Abstract
The area of Machine Learning as a Service (MLaaS) is experiencing increasedimplementation due to recent advancements in the AI (Artificial Intelligence)industry. However, this spike has prompted concerns regarding AI defensemechanisms, specifically regarding potential covert attacks from third-partyproviders that cannot be entirely trusted. Recent research has uncovered thatauditory backdoors may use certain modifications as their initiating mechanism.DynamicTrigger is introduced as a methodology for carrying out dynamic backdoorattacks that use cleverly designed tweaks to ensure that corrupted samples areindistinguishable from clean. By utilizing fluctuating signal sampling ratesand masking speaker identities through dynamic sound triggers (such as theclapping of hands), it is possible to deceive speech recognition systems (ASR).Our empirical testing demonstrates that DynamicTrigger is both potent andstealthy, achieving impressive success rates during covert attacks whilemaintaining exceptional accuracy with non-poisoned datasets.