[Paper Review] Exact Analysis of k-Connectivity in Secure Sensor Networks with Unreliable Links

Jun Zhao, Osman Yağan+1 more|arXiv (Cornell University)|Sep 21, 2014

Security in Wireless Sensor Networks33 references3 citations

TL;DR

This paper provides the asymptotically exact probability of $k$-connectivity in secure wireless sensor networks using the Eschenauer–Gligor key predistribution scheme with unreliable on/off links. By modeling links as independent channels and deriving a precise analytical expression, it enables accurate design of networks resilient to $k-1$ node or link failures, filling a critical gap beyond zero-one laws.

ABSTRACT

The Eschenauer--Gligor (EG) random key predistribution scheme has been widely recognized as a typical approach to secure communications in wireless sensor networks (WSNs). However, there is a lack of precise probability analysis on the reliable connectivity of WSNs under the EG scheme. To address this, we rigorously derive the asymptotically exact probability of $k$-connectivity in WSNs employing the EG scheme with unreliable links represented by independent on/off channels, where $k$-connectivity ensures that the network remains connected despite the failure of any $(k-1)$ sensors or links. Our analytical results are confirmed via numerical experiments, and they provide precise guidelines for the design of secure WSNs that exhibit a desired level of reliability against node and link failures.

Motivation & Objective

• Address the lack of precise probability analysis for $k$-connectivity in secure wireless sensor networks (WSNs) under the Eschenauer–Gligor (EG) key predistribution scheme.
• Overcome the limitations of zero-one laws, which only indicate high-probability or high-unlikelihood of $k$-connectivity without quantifying the exact probability.
• Provide a complete and analytically precise understanding of how network parameters affect $k$-connectivity, especially in the presence of unreliable links.
• Enable network designers to achieve desired reliability levels against node and link failures in hostile or unattended WSN deployments.
• Quantify the sensitivity of $k$-connectivity to parameter variations by determining the width of the phase transition in connectivity behavior.

Proposed method

• Model the secure WSN as the intersection of a random key graph $G(n, K_n, P_n)$ and an Erdős–Rényi random graph $G(n, p_n)$, representing key sharing and unreliable links, respectively.
• Derive the asymptotically exact probability of $k$-connectivity using advanced random graph theory, focusing on the regime where $\frac{K_n^2}{P_n} = \frac{\ln n + (k-1)\ln\ln n + \alpha_n}{n}$ and $p_n$ is a function of $n$.
• Establish a closed-form expression for the limiting probability of $k$-connectivity as $e^{-\frac{e^{-\alpha^*}}{(k-1)!}}$ when $\lim_{n\to\infty} \alpha_n = \alpha^*$, where $\alpha^*$ is a finite constant.
• Leverage tools from random intersection graphs and Poisson approximation techniques to analyze the joint effect of key predistribution and link unreliability.
• Confirm the analytical results via extensive numerical experiments, validating the accuracy of the derived expression across varying network parameters.
• Extend prior zero-one laws to a one-law framework, providing a continuous and precise characterization of $k$-connectivity probability across all parameter regimes.

Experimental results

Research questions

• RQ1 What is the exact asymptotic probability that a secure WSN using the EG scheme remains $k$-connected when links are unreliable and fail independently?
• RQ2 How does the probability of $k$-connectivity vary with the key pool size $P_n$, number of keys per node $K_n$, and link reliability $p_n$?
• RQ3 Can the phase transition width in $k$-connectivity be quantified, and how sensitive is the connectivity probability to small changes in $K_n$, $P_n$, or $p_n$?
• RQ4 How does the proposed exact probability expression improve upon existing zero-one laws in guiding the design of reliable and secure WSNs?
• RQ5 To what extent does the interplay between key predistribution and random link failures affect the robustness of $k$-connectivity in large-scale WSNs?

Key findings

• The asymptotically exact probability that the network is $k$-connected converges to $e^{-\frac{e^{-\alpha^*}}{(k-1)!}}$ when $\lim_{n\to\infty} \alpha_n = \alpha^*$, where $\alpha_n = \frac{K_n^2}{P_n} \cdot n - \ln n - (k-1)\ln\ln n$.
• This result provides a continuous and precise characterization of $k$-connectivity, enabling accurate network design for desired reliability levels.
• The width of the phase transition in $k$-connectivity is quantified by the sensitivity of the probability to changes in $\alpha_n$, which is now analytically captured.
• Numerical experiments confirm the analytical predictions, showing strong agreement between theory and simulation across different values of $n$, $K_n$, $P_n$, and $p_n$.
• The derived expression generalizes the Erdős–Rényi $k$-connectivity result to the intersection of random key and random link graphs, establishing a novel connection between two major random graph models.
• The work resolves a key gap in the literature by providing the first asymptotically exact probability for $k$-connectivity in secure WSNs with unreliable links, surpassing previous zero-one laws in utility and precision.