Detecting acoustic backdoor transmission of inaudible messages using deep learning

Silvija Kokalj-Filipovic, Morriel Kasher, Michael Zhao, Predrag Spasojevic

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The novel secret inaudible acoustic communication channel [11], referred to as the BackDoor channel, is a method of embedding inaudible signals in acoustic data that is likely to be processed by a trained deep neural net. In this paper we perform preliminary studies of the detectability of such a communication channel by deep learning algorithms that are trained on the original acoustic data used for such a secret exploit. The BackDoor channel embeds inaudible messages by modulating them with a sinewave of 40kHz and transmitting using ultrasonic speakers. The received composite signal is used to generate the Backdoor dataset for evaluation of our neural net. The audible samples are played back and recorded as a baseline dataset for training. The Backdoor dataset is used to evaluate the impact that the BackDoor channel has on the classification of the acoustic data, and we show that the accuracy of the classifier is degraded. The degradation depends on the type of deep classifier and it appears to impact less the classifiers that are trained using autoencoders. We also propose statistics that can be used to detect the out-of-distribution samples created as a result of the BackDoor channel, such as the log likelihood of the variational autoencoder used to pre-train the classifier or the empirical entropy of the classifier's output layer. The preliminary results presented in this paper indicate that the use of deep learning classifiers as detectors of the BackDoor secret channel merits further research.

Original languageEnglish (US)
Title of host publicationWiseML 2020 - Proceedings of the 2nd ACM Workshop on Wireless Security and Machine Learning
PublisherAssociation for Computing Machinery
Pages80-85
Number of pages6
ISBN (Electronic)9781450380072
DOIs
StatePublished - Jul 13 2020
Event2nd ACM Workshop on Wireless Security and Machine Learning, WiseML 2020 - Linz, Virtual, Austria
Duration: Jul 13 2020 → …

Publication series

NameWiseML 2020 - Proceedings of the 2nd ACM Workshop on Wireless Security and Machine Learning

Conference

Conference2nd ACM Workshop on Wireless Security and Machine Learning, WiseML 2020
Country/TerritoryAustria
CityLinz, Virtual
Period7/13/20 → …

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications
  • Software
  • Artificial Intelligence

Keywords

  • BackDoor channel
  • deep learning
  • inaudible voice commands
  • neural networks
  • ultrasonic acoustics
  • ultrasound injection

Fingerprint

Dive into the research topics of 'Detecting acoustic backdoor transmission of inaudible messages using deep learning'. Together they form a unique fingerprint.

Cite this