Combining Crown Structures for Vulnerability Measures

TL;DR

This paper develops new kernelization algorithms for graph vulnerability measures, combining crown decompositions to improve kernel sizes for vertex integrity and component order connectivity, with applications to vertex cover and claw-free graphs.

Contribution

It introduces novel kernelization algorithms that leverage combined crown decompositions, extending existing techniques and improving kernel bounds for key vulnerability measures.

Findings

Vertex kernel for VI improved from p^3 to p^2

Kernel for wVI improved from p^3 to 3(p^2 + p^{1.5} p_{ ext{ell}})

Kernel for wCOC improved from O(k^2W + kW^2) to 3μ(k + √μW)

Abstract

Over the past decades, various metrics have emerged in graph theory to grasp the complex nature of network vulnerability. In this paper, we study two specific measures: (weighted) vertex integrity (wVI) and (weighted) component order connectivity (wCOC). These measures not only evaluate the number of vertices required to decompose a graph into fragments, but also take into account the size of the largest remaining component. The main focus of our paper is on kernelization algorithms tailored to both measures. We capitalize on the structural attributes inherent in different crown decompositions, strategically combining them to introduce novel kernelization algorithms that advance the current state of the field. In particular, we extend the scope of the balanced crown decomposition provided by Casel et al.~[7] and expand the applicability of crown decomposition techniques. In summary, we improve the vertex kernel of VI from $p^3$ to $p^2$, and of wVI from $p^3$ to $3(p^2 + p^{1.5} p_{\ell})$, where $p_{\ell} < p$ represents the weight of the heaviest component after removing a solution. For wCOC we improve the vertex kernel from $\mathcal{O}(k^2W + kW^2)$ to $3\mu(k + \sqrt{\mu}W)$, where $\mu = \max(k,W)$. We also give a combinatorial algorithm that provides a $2kW$ vertex kernel in FPT-runtime when parameterized by $r$, where $r \leq k$ is the size of a maximum $(W+1)$-packing. We further show that the algorithm computing the $2kW$ vertex kernel for COC can be transformed into a polynomial algorithm for two special cases, namely when $W=1$, which corresponds to the well-known vertex cover problem, and for claw-free graphs. In particular, we show a new way to obtain a $2k$ vertex kernel (or to obtain a 2-approximation) for the vertex cover problem by only using crown structures.