three metrics, which are human audibility, throughput, and jamming effectiveness. For audibility, the authors asked a group of 7 people sitting around a table to report the BackDoor sound loudness. All seven people confirmed that the sound is inaudible. BackDoor achieved double the throughput of the near-ultrasound band. For jamming effectiveness, the authors played 2000 English words to 7 humans who were able to interpret less than 15% of them correctly. The proposed solution opens the door to new applications like jamming spy microphones for privacy, acoustic watermarking, and inaudible data communication.

Things I liked

• Their solution does not require specific hardware. All current microphones can make use of this capability
• Their solution provides more spectrum for communication without the possibility of harming pets and infants (which is a concern in near ultrasound range).
• Provided even higher data rate than near ultrasound range(from 20KHz to 24KHz)
• Presented an elegant idea from the high conceptual level of taking advantage of something viewed as a con like microphones’ non-linearity to help in implementing a neat idea.
• Provided beneficial background information on microphone systems by explaining the recording system and the modules it requires and the components needed. They also presented a graphical pipeline that helps reveal the sound recording signal flow

Things I did not like:

• I am not sure the authors should have mentioned the vulnerability of jamming recording microphones with specific details; this can help robbers jam phone microphones and prevent their owners from calling an emergency number during an attack
• The authors tested only seven individuals to confirm that the sound is uninterpretable. I am not sure that this sample size is enough.

Further research

Can we reverse engineer recordings by identifying shadows and negating these shadows via audio editing software to interpret the record?