ConvXformer: Differentially Private Hybrid ConvNeXt-Transformer for Inertial Navigation

2025-10-22

AI Summary

To address privacy leakage risks and the accuracy–privacy trade-off in deep learning–based inertial navigation under GPS-denied environments, this paper proposes a hierarchical hybrid architecture satisfying (ε,δ)-differential privacy. The architecture synergistically integrates ConvNeXt for local feature modeling and Transformer for global dependency capture. It further introduces three novel mechanisms: adaptive gradient clipping, gradient-alignment-based noise injection, and truncated singular value decomposition (SVD)-driven low-rank gradient compression. These components jointly enable noise-aware privacy protection on high-frequency inertial sequences, effectively mitigating performance degradation caused by excessive noise in conventional differential privacy approaches. Extensive experiments on OxIOD, RIDI, RoNIN, and our custom Mech-IO dataset—characterized by strong magnetic interference—demonstrate an average localization accuracy improvement exceeding 40%, alongside enhanced robustness and strengthened privacy guarantees.

Technology Category

• Machine Learning: Privacy
• Natural Language Processing: Safety and Robustness
• Computer Vision: Bias, Fairness & Privacy

Application Category

• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Data-driven inertial sequence learning has revolutionized navigation in GPS-denied environments, offering superior odometric resolution compared to traditional Bayesian methods. However, deep learning-based inertial tracking systems remain vulnerable to privacy breaches that can expose sensitive training data. hl{Existing differential privacy solutions often compromise model performance by introducing excessive noise, particularly in high-frequency inertial measurements.} In this article, we propose ConvXformer, a hybrid architecture that fuses ConvNeXt blocks with Transformer encoders in a hierarchical structure for robust inertial navigation. We propose an efficient differential privacy mechanism incorporating adaptive gradient clipping and gradient-aligned noise injection (GANI) to protect sensitive information while ensuring model performance. Our framework leverages truncated singular value decomposition for gradient processing, enabling precise control over the privacy-utility trade-off. Comprehensive performance evaluations on benchmark datasets (OxIOD, RIDI, RoNIN) demonstrate that ConvXformer surpasses state-of-the-art methods, achieving more than 40% improvement in positioning accuracy while ensuring $(ε,δ)$-differential privacy guarantees. To validate real-world performance, we introduce the Mech-IO dataset, collected from the mechanical engineering building at KAIST, where intense magnetic fields from industrial equipment induce significant sensor perturbations. This demonstrated robustness under severe environmental distortions makes our framework well-suited for secure and intelligent navigation in cyber-physical systems.

Problem

• Protecting sensitive training data in deep learning-based inertial navigation systems
• Reducing performance degradation from excessive noise in differential privacy solutions
• Maintaining positioning accuracy under severe environmental distortions and sensor perturbations

Innovation

• Hybrid ConvNeXt-Transformer architecture for navigation
• Adaptive gradient clipping with noise injection mechanism
• Truncated singular value decomposition for privacy control