۱- Introduction

۲- Traffic metrics

۳- Cascading model

۴- Analysis on the invulnerability of WSNs

۵- Conclusions

References

Abstract

Existing cascading models for wireless sensor networks (WSNs) cannot correctly reflect the traffic feature of WSNs. In this work, we build a more practical cascading model for WSNs, in which the network load is defined according to two new traffic metrics “sink-oriented node betweenness” and “sink-oriented link betweenness” and the cascading process is jointly propelled by the load redistribution of sensor nodes and wireless links. In addition, load-redistribution schemes are designed according to the principle of “idle capacity”. Simulation results show that the network invulnerability is positively related to the tolerance coefficient and negatively related to the exponential coefficient. The minimum costs needed to resist intentional node attacks are more expensive than the costs needed when facing intentional link attacks.

Introduction

In recent years, owing to the significance of network safety in our daily life, network invulnerability has attracted a large amount of interests from many researchers [1–۳]. As an important part of the Internet of Things (IoTs), Wireless Sensor Networks (WSNs) have also received widespread attention about the network invulnerability because of their unattended deployment environment and vulnerability to node/link failures [4,5]. Since the earlier studies mainly focus on the static invulnerability from a topological perspective, recently cascading failures induced by the dynamic load redistribution in WSNs have been significantly concerned and widely investigated [6–۱۰]. In existing works related to the cascading invulnerability of WSNs, they usually assumed that each sensor node takes a certain degree of traffic load due to data delivery tasks [11–۱۵]. The traffic load is usually represented by degree or betweenness values. Due to the limited hardware costs, each sensor node can only have limited capacity to tackle its own load. If the real-time load is beyond its capacity, the sensor node is highly likely to fail due to buffer overflows or channel congestions. When a sensor node fails, those nodes who transmit data through it will choose a new path to accomplish the data delivery, further leading to the redistribution of the network load. We can easily discover that in existing cascading models the load-redistribution process can only spread from node to node and can only be triggered by node failures.