Self-Supervised Federated GNSS Spoofing Detection with Opportunistic Data

Abstract: Global navigation satellite systems (GNSS) are vulnerable to spoofing attacks, with adversarial signals manipulating the location or time information of receivers, potentially causing severe disruptions. The task of discerning the spoofing signals from benign ones is naturally relevant for machine learning, thus recent interest in applying it for detection. While deep learning-based methods are promising, they require extensive labeled datasets, consume significant computational resources, and raise privacy concerns due to the sensitive nature of position data. This is why this paper proposes a self-supervised federated learning framework for GNSS spoofing detection. It consists of a cloud server and local mobile platforms. Each mobile platform employs a self-supervised anomaly detector using long short-term memory (LSTM) networks. Labels for training are generated locally through a spoofing-deviation prediction algorithm, ensuring privacy. Local models are trained independently, and only their parameters are uploaded to the cloud server, which aggregates them into a global model using FedAvg. The updated global model is then distributed back to the mobile platforms and trained iteratively. The evaluation shows that our self-supervised federated learning framework outperforms position-based and deep learning-based methods in detecting spoofing attacks while preserving data privacy.

Evaluation of Self-Supervised Federated GNSS Spoofing Detection

The paper "Self-supervised federated GNSS spoofing detection with opportunistic data" addresses the critical challenge of GNSS spoofing, which poses significant risks to secure localization systems. This threat undermines the accuracy of positioning information crucial for various applications. The authors propose a self-supervised federated learning framework, a novel approach to effectively detect GNSS spoofing attacks while enhancing privacy preservation.

Core Proposal

Central to this paper is the establishment of a federated learning architecture designed to identify GNSS spoofing with self-supervised anomaly detection models implemented on distributed mobile platforms. Each platform utilizes LSTM networks trained on local datasets comprising GNSS signals and auxiliary opportunistic data. This decentralized model training strategy ensures privacy by eliminating the need to transfer sensitive positional data to a central server. Instead, only model parameters are aggregated using the FedAvg algorithm to construct a global detection model.

Key Results

The experimental results demonstrated promising efficacy, with the federated learning approach achieving superior detection performance compared to traditional position-based and centralized deep learning methods. Specifically, the federated model yielded an AUC of 87.4%, surpassing position-based detection methods with an AUC of 83.5% and centralized deep learning with an AUC of 86.6%.

Data and Evaluation

Using real-world GNSS spoofing data from Jammertest 2024, the framework was rigorously tested under various conditions, including same-device testing, same-model training, and cross-model generalization, involving both IID and non-IID data scenarios. Such comprehensive evaluations underscored the robustness and adaptability of the detection mechanism across different smartphone models, demonstrating a capacity to generalize effectively even across distinct datasets.

Implications and Future Work

This research paves the way for significant advancements in secure localization techniques, particularly where privacy considerations prevent extensive centralized data management. The integration of self-supervised learning resolves the dependency on labeled data, a common limitation in deep learning applications. Future directions should explore real-world deployment complexities, such as communication-induced network delays and collaborative adversarial models, ensuring further refinements in model security and performance resilience.

The paper's methodological innovations and empirical validations present a robust framework for improving GNSS security, reflecting substantial progress in federated learning applications within cybersecurity and privacy-aware settings. Continued exploration in this domain promises further enhancements in spoofing detection accuracy and infrastructure adaptability.