Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Online Reliable Anomaly Detection via Neuromorphic Sensing and Communications

Junya Shiraishi, Jiechen Chen, Osvaldo Simeone, Petar Popovski

TL;DR

This work tackles online anomaly detection in neuromorphic sensor networks by combining a pull-based querying scheme with online e-value statistics to guarantee decaying-memory false discovery rate control, without requiring prior knowledge of anomaly rates. It integrates a best-arm identification–based scheduling policy (track-and-stop) to adaptively select informative sensors under uplink constraints, improving detection latency and reducing communications. The framework demonstrates reliable anomaly detection with FDR guarantees and efficient sensor scheduling across varying system parameters, channel conditions, and anomaly priors. Practically, this approach enables low-power, real-time monitoring in domains like brain–machine interfaces and remote environmental sensing while maintaining trustworthy alarms.

Abstract

This paper proposes a low-power online anomaly detection framework based on neuromorphic wireless sensor networks, encompassing possible use cases such as brain-machine interfaces and remote environmental monitoring. In the considered system, a central reader node actively queries a subset of neuromorphic sensor nodes (neuro-SNs) at each time frame. The neuromorphic sensors are event-driven, producing spikes in correspondence to relevant changes in the monitored system. The queried neuro-SNs respond to the reader with impulse radio (IR) transmissions that directly encode the sensed local events. The reader processes these event-driven signals to determine whether the monitored environment is in a normal or anomalous state, while rigorously controlling the false discovery rate (FDR) of detections below a predefined threshold. The proposed approach employs an online hypothesis testing method with e-values to maintain FDR control without requiring knowledge of the anomaly rate, and it dynamically optimizes the sensor querying strategy by casting it as a best-arm identification problem in a multi-armed bandit framework. Extensive performance evaluation demonstrates that the proposed method can reliably detect anomalies under stringent FDR requirements, while efficiently scheduling sensor communications and achieving low detection latency.

Online Reliable Anomaly Detection via Neuromorphic Sensing and Communications

TL;DR

This work tackles online anomaly detection in neuromorphic sensor networks by combining a pull-based querying scheme with online e-value statistics to guarantee decaying-memory false discovery rate control, without requiring prior knowledge of anomaly rates. It integrates a best-arm identification–based scheduling policy (track-and-stop) to adaptively select informative sensors under uplink constraints, improving detection latency and reducing communications. The framework demonstrates reliable anomaly detection with FDR guarantees and efficient sensor scheduling across varying system parameters, channel conditions, and anomaly priors. Practically, this approach enables low-power, real-time monitoring in domains like brain–machine interfaces and remote environmental sensing while maintaining trustworthy alarms.

Abstract

This paper proposes a low-power online anomaly detection framework based on neuromorphic wireless sensor networks, encompassing possible use cases such as brain-machine interfaces and remote environmental monitoring. In the considered system, a central reader node actively queries a subset of neuromorphic sensor nodes (neuro-SNs) at each time frame. The neuromorphic sensors are event-driven, producing spikes in correspondence to relevant changes in the monitored system. The queried neuro-SNs respond to the reader with impulse radio (IR) transmissions that directly encode the sensed local events. The reader processes these event-driven signals to determine whether the monitored environment is in a normal or anomalous state, while rigorously controlling the false discovery rate (FDR) of detections below a predefined threshold. The proposed approach employs an online hypothesis testing method with e-values to maintain FDR control without requiring knowledge of the anomaly rate, and it dynamically optimizes the sensor querying strategy by casting it as a best-arm identification problem in a multi-armed bandit framework. Extensive performance evaluation demonstrates that the proposed method can reliably detect anomalies under stringent FDR requirements, while efficiently scheduling sensor communications and achieving low detection latency.
Paper Structure (27 sections, 1 theorem, 30 equations, 12 figures, 1 table)

This paper contains 27 sections, 1 theorem, 30 equations, 12 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Main Contributions
  4. System model
  5. Setting
  6. Observation Models
  7. Scheduling and Impulse Radio Transmission
  8. Online Anomaly Detection
  9. Hypothesis Testing Framework
  10. Decaying Memory Time-Averaged False Discovery Rate
  11. Decaying Memory Time-Averaged True Discovery Rate
  12. Construction of E-Values
  13. Design of the Rejection Thresholds
  14. Scheduling Policy Optimization
  15. Multi-Armed Bandit Formulation
  16. ...and 12 more sections

Key Result

Proposition 1

When the decision threshold $\alpha_f$ is updated according to eq:alpha_update_rule at each frame $f$, the anomaly detection rule in eq:test_estimation ensures that the decaying-memory time-averaged FDR defined in fdrl is controlled at level $\alpha$ for all frames $f$, thus satisfying condition eq:

Figures (12)

  • Figure 1: In the setup under study, a reader interrogates a subset $\mathcal{N}_f$ of the $K$ neuromorphic sensor nodes (neuro-SN) at each frame $f$ in order to determine whether the behavior of the monitored system is normal or anomalous. The binary variable $S_f$ indicates the presence or absence of an anomaly at frame $f$, while $\hat{S}_f$ represents the corresponding estimate of the reader.
  • Figure 2: Frame structure illustrating the query transmission from the reader (downlink), sensing and impulse radio (IR) transmission by the selected neuro-SN during $L$ uplink slots, and anomaly detection at the reader.
  • Figure 3: Each $k$-th sensor is characterized by a known spiking probability $q^{k}_{0}$ in normal conditions, and a higher, unknown spiking probability $q^{k}_{1}\geq q^{k}_{0}$ in the presence of an anomaly.
  • Figure 4: Illustration of a binary asymmetric channel.
  • Figure 5: and versus the time frame $f$ for fixed anomaly decision threshold $\alpha_f = \alpha$ for all $f \in [1,F]$ and for the proposed dynamic anomaly decision threshold based on \ref{['eq:alpha_update_rule']}, where we set $K =1$ and $\pi_1 = 0.05$.
  • ...and 7 more figures

Theorems & Definitions (1)

  • Proposition 1: rebjock2021online