Abstract
Audio event detection (AED) systems can leverage the power of specialized algorithms for detecting the presence of a specific sound of interest within audio captured from the environment. More recent approaches rely on deep learning algorithms, such as convolutional neural networks and convolutional recurrent neural networks. Given these conditions, it is important to assess how vulnerable these systems can be to attacks. As such, we develop AED-suited convolutional neural networks and convolutional recurrent neural networks, and attack them next with white noise disturbances, conceived to be simple and straightforward to be implemented and employed, even by non-tech savvy attackers. We develop this work under a safety-oriented scenario (AED systems for safety-related sounds, such as gunshots), and we show that an attacker can use such disturbances to avoid detection by up to 100 percent success. Prior work has shown that attackers can mislead image classification tasks; however, this work focuses on attacks against AED systems by tampering with their audio rather than image components. This work brings awareness to the designers and manufacturers of AED systems, as these solutions are vulnerable, yet may be trusted by individuals and families.
Highlights
Several studies have shown that unwanted noise can have a detrimental effect on classifier performance [27,28]. We focus on this niche, and as such, study how to attack deep learning Audio event detection (AED) systems, focusing on employing simple, accessible and easy-to-reproduce disturbances made of white noise, to be used as a means of disrupting the classifiers
We present the results obtained from the several experiments that show the robustness of convolutional neural networks (CNN) and convolutional recurrent neural networks (CRNN) classifiers against the white noise attacks
We tested CNN and CRNN algorithms for AED, and while their detection performance was reasonable under ideal circumstances, a sharp drop in it was seen, even when little white noise was injected into the test audio samples
Summary
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. IoT-based cyber–physical systems (CPSs) have been developed to advance personalized health care, emergency response, traffic flow management, and electric power generation and delivery [1,2,3,4,5,6,7]. IoT-based CPSs include smart networked systems with embedded sensors, processors and actuators that sense and interact with the physical world. A core element of IoT-based CPSs is decision making, which analyzes the data obtained from sensors. As more sensors became pervasive, generating huge volumes of data, deep learning algorithms, in particular, more recent neural networks, are becoming pervasive, playing a vital role in the good performance of decision-making in real-life
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
