Precision-Varying Prediction (PVP): Robustifying ASR systems against adversarial attacks

2026-03-23

AI Summary

This work proposes a training-free, unified defense strategy against adversarial attacks on automatic speech recognition (ASR) systems, which are known to be highly vulnerable and lack robustness. The approach leverages the sensitivity discrepancy of adversarial examples across model precisions during inference by randomly sampling precision levels and constructing a detection mechanism based on a Gaussian classifier. For the first time, this method reveals and exploits the distinctive response characteristics of adversarial samples under variable-precision inference. Extensive evaluations demonstrate that the proposed technique significantly enhances system robustness across diverse ASR architectures and attack types while enabling efficient detection of adversarial inputs.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Computer Vision: Adversarial Attacks & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

With the increasing deployment of automated and agentic systems, ensuring the adversarial robustness of automatic speech recognition (ASR) models has become critical. We observe that changing the precision of an ASR model during inference reduces the likelihood of adversarial attacks succeeding. We take advantage of this fact to make the models more robust by simple random sampling of the precision during prediction. Moreover, the insight can be turned into an adversarial example detection strategy by comparing outputs resulting from different precisions and leveraging a simple Gaussian classifier. An experimental analysis demonstrates a significant increase in robustness and competitive detection performance for various ASR models and attack types.

Problem

• adversarial robustness
• automatic speech recognition
• adversarial attacks
• ASR systems

Innovation

• Precision-Varying Prediction
• adversarial robustness
• automatic speech recognition
• adversarial example detection
• numerical precision