Federated Quantum Kernel Learning for Anomaly Detection in Multivariate IoT Time-Series

TL;DR

Federated Quantum Kernel Learning for Anomaly Detection in Multivariate IoT Time-Series addresses privacy and communication constraints in IIoT anomaly detection by combining quantum kernel embeddings with federated aggregation to build a global Gram matrix $K_{ ext{global}}$ and train a quantum-enhanced classifier. The approach decentralizes quantum kernel computations to edge devices and centrally optimizes using a classical kernel method, demonstrating competitive or superior generalization to complex temporal correlations while reducing data exchange compared with classical federated baselines. The work contributes a concrete FQKL framework, synthetic benchmarking on parity-based non-linear dependencies, and analysis of scalability across clients and data sizes, highlighting the potential of quantum kernels in distributed IoT settings. Together, these results indicate a viable path toward scalable, privacy-preserving quantum-classical pipelines for robust anomaly detection in heterogeneous industrial networks.

Abstract

The rapid growth of industrial Internet of Things (IIoT) systems has created new challenges for anomaly detection in high-dimensional, multivariate time-series, where privacy, scalability, and communication efficiency are critical. Classical federated learning approaches mitigate privacy concerns by enabling decentralized training, but they often struggle with highly non-linear decision boundaries and imbalanced anomaly distributions. To address this gap, we propose a Federated Quantum Kernel Learning (FQKL) framework that integrates quantum feature maps with federated aggregation to enable distributed, privacy-preserving anomaly detection across heterogeneous IoT networks. In our design, quantum edge nodes locally compute compressed kernel statistics using parameterized quantum circuits and share only these summaries with a central server, which constructs a global Gram matrix and trains a decision function (e.g., Fed-QSVM). Experimental results on synthetic IIoT benchmarks demonstrate that FQKL achieves superior generalization in capturing complex temporal correlations compared to classical federated baselines, while significantly reducing communication overhead. This work highlights the promise of quantum kernels in federated settings, advancing the path toward scalable, robust, and quantum-enhanced intelligence for next-generation IoT infrastructures.

Figures (5)

• Figure 1: Proposed architecture of Federated Quantum Kernel Learning for anomaly detection in multivariate IoT time-series. Quantum edge nodes compute local kernel statistics using a fixed quantum feature map and share only compressed results with the Quantum Federated Server, which aggregates them into a global Gram matrix to train a decision function (e.g., Fed-QSVM). The optimized global model is then disseminated back to all edge nodes, enabling privacy-preserving and scalable anomaly detection across heterogeneous IoT networks.
• Figure 2: Multivariate IoT time-series from 8 sensors (indexed 0–7). Normal patterns are shown in blue, while anomaly intervals are highlighted in red. The aligned subplot view emphasizes temporal correlations across sensors.
• Figure 3: Illustration of the synthetic IIoT dataset. (a) Example raw sensor waveforms with shaded anomaly intervals. (b) Phase-change indicator matrix, where binary values capture the up/down state of selected sensors over time. (c) Comparison between the derived parity (XOR of selected phase-change indicators) and the ground-truth anomaly labels. Anomalies are not visible as simple amplitude outliers but instead emerge from distributed high-order correlations, motivating the use of quantum-enhanced kernels in federated IIoT anomaly detection.
• Figure 4: Benchmarking of federated quantum SVM (Fed-QSVM) against classical baselines (Fed-SVM, Fed-RF). (a) Fed-QSVM sustains accuracy under increasing parity order, unlike classical models that degrade with higher feature interactions. (b) Fed-QSVM remains robust with growing client numbers, whereas Fed-RF shows higher variance and Fed-SVM degrades moderately. (c) While requiring more samples in the low-data regime, Fed-QSVM surpasses classical baselines with larger datasets, demonstrating superior scalability in distributed IoT anomaly detection.
• Figure 5: Communication–accuracy trade-off in federated learning. (a) Parity scenario: quantum kernels offer clear communication efficiency over classical baselines. (b) Periodic scenario: all models achieve high accuracy with overlapping communication ranges.