Online Fault-Tolerant Dynamic Event Region Detection in Sensor Networks via Trust Model

Jiejie Wang, Bin Liu

Abstract

This paper proposes a Bayesian modeling approach to address the problem of online fault-tolerant dynamic event region detection in wireless sensor networks. In our model every network node is associated with a virtual community and a trust index, which quantitatively measures the trustworthiness of this node in its community. If a sensor node's trust value is smaller than a threshold, it suggests that this node encounters a fault and thus its sensor reading can not be trusted at this moment. This concept of sensor node trust discriminates our model with the other alternatives, e.g., the Markov random fields. The practical issues, including spatiotemporal correlations of neighbor nodes' sensor readings, the presence of sensor faults and the requirement of online processing are linked together by the concept trust and are all taken into account in the modeling stage. Based on the proposed model, the trust value of each node is updated online by a particle filter algorithm upon the arrival of new observations. The decision on whether a node is located in the event region is made based upon the current estimate of this node's trust value. Experimental results demonstrate that the proposed solution can provide striking better performance than existent methods in terms of error rate in detecting the event region.

Introduction

In Wireless Sensor Networks (WSNs), Event Region Detection (ERD) is critical for identifying areas affected by phenomena such as forest fires, gas leaks, or structural failures. However, two primary challenges hinder effective detection: the dynamic nature of real-world events, which move and spread over time, and the inherent unreliability of low-cost sensors deployed in harsh environments. Faulty sensors often produce false alarms or missed detections, complicating the distinction between a true environmental event and a localized hardware failure. Existing solutions frequently struggle to handle both moving boundaries and high sensor failure rates in an online, real-time fashion.

Methodology: The Bayesian Trust Model (BTM)

The proposed approach utilizes a decentralized Bayesian Trust Model (BTM) where each node belongs to a "virtual community" defined by its immediate spatial neighbors. The core of this model is the Trust Index ($\lambda$), a value between 0 and 1 that represents a node's reliability.

To estimate this hidden trust state from noisy observations, the system employs Particle Filtering (a sequential Monte Carlo method). Each node runs its own filter, making the solution distributed and scalable. The trust index is updated based on a voting mechanism: a node's reading is compared to its neighbors, and if the difference falls within a permissible threshold ($r$), the neighbors "vote" in favor of the node's reliability.

Final decision logic is bifurcated based on the trust estimate:

  1. High Trust: The node's raw data is used directly to determine event status.
  2. Low Trust: The system assumes a fault and instead uses a "trust-averaged" estimate derived from the node's neighbors to infer the event status at that location.

Experimental Evaluation

The model was evaluated using synthetic datasets simulating a 10x10 meter field with 100 uniformly distributed nodes. The simulation modeled a "diffuse event" that spreads outward from the center over five discrete time steps ($k=1$ to $k=5$).

Evolution of the dynamic event region over five time steps (k=1 to k=5).

To test robustness, the authors simulated four common sensor failure types: offset faults, stuck-at faults, variance degradation, and sleeper attacks.

Results and Performance Analysis

The BTM was compared against two baseline detectors: the traditional CUSUM (Cumulative Sum) and a Hybrid model combining CUSUM with Markov Random Fields.

As shown in the performance data, the proposed BTM significantly outperforms the alternatives. In scenarios with a 10% failure rate, the BTM maintained an average error rate between 0 and 0.01, while CUSUM and Hybrid models showed significantly higher error rates (up to 0.06 and 0.02, respectively).

Average Error Rate comparison across different time steps (k) for CUSUM, Hybrid, and the Proposed BTM.

Key findings include:

  • Robustness: The BTM's error rate grew much more slowly than baselines as sensor failure rates increased toward 40%.
  • Boundary Accuracy: The model was particularly effective at accurately identifying "boundary nodes" where data is most ambiguous.
  • Efficiency: By using only 20 particles for the 1D trust variable, the model remains computationally feasible for resource-constrained sensor hardware.

Conclusion and Future Directions

The Bayesian Trust Model provides a robust framework for decoupling sensor faults from environmental events through social-analogy consensus. While effective, future work is needed to address potential communication issues like packet loss, adapt the model for mobile sensor topologies, and extend the binary classification to multi-class monitoring for varying event intensities.

Summary updated

Abstract

This paper proposes a Bayesian modeling approach to address the problem of online fault-tolerant dynamic event region detection in wireless sensor networks. In our model every network node is associated with a virtual community and a trust index, which quantitatively measures the trustworthiness of this node in its community. If a sensor node's trust value is smaller than a threshold, it suggests that this node encounters a fault and thus its sensor reading can not be trusted at this moment. This concept of sensor node trust discriminates our model with the other alternatives, e.g.,the Markov random fields. The practical issues, including spatiotemporal correlations of neighbor nodes' sensor readings, the presence of sensor faults and the requirement of online processing are linked together by the concept trust and are all taken into account in the modeling stage. Based on the proposed model, the trust value of each node is updated online by a particle filter algorithm upon the arrival of new observations. The decision on whether a node is located in the event region is made based upon the current estimate of this node's trust value. Experimental results demonstrate that the proposed solution can provide striking better performance than existent methods in terms of error rate in detecting the event region.
BibTeX
@inproceedings{wang2017online,
  author = {Jiejie Wang and Bin Liu},
  title = {{Online Fault-Tolerant Dynamic Event Region Detection in Sensor Networks via Trust Model}},
  booktitle = {WCNC},
  year = {2017},
  doi = {10.1109/wcnc.2017.7925627},
}

Loading related content...

Figure