FedWiLoc: Federated Learning for Privacy-Preserving WiFi Indoor Localization

TL;DR

FedWiLoc tackles privacy-preserving indoor localization by combining a split encoder architecture executed at APs with on-device decoders, and a federated training regime for environment-robust encoder learning. A geometric loss couplet enforces cross-AP angular consistency, enabling generalization across spaces and bandwidths without environment-specific retraining. Empirical results across six environments show substantial improvements in median localization error and strong generalization, while maintaining strong privacy guarantees during both training and inference. The approach advances practical deployment of private Wi-Fi localization for IoT, robotics, and service applications.

Abstract

Current data-driven Wi-Fi-based indoor localization systems face three critical challenges: protecting user privacy, achieving accurate predictions in dynamic multipath environments, and generalizing across different deployments. Traditional Wi-Fi localization systems often compromise user privacy, particularly when facing compromised access points (APs) or man-in-the-middle attacks. As IoT devices proliferate in indoor environments, developing solutions that deliver accurate localization while robustly protecting privacy has become imperative. We introduce FedWiLoc, a privacy-preserving indoor localization system that addresses these challenges through three key innovations. First, FedWiLoc employs a split architecture where APs process Channel State Information (CSI) locally and transmit only privacy-preserving embedding vectors to user devices, preventing raw CSI exposure. Second, during training, FedWiLoc uses federated learning to collaboratively train the model across APs without centralizing sensitive user data. Third, we introduce a geometric loss function that jointly optimizes angle-of-arrival predictions and location estimates, enforcing geometric consistency to improve accuracy in challenging multipath conditions. Extensive evaluation across six diverse indoor environments spanning over 2,000 sq. ft. demonstrates that FedWiLoc outperforms state-of-the-art methods by up to 61.9% in median localization error while maintaining strong privacy guarantees throughout both training and inference.

Figures (10)

• Figure 1: Current systems send the user's measurement across different Wi-Fi Access Points(APs) to a central server to computes the user location. In contrast in FedWiLoc has a Federated learning architecture that enables the user device to decode its location like GPS.
• Figure 2: Model Architecture: Shows the encoder-decoder architecture of FedWiLoc. We use Resnet-34 modules for the encoder resnet-pt and the decoder receives the $N_{AP}$ embeddings and predicts $\{\theta_1,\theta_2,\cdots,\theta_{N_{AP}}\}$ that are then used along with $\{\mathbf{p}_1,\mathbf{p}_2,\cdots,\mathbf{p}_{N_{AP}}\}$ to perform triangulation. The triangulation loss is combined till the shared decoder but splits to propagate appropriate gradients through each AP's Encoder. The $\sin(\theta)$ and $\cos(\theta)$ losses are independent for each AP.
• Figure 3: Federated Learning
• Figure 4: FedWiLoc Results: FedWiLoc outperforms state-of-the-art models (DLoc, Spotfi) in localization in multiple environments of varying sizes
• Figure 5: AoA errors across various environments: Histograms of AoA error (Ground truth - Predicted) distributions across all the 6 different environments FedWiLoc has been tested against