Sec5GLoc: Securing 5G Indoor Localization via Adversary-Resilient Deep Learning Architecture

TL;DR

Sec5GLoc addresses secure indoor localization in 5G by integrating CIR fingerprinting with physics-based cues and an adaptive attention mechanism to resist spoofing and adversarial perturbations. The method is framed as a robust min-max problem, formalized as $\min_{f} \max_{\Delta} \left\| f(\Delta(\mathbf{H})) - \mathbf{L} \right\|$, and solved with a CNN-based CIR extractor, multi-head attention fusion, and anchor/TDoA features to enforce geometric consistency. Empirical results on a public 5G indoor dataset show sub-meter accuracy (mean errors around $0.58$ m in mixed LOS/NLOS and $0.34$ m in NLOS) and clear robustness to simulated attacks, outperforming classic baselines. The work demonstrates practical, real-time localization with built-in resilience and provides a foundation for secure RF localization in 5G and beyond.

Abstract

Emerging 5G millimeter-wave and sub-6 GHz networks enable high-accuracy indoor localization, but security and privacy vulnerabilities pose serious challenges. In this paper, we identify and address threats including location spoofing and adversarial signal manipulation against 5G-based indoor localization. We formalize a threat model encompassing attackers who inject forged radio signals or perturb channel measurements to mislead the localization system. To defend against these threats, we propose an adversary-resilient localization architecture that combines deep learning fingerprinting with physical domain knowledge. Our approach integrates multi-anchor Channel Impulse Response (CIR) fingerprints with Time Difference of Arrival (TDoA) features and known anchor positions in a hybrid Convolutional Neural Network (CNN) and multi-head attention network. This design inherently checks geometric consistency and dynamically down-weights anomalous signals, making localization robust to tampering. We formulate the secure localization problem and demonstrate, through extensive experiments on a public 5G indoor dataset, that the proposed system achieves a mean error approximately 0.58 m under mixed Line-of-Sight (LOS) and Non-Line-of-Sight (NLOS) trajectories in benign conditions and gracefully degrades to around 0.81 m under attack scenarios. We also show via ablation studies that each architecture component (attention mechanism, TDoA, etc.) is critical for both accuracy and resilience, reducing errors by 4-5 times compared to baselines. In addition, our system runs in real-time, localizing the user in just 1 ms on a simple CPU. The code has been released to ensure reproducibility (https://github.com/sec5gloc/Sec5GLoc).

Figures (5)

• Figure 1: Problem formulation of secure and robust 5G localization under adversarial signal perturbations and physical-layer threats.
• Figure 2: Sec5GLoc model: CIR features are extracted using shared CNN; anchor positions and TDoA features are embedded and fused via attention, then passed to a regression head to predict the location.
• Figure 3: Scatter plot of the predictions versus ground truth positions in the test environment under mixed LOS and NLOS conditions.
• Figure 4: Heatmap of average localization errors for Sec5GLoc in the test environment under mixed LOS and NLOS conditions, with a spoofing attack targeting the most influential anchor.
• Figure 5: CDFs of localization errors on the mixed LOS and NLOS test trajectory for Sec5GLoc with Gaussian noise added to the most influential anchor.