Adversary-Aware Private Inference over Wireless Channels

2025-10-23

AI Summary

To address the privacy risk in wireless edge AI sensing—where transmitted features may enable sensitive data reconstruction—this paper proposes the first fine-grained privacy-preserving framework tailored for single-feature streams. Unlike conventional differential privacy, which suffers from limited applicability in continuous, dynamic feature settings, our framework jointly integrates learnable feature transformation and adversarial-aware perturbation, enabling channel-adaptive feature distortion between non-colocated sensors and model servers. It simultaneously preserves downstream inference accuracy and thwarts reconstruction attacks. Experimental evaluations across diverse wireless sensing tasks—including gesture recognition and respiration monitoring—demonstrate strong privacy protection, reducing reconstruction PSNR by over 12 dB, while retaining more than 95% of the original model’s accuracy. The approach significantly outperforms existing baseline methods.

Technology Category

• Machine Learning: Privacy
• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Security and Privacy: Privacy-enhancing technologies
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

AI-based sensing at wireless edge devices has the potential to significantly enhance Artificial Intelligence (AI) applications, particularly for vision and perception tasks such as in autonomous driving and environmental monitoring. AI systems rely both on efficient model learning and inference. In the inference phase, features extracted from sensing data are utilized for prediction tasks (e.g., classification or regression). In edge networks, sensors and model servers are often not co-located, which requires communication of features. As sensitive personal data can be reconstructed by an adversary, transformation of the features are required to reduce the risk of privacy violations. While differential privacy mechanisms provide a means of protecting finite datasets, protection of individual features has not been addressed. In this paper, we propose a novel framework for privacy-preserving AI-based sensing, where devices apply transformations of extracted features before transmission to a model server.

Problem

• Protecting feature privacy in wireless edge AI sensing
• Preventing adversary reconstruction of sensitive personal data
• Securing feature transmission between sensors and model servers

Innovation

• Adversary-aware private inference over wireless channels
• Feature transformation before transmission to model server
• Protection of individual features against reconstruction attacks