Can You Hear It? Backdoor Attacks via Ultrasonic Triggers
This poses a security threat for users of speech recognition systems by making attacks stealthy and hard to detect, representing an incremental advance in backdoor attack methods.
The paper tackles backdoor attacks on automatic speech recognition systems using inaudible ultrasonic triggers, achieving a 100% attack success rate with less than 1% poisoned data and demonstrating real-world effectiveness on Android hardware.
This work explores backdoor attacks for automatic speech recognition systems where we inject inaudible triggers. By doing so, we make the backdoor attack challenging to detect for legitimate users, and thus, potentially more dangerous. We conduct experiments on two versions of a speech dataset and three neural networks and explore the performance of our attack concerning the duration, position, and type of the trigger. Our results indicate that less than 1% of poisoned data is sufficient to deploy a backdoor attack and reach a 100% attack success rate. We observed that short, non-continuous triggers result in highly successful attacks. However, since our trigger is inaudible, it can be as long as possible without raising any suspicions making the attack more effective. Finally, we conducted our attack in actual hardware and saw that an adversary could manipulate inference in an Android application by playing the inaudible trigger over the air.