Parameterized Critical Node Cut Revisited
Abstract
Given a graph $G$ and integers $k, x \geq 0$, the Critical Node Cut problem asks whether it is possible to delete at most $k$ vertices from $G$ such that the number of remaining pairs of connected vertices is at most $x$. This problem generalizes Vertex Cover (when $x = 0$), and has applications in network design, epidemiology, and social network analysis. We investigate the parameterized complexity of Critical Node Cut under various structural parameters. We first significantly strengthen existing hardness results by proving W[1]-hardness even when parameterized by the combined parameter $k + \mathrm{fes} + \Delta + \mathrm{pw}$, where $\mathrm{fes}$ is the feedback edge set number, $\Delta$ the maximum degree, and $\mathrm{pw}$ the pathwidth of the input graph. We then identify three structural parameters--max-leaf number, vertex integrity, and modular-width--that render the problem fixed-parameter tractable. Furthermore, leveraging a technique introduced by Lampis [ICALP '14], we develop an FPT approximation scheme that, for any $\varepsilon > 0$, computes a $(1+\varepsilon)$-approximate solution in time $(\mathrm{tw} / \varepsilon)^{\mathcal{O}(\mathrm{tw})} n^{\mathcal{O}(1)}$, where $\mathrm{tw}$ denotes the treewidth of the input graph. Finally, we show that Critical Node Cut does not admit a polynomial kernel when parameterized by vertex cover number, unless standard complexity assumptions fail. Overall, our results significantly sharpen the known complexity landscape of Critical Node Cut.