Tight bounds on the performance of dynamic directed cutset data structures based on OMv

Efficient data structures for computing the cutset of a set of nodes in a graph undergoing dynamic edge insertions/deletions are well-studied in undirected graphs. We study this problem in directed graphs and show a reduction from the Online Boolean Matrix-Vector Multiplication Conjecture (OMv) introduced by Henzinger et al. [STOC’15]. We prove conditional on OMv that a dynamic data structure computing the directed cutset of a set of nodes cannot have an amortized time of both O(n2−ε) for a query operation and O(n1−ε) for an update operation, for any constant ε > 0, even when an adversary’s operations are restricted to the incremental or decremental settings. We further give algorithms to match these lower bounds.